Sphinx 2.0.4-release reference manual

Free open-source SQL full-text search engine

Table of Contents

1. Introduction
1.1. About
1.2. Sphinx features
1.3. Where to get Sphinx
1.4. License
1.5. Credits
1.6. History
2. Installation
2.1. Supported systems
2.2. Required tools
2.3. Installing Sphinx on Linux
2.4. Installing Sphinx on Windows
2.5. Known installation issues
2.6. Quick Sphinx usage tour
3. Indexing
3.1. Data sources
3.2. Full-text fields
3.3. Attributes
3.4. MVA (multi-valued attributes)
3.5. Indexes
3.6. Restrictions on the source data
3.7. Charsets, case folding, and translation tables
3.8. SQL data sources (MySQL, PostgreSQL)
3.9. xmlpipe data source
3.10. xmlpipe2 data source
3.11. Live index updates
3.12. Delta index updates
3.13. Index merging
4. Real-time indexes
4.1. RT indexes overview
4.2. Known caveats with RT indexes
4.3. RT index internals
4.4. Binary logging
5. Searching
5.1. Matching modes
5.2. Boolean query syntax
5.3. Extended query syntax
5.4. Search results ranking
5.5. Expressions, functions, and operators
5.5.1. Operators
5.5.2. Numeric functions
5.5.3. Date and time functions
5.5.4. Type conversion functions
5.5.5. Comparison functions
5.5.6. Miscellaneous functions
5.6. Sorting modes
5.7. Grouping (clustering) search results
5.8. Distributed searching
5.9. searchd query log formats
5.9.1. Plain log format
5.9.2. SphinxQL log format
5.10. MySQL protocol support and SphinxQL
5.11. Multi-queries
5.12. Collations
5.13. User-defined functions (UDF)
6. Command line tools reference
6.1. indexer command reference
6.2. searchd command reference
6.3. search command reference
6.4. spelldump command reference
6.5. indextool command reference
7. SphinxQL reference
7.1. SELECT syntax
7.2. SHOW META syntax
7.3. SHOW WARNINGS syntax
7.4. SHOW STATUS syntax
7.5. INSERT and REPLACE syntax
7.6. DELETE syntax
7.7. SET syntax
7.9. BEGIN, COMMIT, and ROLLBACK syntax
7.10. CALL SNIPPETS syntax
7.11. CALL KEYWORDS syntax
7.12. SHOW TABLES syntax
7.13. DESCRIBE syntax
7.14. CREATE FUNCTION syntax
7.15. DROP FUNCTION syntax
7.16. SHOW VARIABLES syntax
7.17. SHOW COLLATION syntax
7.18. UPDATE syntax
7.19. ATTACH INDEX syntax
7.20. FLUSH RTINDEX syntax
7.21. Multi-statement queries
7.22. Comment syntax
7.23. List of SphinxQL reserved keywords
7.24. SphinxQL upgrade notes, version 2.0.1-beta
8. API reference
8.1. General API functions
8.1.1. GetLastError
8.1.2. GetLastWarning
8.1.3. SetServer
8.1.4. SetRetries
8.1.5. SetConnectTimeout
8.1.6. SetArrayResult
8.1.7. IsConnectError
8.2. General query settings
8.2.1. SetLimits
8.2.2. SetMaxQueryTime
8.2.3. SetOverride
8.2.4. SetSelect
8.3. Full-text search query settings
8.3.1. SetMatchMode
8.3.2. SetRankingMode
8.3.3. SetSortMode
8.3.4. SetWeights
8.3.5. SetFieldWeights
8.3.6. SetIndexWeights
8.4. Result set filtering settings
8.4.1. SetIDRange
8.4.2. SetFilter
8.4.3. SetFilterRange
8.4.4. SetFilterFloatRange
8.4.5. SetGeoAnchor
8.5. GROUP BY settings
8.5.1. SetGroupBy
8.5.2. SetGroupDistinct
8.6. Querying
8.6.1. Query
8.6.2. AddQuery
8.6.3. RunQueries
8.6.4. ResetFilters
8.6.5. ResetGroupBy
8.7. Additional functionality
8.7.1. BuildExcerpts
8.7.2. UpdateAttributes
8.7.3. BuildKeywords
8.7.4. EscapeString
8.7.5. Status
8.7.6. FlushAttributes
8.8. Persistent connections
8.8.1. Open
8.8.2. Close
9. MySQL storage engine (SphinxSE)
9.1. SphinxSE overview
9.2. Installing SphinxSE
9.2.1. Compiling MySQL 5.0.x with SphinxSE
9.2.2. Compiling MySQL 5.1.x with SphinxSE
9.2.3. Checking SphinxSE installation
9.3. Using SphinxSE
9.4. Building snippets (excerpts) via MySQL
10. Reporting bugs
11. sphinx.conf options reference
11.1. Data source configuration options
11.1.1. type
11.1.2. sql_host
11.1.3. sql_port
11.1.4. sql_user
11.1.5. sql_pass
11.1.6. sql_db
11.1.7. sql_sock
11.1.8. mysql_connect_flags
11.1.9. mysql_ssl_cert, mysql_ssl_key, mysql_ssl_ca
11.1.10. odbc_dsn
11.1.11. sql_query_pre
11.1.12. sql_query
11.1.13. sql_joined_field
11.1.14. sql_query_range
11.1.15. sql_range_step
11.1.16. sql_query_killlist
11.1.17. sql_attr_uint
11.1.18. sql_attr_bool
11.1.19. sql_attr_bigint
11.1.20. sql_attr_timestamp
11.1.21. sql_attr_str2ordinal
11.1.22. sql_attr_float
11.1.23. sql_attr_multi
11.1.24. sql_attr_string
11.1.25. sql_attr_str2wordcount
11.1.26. sql_column_buffers
11.1.27. sql_field_string
11.1.28. sql_field_str2wordcount
11.1.29. sql_file_field
11.1.30. sql_query_post
11.1.31. sql_query_post_index
11.1.32. sql_ranged_throttle
11.1.33. sql_query_info
11.1.34. xmlpipe_command
11.1.35. xmlpipe_field
11.1.36. xmlpipe_field_string
11.1.37. xmlpipe_field_wordcount
11.1.38. xmlpipe_attr_uint
11.1.39. xmlpipe_attr_bigint
11.1.40. xmlpipe_attr_bool
11.1.41. xmlpipe_attr_timestamp
11.1.42. xmlpipe_attr_str2ordinal
11.1.43. xmlpipe_attr_float
11.1.44. xmlpipe_attr_multi
11.1.45. xmlpipe_attr_multi_64
11.1.46. xmlpipe_attr_string
11.1.47. xmlpipe_fixup_utf8
11.1.48. mssql_winauth
11.1.49. mssql_unicode
11.1.50. unpack_zlib
11.1.51. unpack_mysqlcompress
11.1.52. unpack_mysqlcompress_maxsize
11.2. Index configuration options
11.2.1. type
11.2.2. source
11.2.3. path
11.2.4. docinfo
11.2.5. mlock
11.2.6. morphology
11.2.7. dict
11.2.8. index_sp
11.2.9. index_zones
11.2.10. min_stemming_len
11.2.11. stopwords
11.2.12. wordforms
11.2.13. exceptions
11.2.14. min_word_len
11.2.15. charset_type
11.2.16. charset_table
11.2.17. ignore_chars
11.2.18. min_prefix_len
11.2.19. min_infix_len
11.2.20. prefix_fields
11.2.21. infix_fields
11.2.22. enable_star
11.2.23. ngram_len
11.2.24. ngram_chars
11.2.25. phrase_boundary
11.2.26. phrase_boundary_step
11.2.27. html_strip
11.2.28. html_index_attrs
11.2.29. html_remove_elements
11.2.30. local
11.2.31. agent
11.2.32. agent_blackhole
11.2.33. agent_connect_timeout
11.2.34. agent_query_timeout
11.2.35. preopen
11.2.36. ondisk_dict
11.2.37. inplace_enable
11.2.38. inplace_hit_gap
11.2.39. inplace_docinfo_gap
11.2.40. inplace_reloc_factor
11.2.41. inplace_write_factor
11.2.42. index_exact_words
11.2.43. overshort_step
11.2.44. stopword_step
11.2.45. hitless_words
11.2.46. expand_keywords
11.2.47. blend_chars
11.2.48. blend_mode
11.2.49. rt_mem_limit
11.2.50. rt_field
11.2.51. rt_attr_uint
11.2.52. rt_attr_bigint
11.2.53. rt_attr_float
11.2.54. rt_attr_multi
11.2.55. rt_attr_multi_64
11.2.56. rt_attr_timestamp
11.2.57. rt_attr_string
11.3. indexer program configuration options
11.3.1. mem_limit
11.3.2. max_iops
11.3.3. max_iosize
11.3.4. max_xmlpipe2_field
11.3.5. write_buffer
11.3.6. max_file_field_buffer
11.3.7. on_file_field_error
11.4. searchd program configuration options
11.4.1. listen
11.4.2. address
11.4.3. port
11.4.4. log
11.4.5. query_log
11.4.6. query_log_format
11.4.7. read_timeout
11.4.8. client_timeout
11.4.9. max_children
11.4.10. pid_file
11.4.11. max_matches
11.4.12. seamless_rotate
11.4.13. preopen_indexes
11.4.14. unlink_old
11.4.15. attr_flush_period
11.4.16. ondisk_dict_default
11.4.17. max_packet_size
11.4.18. mva_updates_pool
11.4.19. crash_log_path
11.4.20. max_filters
11.4.21. max_filter_values
11.4.22. listen_backlog
11.4.23. read_buffer
11.4.24. read_unhinted
11.4.25. max_batch_queries
11.4.26. subtree_docs_cache
11.4.27. subtree_hits_cache
11.4.28. workers
11.4.29. dist_threads
11.4.30. binlog_path
11.4.31. binlog_flush
11.4.32. binlog_max_log_size
11.4.33. collation_server
11.4.34. collation_libc_locale
11.4.35. plugin_dir
11.4.36. mysql_version_string
11.4.37. rt_flush_period
11.4.38. thread_stack
11.4.39. expansion_limit
11.4.40. compat_sphinxql_magics
11.4.41. watchdog
11.4.42. prefork_rotation_throttle
A. Sphinx revision history
A.1. Version 2.0.4-release, 02 mar 2012
A.2. Version 2.0.3-release, 23 dec 2011
A.3. Version 2.0.2-beta, 15 nov 2011
A.4. Version 2.0.1-beta, 22 apr 2011
A.5. Version 1.10-beta, 19 jul 2010
A.6. Version 0.9.9-release, 02 dec 2009
A.7. Version 0.9.9-rc2, 08 apr 2009
A.8. Version 0.9.9-rc1, 17 nov 2008
A.9. Version, 30 oct 2008
A.10. Version 0.9.8, 14 jul 2008
A.11. Version 0.9.7, 02 apr 2007
A.12. Version 0.9.7-rc2, 15 dec 2006
A.13. Version 0.9.7-rc1, 26 oct 2006
A.14. Version 0.9.6, 24 jul 2006
A.15. Version 0.9.6-rc1, 26 jun 2006

Chapter 1. Introduction

1.1. About

Sphinx is a full-text search engine, publicly distributed under GPL version 2. Commercial licensing (eg. for embedded use) is available upon request.

Technically, Sphinx is a standalone software package provides fast and relevant full-text search functionality to client applications. It was specially designed to integrate well with SQL databases storing the data, and to be easily accessed scripting languages. However, Sphinx does not depend on nor require any specific database to function.

Applications can access Sphinx search daemon (searchd) using any of the three different access methods: a) via native search API (SphinxAPI), b) via Sphinx own implementation of MySQL network protocol (using a small SQL subset called SphinxQL), or c) via MySQL server with a pluggable storage engine (SphinxSE).

Official native SphinxAPI implementations for PHP, Perl, Ruby, and Java are included within the distribution package. API is very lightweight so porting it to a new language is known to take a few hours or days. Third party API ports and plugins exist for Perl, C#, Haskell, Ruby-on-Rails, and possibly other languages and frameworks.

Starting version 1.10-beta, Sphinx supports two different indexing backends: "disk" index backend, and "realtime" (RT) index backend. Disk indexes support online full-text index rebuilds, but online updates can only be done on non-text (attribute) data. RT indexes additionally allow for online full-text index updates. Previous versions only supported disk indexes.

Data can be loaded into disk indexes using a so-called data source. Built-in sources can fetch data directly from MySQL, PostgreSQL, ODBC compliant database (MS SQL, Oracle, etc), or a pipe in a custom XML format. Adding new data sources drivers (eg. to natively support other DBMSes) is designed to be as easy as possible. RT indexes, as of 1.10-beta, can only be populated using SphinxQL.

As for the name, Sphinx is an acronym which is officially decoded as SQL Phrase Index. Yes, I know about CMU's Sphinx project.

1.2. Sphinx features

Key Sphinx features are:

  • high indexing and searching performance;

  • advanced indexing and querying tools (flexible and feature-rich text tokenizer, querying language, several different ranking modes, etc);

  • advanced result set post-processing (SELECT with expressions, WHERE, ORDER BY, GROUP BY etc over text search results);

  • proven scalability up to billions of documents, terabytes of data, and thousands of queries per second;

  • easy integration with SQL and XML data sources, and SphinxAPI, SphinxQL, or SphinxSE search interfaces;

  • easy scaling with distributed searches.

To expand a bit, Sphinx:

  • has high indexing speed (upto 10-15 MB/sec per core on an internal benchmark);

  • has high search speed (upto 150-250 queries/sec per core against 1,000,000 documents, 1.2 GB of data on an internal benchmark);

  • has high scalability (biggest known cluster indexes over 3,000,000,000 documents, and busiest one peaks over 50,000,000 queries/day);

  • provides good relevance ranking through combination of phrase proximity ranking and statistical (BM25) ranking;

  • provides distributed searching capabilities;

  • provides document excerpts (snippets) generation;

  • provides searching from within application with SphinxAPI or SphinxQL interfaces, and from within MySQL with pluggable SphinxSE storage engine;

  • supports boolean, phrase, word proximity and other types of queries;

  • supports multiple full-text fields per document (upto 32 by default);

  • supports multiple additional attributes per document (ie. groups, timestamps, etc);

  • supports stopwords;

  • supports morphological word forms dictionaries;

  • supports tokenizing exceptions;

  • supports both single-byte encodings and UTF-8;

  • supports stemming (stemmers for English, Russian and Czech are built-in; and stemmers for French, Spanish, Portuguese, Italian, Romanian, German, Dutch, Swedish, Norwegian, Danish, Finnish, Hungarian, are available by building third party libstemmer library);

  • supports MySQL natively (all types of tables, including MyISAM, InnoDB, NDB, Archive, etc are supported);

  • supports PostgreSQL natively;

  • supports ODBC compliant databases (MS SQL, Oracle, etc) natively;

  • ...has 50+ other features not listed here, refer to API and configuration manual!

1.3. Where to get Sphinx

Sphinx is available through its official Web site at http://sphinxsearch.com/.

Currently, Sphinx distribution tarball includes the following software:

  • indexer: an utility which creates fulltext indexes;

  • search: a simple command-line (CLI) test utility which searches through fulltext indexes;

  • searchd: a daemon which enables external software (eg. Web applications) to search through fulltext indexes;

  • sphinxapi: a set of searchd client API libraries for popular Web scripting languages (PHP, Python, Perl, Ruby).

  • spelldump: a simple command-line tool to extract the items from an ispell or MySpell (as bundled with OpenOffice) format dictionary to help customize your index, for use with wordforms.

  • indextool: an utility to dump miscellaneous debug information about the index, added in version 0.9.9-rc2.

1.4. License

This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. See COPYING file for details.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

Non-GPL licensing (for OEM/ISV embedded use) can also be arranged, please contact us to discuss commercial licensing possibilities.

1.5. Credits


Sphinx initial author (and a benevolent dictator ever since):


Past and present employees of Sphinx Technologies Inc who should be noted on their work on Sphinx (in alphabetical order):

  • Alexander Klimenko

  • Alexey Dvoichenkov

  • Alexey Vinogradov

  • Ilya Kuznetsov

  • Stanislav Klinov


People who contributed to Sphinx and their contributions (in no particular order):

  • Robert "coredev" Bengtsson (Sweden), initial version of PostgreSQL data source

  • Len Kranendonk, Perl API

  • Dmytro Shteflyuk, Ruby API

Many other people have contributed ideas, bug reports, fixes, etc. Thank you!

1.6. History

Sphinx development was started back in 2001, because I didn't manage to find an acceptable search solution (for a database driven Web site) which would meet my requirements. Actually, each and every important aspect was a problem:

  • search quality (ie. good relevance)

    • statistical ranking methods performed rather bad, especially on large collections of small documents (forums, blogs, etc)

  • search speed

    • especially if searching for phrases which contain stopwords, as in "to be or not to be"

  • moderate disk and CPU requirements when indexing

    • important in shared hosting enivronment, not to mention the indexing speed.

Despite the amount of time passed and numerous improvements made in the other solutions, there's still no solution which I personally would be eager to migrate to.

Considering that and a lot of positive feedback received from Sphinx users during last years, the obvious decision is to continue developing Sphinx (and, eventually, to take over the world).

Chapter 2. Installation

2.1. Supported systems

Most modern UNIX systems with a C++ compiler should be able to compile and run Sphinx without any modifications.

Currently known systems Sphinx has been successfully running on are:

  • Linux 2.4.x, 2.6.x (many various distributions)

  • Windows 2000, XP, 7

  • FreeBSD 4.x, 5.x, 6.x, 7.x, 8.x

  • NetBSD 1.6, 3.0

  • Solaris 9, 11

  • Mac OS X

CPU architectures known to work include i386 (aka x86), amd64 (aka x86_64), SPARC64, and ARM.

Chances are good that Sphinx should work on other Unix platforms and/or CPU architectures just as well. Please report any other platforms that worked for you!

All platforms are production quality. There are no principal functional limitations on any platform.

2.2. Required tools

On UNIX, you will need the following tools to build and install Sphinx:

  • a working C++ compiler. GNU gcc is known to work.

  • a good make program. GNU make is known to work.

On Windows, you will need Microsoft Visual C/C++ Studio .NET 2005 or above. Other compilers/environments will probably work as well, but for the time being, you will have to build makefile (or other environment specific project files) manually.

2.3. Installing Sphinx on Linux

  1. Extract everything from the distribution tarball (haven't you already?) and go to the sphinx subdirectory. (We are using version 2.0.1-beta here for the sake of example only; be sure to change this to a specific version you're using.)

    $ tar xzvf sphinx-2.0.1-beta.tar.gz
    $ cd sphinx

  2. Run the configuration program:

    $ ./configure

    There's a number of options to configure. The complete listing may be obtained by using --help switch. The most important ones are:

    • --prefix, which specifies where to install Sphinx; such as --prefix=/usr/local/sphinx (all of the examples use this prefix)

    • --with-mysql, which specifies where to look for MySQL include and library files, if auto-detection fails;

    • --with-pgsql, which specifies where to look for PostgreSQL include and library files.

  3. Build the binaries:

    $ make

  4. Install the binaries in the directory of your choice: (defaults to /usr/local/bin/ on *nix systems, but is overridden with configure --prefix)

    $ make install

2.4. Installing Sphinx on Windows

Installing Sphinx on a Windows server is often easier than installing on a Linux environment; unless you are preparing code patches, you can use the pre-compiled binary files from the Downloads area on the website.

  1. Extract everything from the .zip file you have downloaded - sphinx-2.0.1-beta-win32.zip, or sphinx-2.0.1-beta-win32-pgsql.zip if you need PostgresSQL support as well. (We are using version 2.0.1-beta here for the sake of example only; be sure to change this to a specific version you're using.) You can use Windows Explorer in Windows XP and up to extract the files, or a freeware package like 7Zip to open the archive.

    For the remainder of this guide, we will assume that the folders are unzipped into C:\Sphinx, such that searchd.exe can be found in C:\Sphinx\bin\searchd.exe. If you decide to use any different location for the folders or configuration file, please change it accordingly.

  2. Edit the contents of sphinx.conf.in - specifically entries relating to @[email protected] - to paths suitable for your system.

  3. Install the searchd system as a Windows service:

    C:\Sphinx\bin> C:\Sphinx\bin\searchd --install --config C:\Sphinx\sphinx.conf.in --servicename SphinxSearch

  4. The searchd service will now be listed in the Services panel within the Management Console, available from Administrative Tools. It will not have been started, as you will need to configure it and build your indexes with indexer before starting the service. A guide to do this can be found under Quick tour.

    During the next steps of the install (which involve running indexer pretty much as you would on Linux) you may find that you get an error relating to libmysql.dll not being found. If you have MySQL installed, you should find a copy of this library in your Windows directory, or sometimes in Windows\System32, or failing that in the MySQL core directories. If you do receive an error please copy libmysql.dll into the bin directory.

2.5. Known installation issues

If configure fails to locate MySQL headers and/or libraries, try checking for and installing mysql-devel package. On some systems, it is not installed by default.

If make fails with a message which look like

/bin/sh: g++: command not found
make[1]: *** [libsphinx_a-sphinx.o] Error 127

try checking for and installing gcc-c++ package.

If you are getting compile-time errors which look like

sphinx.cpp:67: error: invalid application of `sizeof' to
    incomplete type `Private::SizeError<false>'

this means that some compile-time type size check failed. The most probable reason is that off_t type is less than 64-bit on your system. As a quick hack, you can edit sphinx.h and replace off_t with DWORD in a typedef for SphOffset_t, but note that this will prohibit you from using full-text indexes larger than 2 GB. Even if the hack helps, please report such issues, providing the exact error message and compiler/OS details, so I could properly fix them in next releases.

If you keep getting any other error, or the suggestions above do not seem to help you, please don't hesitate to contact me.

2.6. Quick Sphinx usage tour

All the example commands below assume that you installed Sphinx in /usr/local/sphinx, so searchd can be found in /usr/local/sphinx/bin/searchd.

To use Sphinx, you will need to:

  1. Create a configuration file.

    Default configuration file name is sphinx.conf. All Sphinx programs look for this file in current working directory by default.

    Sample configuration file, sphinx.conf.dist, which has all the options documented, is created by configure. Copy and edit that sample file to make your own configuration: (assuming Sphinx is installed into /usr/local/sphinx/)

    $ cd /usr/local/sphinx/etc
    $ cp sphinx.conf.dist sphinx.conf
    $ vi sphinx.conf

    Sample configuration file is setup to index documents table from MySQL database test; so there's example.sql sample data file to populate that table with a few documents for testing purposes:

    $ mysql -u test < /usr/local/sphinx/etc/example.sql

  2. Run the indexer to create full-text index from your data:

    $ cd /usr/local/sphinx/etc
    $ /usr/local/sphinx/bin/indexer --all

  3. Query your newly created index!

To query the index from command line, use search utility:

$ cd /usr/local/sphinx/etc
$ /usr/local/sphinx/bin/search test

To query the index from your PHP scripts, you need to:

  1. Run the search daemon which your script will talk to:

    $ cd /usr/local/sphinx/etc
    $ /usr/local/sphinx/bin/searchd

  2. Run the attached PHP API test script (to ensure that the daemon was succesfully started and is ready to serve the queries):

    $ cd sphinx/api
    $ php test.php test

  3. Include the API (it's located in api/sphinxapi.php) into your own scripts and use it.

Happy searching!

Chapter 3. Indexing

3.1. Data sources

The data to be indexed can generally come from very different sources: SQL databases, plain text files, HTML files, mailboxes, and so on. From Sphinx point of view, the data it indexes is a set of structured documents, each of which has the same set of fields and attributes. This is similar to SQL, where each row would correspond to a document, and each column to either a field or an attribute.

Depending on what source Sphinx should get the data from, different code is required to fetch the data and prepare it for indexing. This code is called data source driver (or simply driver or data source for brevity).

At the time of this writing, there are built-in drivers for MySQL, PostgreSQL, MS SQL (on Windows), and ODBC. There is also a generic driver called xmlpipe, which runs a specified command and reads the data from its stdout. See Section 3.9, “xmlpipe data source” section for the format description.

There can be as many sources per index as necessary. They will be sequentially processed in the very same order which was specifed in index definition. All the documents coming from those sources will be merged as if they were coming from a single source.

3.2. Full-text fields

Full-text fields (or just fields for brevity) are the textual document contents that get indexed by Sphinx, and can be (quickly) searched for keywords.

Fields are named, and you can limit your searches to a single field (eg. search through "title" only) or a subset of fields (eg. to "title" and "abstract" only). Sphinx index format generally supports up to 256 fields. However, up to version 2.0.1-beta indexes were forcibly limited by 32 fields, because of certain complications in the matching engine. Full support for up to 256 fields was added in version 2.0.2-beta.

Note that the original contents of the fields are not stored in the Sphinx index. The text that you send to Sphinx gets processed, and a full-text index (a special data structure that enables quick searches for a keyword) gets built from that text. But the original text contents are then simply discarded. Sphinx assumes that you store those contents elsewhere anyway.

Moreover, it is impossible to fully reconstruct the original text, because the specific whitespace, capitalization, punctuation, etc will all be lost during indexing. It is theoretically possible to partially reconstruct a given document from the Sphinx full-text index, but that would be a slow process (especially if the CRC dictionary is used, which does not even store the original keywords and works with their hashes instead).

3.3. Attributes

Attributes are additional values associated with each document that can be used to perform additional filtering and sorting during search.

It is often desired to additionally process full-text search results based not only on matching document ID and its rank, but on a number of other per-document values as well. For instance, one might need to sort news search results by date and then relevance, or search through products within specified price range, or limit blog search to posts made by selected users, or group results by month. To do that efficiently, Sphinx allows to attach a number of additional attributes to each document, and store their values in the full-text index. It's then possible to use stored values to filter, sort, or group full-text matches.

Attributes, unlike the fields, are not full-text indexed. They are stored in the index, but it is not possible to search them as full-text, and attempting to do so results in an error.

For example, it is impossible to use the extended matching mode expression @column 1 to match documents where column is 1, if column is an attribute, and this is still true even if the numeric digits are normally indexed.

Attributes can be used for filtering, though, to restrict returned rows, as well as sorting or result grouping; it is entirely possible to sort results purely based on attributes, and ignore the search relevance tools. Additionally, attributes are returned from the search daemon, while the indexed text is not.

A good example for attributes would be a forum posts table. Assume that only title and content fields need to be full-text searchable - but that sometimes it is also required to limit search to a certain author or a sub-forum (ie. search only those rows that have some specific values of author_id or forum_id columns in the SQL table); or to sort matches by post_date column; or to group matching posts by month of the post_date and calculate per-group match counts.

This can be achieved by specifying all the mentioned columns (excluding title and content, that are full-text fields) as attributes, indexing them, and then using API calls to setup filtering, sorting, and grouping. Here as an example.

Example sphinx.conf part:

sql_query = SELECT id, title, content, \
	author_id, forum_id, post_date FROM my_forum_posts
sql_attr_uint = author_id
sql_attr_uint = forum_id
sql_attr_timestamp = post_date

Example application code (in PHP):

// only search posts by author whose ID is 123
$cl->SetFilter ( "author_id", array ( 123 ) );

// only search posts in sub-forums 1, 3 and 7
$cl->SetFilter ( "forum_id", array ( 1,3,7 ) );

// sort found posts by posting date in descending order
$cl->SetSortMode ( SPH_SORT_ATTR_DESC, "post_date" );

Attributes are named. Attribute names are case insensitive. Attributes are not full-text indexed; they are stored in the index as is. Currently supported attribute types are:

  • unsigned integers (1-bit to 32-bit wide);

  • UNIX timestamps;

  • floating point values (32-bit, IEEE 754 single precision);

  • string ordinals (specially computed integers);

  • strings (since 1.10-beta);

  • MVA, multi-value attributes (variable-length lists of 32-bit unsigned integers).

The complete set of per-document attribute values is sometimes referred to as docinfo. Docinfos can either be

  • stored separately from the main full-text index data ("extern" storage, in .spa file), or

  • attached to each occurence of document ID in full-text index data ("inline" storage, in .spd file).

When using extern storage, a copy of .spa file (with all the attribute values for all the documents) is kept in RAM by searchd at all times. This is for performance reasons; random disk I/O would be too slow. On the contrary, inline storage does not require any additional RAM at all, but that comes at the cost of greatly inflating the index size: remember that it copies all attribute value every time when the document ID is mentioned, and that is exactly as many times as there are different keywords in the document. Inline may be the only viable option if you have only a few attributes and need to work with big datasets in limited RAM. However, in most cases extern storage makes both indexing and searching much more efficient.

Search-time memory requirements for extern storage are (1+number_of_attrs)*number_of_docs*4 bytes, ie. 10 million docs with 2 groups and 1 timestamp will take (1+2+1)*10M*4 = 160 MB of RAM. This is PER DAEMON, not per query. searchd will allocate 160 MB on startup, read the data and keep it shared between queries. The children will NOT allocate any additional copies of this data.

3.4. MVA (multi-valued attributes)

MVAs, or multi-valued attributes, are an important special type of per-document attributes in Sphinx. MVAs let you attach sets of numeric values to every document. That is useful to implement article tags, product categories, etc. Filtering and group-by (but not sorting) on MVA attributes is supported.

As of version 2.0.2-beta, MVA values can either be unsigned 32-bit integers (UNSIGNED INTEGER) or signed 64-bit integers (BIGINT). Up to version 2.0.1-beta, only the unsigned 32-bit values were supported.

The set size is not limited, you can have an arbitrary number of values attached to each document as long as RAM permits (.spm file that contains the MVA values will be precached in RAM by searchd). The source data can be taken either from a separate query, or from a document field; see source type in sql_attr_multi. In the first case the query will have to return pairs of document ID and MVA values, in the second one the field will be parsed for integer values. There are absolutely no requirements as to incoming data order; the values will be automatically grouped by document ID (and internally sorted within the same ID) during indexing anyway.

When filtering, a document will match the filter on MVA attribute if any of the values satisfy the filtering condition. (Therefore, documents that pass through exclude filters will not contain any of the forbidden values.) When grouping by MVA attribute, a document will contribute to as many groups as there are different MVA values associated with that document. For instance, if the collection contains exactly 1 document having a 'tag' MVA with values 5, 7, and 11, grouping on 'tag' will produce 3 groups with '@count' equal to 1 and '@groupby' key values of 5, 7, and 11 respectively. Also note that grouping by MVA might lead to duplicate documents in the result set: because each document can participate in many groups, it can be chosen as the best one in in more than one group, leading to duplicate IDs. PHP API historically uses ordered hash on the document ID for the resulting rows; so you'll also need to use SetArrayResult() in order to employ group-by on MVA with PHP API.

3.5. Indexes

To be able to answer full-text search queries fast, Sphinx needs to build a special data structure optimized for such queries from your text data. This structure is called index; and the process of building index from text is called indexing.

Different index types are well suited for different tasks. For example, a disk-based tree-based index would be easy to update (ie. insert new documents to existing index), but rather slow to search. Therefore, Sphinx architecture allows for different index types to be implemented easily.

The only index type which is implemented in Sphinx at the moment is designed for maximum indexing and searching speed. This comes at a cost of updates being really slow; theoretically, it might be slower to update this type of index than than to reindex it from scratch. However, this very frequently could be worked around with muiltiple indexes, see Section 3.11, “Live index updates” for details.

It is planned to implement more index types, including the type which would be updateable in real time.

There can be as many indexes per configuration file as necessary. indexer utility can reindex either all of them (if --all option is specified), or a certain explicitly specified subset. searchd utility will serve all the specified indexes, and the clients can specify what indexes to search in run time.

3.6. Restrictions on the source data

There are a few different restrictions imposed on the source data which is going to be indexed by Sphinx, of which the single most important one is:


If this requirement is not met, different bad things can happen. For instance, Sphinx can crash with an internal assertion while indexing; or produce strange results when searching due to conflicting IDs. Also, a 1000-pound gorilla might eventually come out of your display and start throwing barrels at you. You've been warned.

3.7. Charsets, case folding, and translation tables

When indexing some index, Sphinx fetches documents from the specified sources, splits the text into words, and does case folding so that "Abc", "ABC" and "abc" would be treated as the same word (or, to be pedantic, term).

To do that properly, Sphinx needs to know

  • what encoding is the source text in;

  • what characters are letters and what are not;

  • what letters should be folded to what letters.

This should be configured on a per-index basis using charset_type and charset_table options. charset_type specifies whether the document encoding is single-byte (SBCS) or UTF-8. charset_table specifies the table that maps letter characters to their case folded versions. The characters that are not in the table are considered to be non-letters and will be treated as word separators when indexing or searching through this index.

Note that while default tables do not include space character (ASCII code 0x20, Unicode U+0020) as a letter, it's in fact perfectly legal to do so. This can be useful, for instance, for indexing tag clouds, so that space-separated word sets would index as a single search query term.

Default tables currently include English and Russian characters. Please do submit your tables for other languages!

3.8. SQL data sources (MySQL, PostgreSQL)

With all the SQL drivers, indexing generally works as follows.

Most options, such as database user/host/password, are straightforward. However, there are a few subtle things, which are discussed in more detail here.

Ranged queries

Main query, which needs to fetch all the documents, can impose a read lock on the whole table and stall the concurrent queries (eg. INSERTs to MyISAM table), waste a lot of memory for result set, etc. To avoid this, Sphinx supports so-called ranged queries. With ranged queries, Sphinx first fetches min and max document IDs from the table, and then substitutes different ID intervals into main query text and runs the modified query to fetch another chunk of documents. Here's an example.

Example 3.1. Ranged query usage example

# in sphinx.conf

sql_query_range	= SELECT MIN(id),MAX(id) FROM documents
sql_range_step = 1000
sql_query = SELECT * FROM documents WHERE id>=$start AND id<=$end

If the table contains document IDs from 1 to, say, 2345, then sql_query would be run three times:

  1. with $start replaced with 1 and $end replaced with 1000;

  2. with $start replaced with 1001 and $end replaced with 2000;

  3. with $start replaced with 2000 and $end replaced with 2345.

Obviously, that's not much of a difference for 2000-row table, but when it comes to indexing 10-million-row MyISAM table, ranged queries might be of some help.

sql_post vs. sql_post_index

The difference between post-query and post-index query is in that post-query is run immediately when Sphinx received all the documents, but further indexing may still fail for some other reason. On the contrary, by the time the post-index query gets executed, it is guaranteed that the indexing was succesful. Database connection is dropped and re-established because sorting phase can be very lengthy and would just timeout otherwise.

3.9. xmlpipe data source

xmlpipe data source was designed to enable users to plug data into Sphinx without having to implement new data sources drivers themselves. It is limited to 2 fixed fields and 2 fixed attributes, and is deprecated in favor of Section 3.10, “xmlpipe2 data source” now. For new streams, use xmlpipe2.

To use xmlpipe, configure the data source in your configuration file as follows:

source example_xmlpipe_source
    type = xmlpipe
    xmlpipe_command = perl /www/mysite.com/bin/sphinxpipe.pl

The indexer will run the command specified in xmlpipe_command, and then read, parse and index the data it prints to stdout. More formally, it opens a pipe to given command and then reads from that pipe.

indexer will expect one or more documents in custom XML format. Here's the example document stream, consisting of two documents:

Example 3.2. XMLpipe document stream

<title>test title</title>
this is my document body

<title>another test</title>
this is another document

Legacy xmlpipe legacy driver uses a builtin parser which is pretty fast but really strict and does not actually fully support XML. It requires that all the fields must be present, formatted exactly as in this example, and occur exactly in the same order. The only optional field is timestamp; it defaults to 1.

3.10. xmlpipe2 data source

xmlpipe2 lets you pass arbitrary full-text and attribute data to Sphinx in yet another custom XML format. It also allows to specify the schema (ie. the set of fields and attributes) either in the XML stream itself, or in the source settings.

When indexing xmlpipe2 source, indexer runs the given command, opens a pipe to its stdout, and expects well-formed XML stream. Here's sample stream data:

Example 3.3. xmlpipe2 document stream

<?xml version="1.0" encoding="utf-8"?>

<sphinx:field name="subject"/> 
<sphinx:field name="content"/>
<sphinx:attr name="published" type="timestamp"/>
<sphinx:attr name="author_id" type="int" bits="16" default="1"/>

<sphinx:document id="1234">
<content>this is the main content <![CDATA[[and this <cdata> entry
must be handled properly by xml parser lib]]></content>
<subject>note how field/attr tags can be
in <b class="red">randomized</b> order</subject>
<misc>some undeclared element</misc>

<sphinx:document id="1235">
<subject>another subject</subject>
<content>here comes another document, and i am given to understand,
that in-document field order must not matter, sir</content>

<!-- ... even more sphinx:document entries here ... -->



Arbitrary fields and attributes are allowed. They also can occur in the stream in arbitrary order within each document; the order is ignored. There is a restriction on maximum field length; fields longer than 2 MB will be truncated to 2 MB (this limit can be changed in the source).

The schema, ie. complete fields and attributes list, must be declared before any document could be parsed. This can be done either in the configuration file using xmlpipe_field and xmlpipe_attr_XXX settings, or right in the stream using <sphinx:schema> element. <sphinx:schema> is optional. It is only allowed to occur as the very first sub-element in <sphinx:docset>. If there is no in-stream schema definition, settings from the configuration file will be used. Otherwise, stream settings take precedence.

Unknown tags (which were not declared neither as fields nor as attributes) will be ignored with a warning. In the example above, <misc> will be ignored. All embedded tags and their attributes (such as <b> in <subject> in the example above) will be silently ignored.

Support for incoming stream encodings depends on whether iconv is installed on the system. xmlpipe2 is parsed using libexpat parser that understands US-ASCII, ISO-8859-1, UTF-8 and a few UTF-16 variants natively. Sphinx configure script will also check for libiconv presence, and utilize it to handle other encodings. libexpat also enforces the requirement to use UTF-8 charset on Sphinx side, because the parsed data it returns is always in UTF-8.

XML elements (tags) recognized by xmlpipe2 (and their attributes where applicable) are:


Mandatory top-level element, denotes and contains xmlpipe2 document set.


Optional element, must either occur as the very first child of sphinx:docset, or never occur at all. Declares the document schema. Contains field and attribute declarations. If present, overrides per-source settings from the configuration file.


Optional element, child of sphinx:schema. Declares a full-text field. Known attributes are:

  • "name", specifies the XML element name that will be treated as a full-text field in the subsequent documents.

  • "attr", specifies whether to also index this field as a string or word count attribute. Possible values are "string" and "wordcount". Introduced in version 1.10-beta.


Optional element, child of sphinx:schema. Declares an attribute. Known attributes are:

  • "name", specifies the element name that should be treated as an attribute in the subsequent documents.

  • "type", specifies the attribute type. Possible values are "int", "timestamp", "str2ordinal", "bool", "float" and "multi".

  • "bits", specifies the bit size for "int" attribute type. Valid values are 1 to 32.

  • "default", specifies the default value for this attribute that should be used if the attribute's element is not present in the document.


Mandatory element, must be a child of sphinx:docset. Contains arbitrary other elements with field and attribute values to be indexed, as declared either using sphinx:field and sphinx:attr elements or in the configuration file. The only known attribute is "id" that must contain the unique integer document ID.


Optional element, child of sphinx:docset. Contains a number of "id" elements whose contents are document IDs to be put into a kill-list for this index.

3.11. Live index updates

There are two major approaches to maintaining the full-text index contents up to date. Note, however, that both these approaches deal with the task of full-text data updates, and not attribute updates. Instant attribute updates are supported since version 0.9.8. Refer to UpdateAttributes() API call description for details.

First, you can use disk-based indexes, partition them manually, and only rebuild the smaller partitions (so-called "deltas") frequently. By minimizing the rebuild size, you can reduce the average indexing lag to something as low as 30-60 seconds. This approach was the the only one available in versions 0.9.x. On huge collections it actually might be the most efficient one. Refer to Section 3.12, “Delta index updates” for details.

Second, versions 1.x (starting with 1.10-beta) add support for so-called real-time indexes (RT indexes for short) that on-the-fly updates of the full-text data. Updates on a RT index can appear in the search results in 1-2 milliseconds, ie. 0.001-0.002 seconds. However, RT index are less efficient for bulk indexing huge amounts of data. Refer to Chapter 4, Real-time indexes for details.

3.12. Delta index updates

There's a frequent situation when the total dataset is too big to be reindexed from scratch often, but the amount of new records is rather small. Example: a forum with a 1,000,000 archived posts, but only 1,000 new posts per day.

In this case, "live" (almost real time) index updates could be implemented using so called "main+delta" scheme.

The idea is to set up two sources and two indexes, with one "main" index for the data which only changes rarely (if ever), and one "delta" for the new documents. In the example above, 1,000,000 archived posts would go to the main index, and newly inserted 1,000 posts/day would go to the delta index. Delta index could then be reindexed very frequently, and the documents can be made available to search in a matter of minutes.

Specifying which documents should go to what index and reindexing main index could also be made fully automatic. One option would be to make a counter table which would track the ID which would split the documents, and update it whenever the main index is reindexed.

Example 3.4. Fully automated live updates

# in MySQL
CREATE TABLE sph_counter
    max_doc_id INTEGER NOT NULL

# in sphinx.conf
source main
    # ...
    sql_query_pre = SET NAMES utf8
    sql_query_pre = REPLACE INTO sph_counter SELECT 1, MAX(id) FROM documents
    sql_query = SELECT id, title, body FROM documents \
        WHERE id<=( SELECT max_doc_id FROM sph_counter WHERE counter_id=1 )

source delta : main
    sql_query_pre = SET NAMES utf8
    sql_query = SELECT id, title, body FROM documents \
        WHERE id>( SELECT max_doc_id FROM sph_counter WHERE counter_id=1 )

index main
    source = main
    path = /path/to/main
    # ... all the other settings

# note how all other settings are copied from main,
# but source and path are overridden (they MUST be)
index delta : main
    source = delta
    path = /path/to/delta

Note how we're overriding sql_query_pre in the delta source. We need to explicitly have that override. Otherwise REPLACE query would be run when indexing delta source too, effectively nullifying it. However, when we issue the directive in the inherited source for the first time, it removes all inherited values, so the encoding setup is also lost. So sql_query_pre in the delta can not just be empty; and we need to issue the encoding setup query explicitly once again.

3.13. Index merging

Merging two existing indexes can be more efficient that indexing the data from scratch, and desired in some cases (such as merging 'main' and 'delta' indexes instead of simply reindexing 'main' in 'main+delta' partitioning scheme). So indexer has an option to do that. Merging the indexes is normally faster than reindexing but still not instant on huge indexes. Basically, it will need to read the contents of both indexes once and write the result once. Merging 100 GB and 1 GB index, for example, will result in 202 GB of IO (but that's still likely less than the indexing from scratch requires).

The basic command syntax is as follows:

indexer --merge DSTINDEX SRCINDEX [--rotate]

Only the DSTINDEX index will be affected: the contents of SRCINDEX will be merged into it. --rotate switch will be required if DSTINDEX is already being served by searchd. The initially devised usage pattern is to merge a smaller update from SRCINDEX into DSTINDEX. Thus, when merging the attributes, values from SRCINDEX will win if duplicate document IDs are encountered. Note, however, that the "old" keywords will not be automatically removed in such cases. For example, if there's a keyword "old" associated with document 123 in DSTINDEX, and a keyword "new" associated with it in SRCINDEX, document 123 will be found by both keywords after the merge. You can supply an explicit condition to remove documents from DSTINDEX to mitigate that; the relevant switch is --merge-dst-range:

indexer --merge main delta --merge-dst-range deleted 0 0

This switch lets you apply filters to the destination index along with merging. There can be several filters; all of their conditions must be met in order to include the document in the resulting mergid index. In the example above, the filter passes only those records where 'deleted' is 0, eliminating all records that were flagged as deleted (for instance, using UpdateAttributes() call).

Chapter 4. Real-time indexes

Real-time indexes (or RT indexes for brevity) are a new backend that lets you insert, update, or delete documents (rows) on the fly. RT indexes were added in version 1.10-beta. While querying of RT indexes is possible using any of the SphinxAPI, SphinxQL, or SphinxSE, updating them is only possible via SphinxQL at the moment. Full SphinxQL reference is available in Chapter 7, SphinxQL reference.

4.1. RT indexes overview

RT indexes should be declared in sphinx.conf, just as every other index type. Notable differences from the regular, disk-based indexes are that a) data sources are not required and ignored, and b) you should explicitly enumerate all the text fields, not just attributes. Here's an example:

Example 4.1. RT index declaration

index rt
	type = rt
	path = /usr/local/sphinx/data/rt
	rt_field = title
	rt_field = content
	rt_attr_uint = gid

As of 2.0.1-beta and above, RT indexes are production quality, despite a few missing features.

RT index can be accessed using MySQL protocol. INSERT, REPLACE, DELETE, and SELECT statements against RT index are supported. For instance, this is an example session with the sample index above:

$ mysql -h -P 9306
Welcome to the MySQL monitor.  Commands end with ; or \g.
Your MySQL connection id is 1
Server version: 1.10-dev (r2153)

Type 'help;' or '\h' for help. Type '\c' to clear the buffer.

mysql> INSERT INTO rt VALUES ( 1, 'first record', 'test one', 123 );
Query OK, 1 row affected (0.05 sec)

mysql> INSERT INTO rt VALUES ( 2, 'second record', 'test two', 234 );
Query OK, 1 row affected (0.00 sec)

mysql> SELECT * FROM rt;
| id   | weight | gid  |
|    1 |      1 |  123 |
|    2 |      1 |  234 |
2 rows in set (0.02 sec)

mysql> SELECT * FROM rt WHERE MATCH('test');
| id   | weight | gid  |
|    1 |   1643 |  123 |
|    2 |   1643 |  234 |
2 rows in set (0.01 sec)

mysql> SELECT * FROM rt WHERE MATCH('@title test');
Empty set (0.00 sec)

Both partial and batch INSERT syntaxes are supported, ie. you can specify a subset of columns, and insert several rows at a time. Deletions are also possible using DELETE statement; the only currently supported syntax is DELETE FROM <index> WHERE id=<id>. REPLACE is also supported, enabling you to implement updates.

mysql> INSERT INTO rt ( id, title ) VALUES ( 3, 'third row' ), ( 4, 'fourth entry' );
Query OK, 2 rows affected (0.01 sec)

mysql> SELECT * FROM rt;
| id   | weight | gid  |
|    1 |      1 |  123 |
|    2 |      1 |  234 |
|    3 |      1 |    0 |
|    4 |      1 |    0 |
4 rows in set (0.00 sec)

mysql> DELETE FROM rt WHERE id=2;
Query OK, 0 rows affected (0.00 sec)

mysql> SELECT * FROM rt WHERE MATCH('test');
| id   | weight | gid  |
|    1 |   1500 |  123 |
1 row in set (0.00 sec)

mysql> INSERT INTO rt VALUES ( 1, 'first record on steroids', 'test one', 123 );
ERROR 1064 (42000): duplicate id '1'

mysql> REPLACE INTO rt VALUES ( 1, 'first record on steroids', 'test one', 123 );
Query OK, 1 row affected (0.01 sec)

mysql> SELECT * FROM rt WHERE MATCH('steroids');
| id   | weight | gid  |
|    1 |   1500 |  123 |
1 row in set (0.01 sec)

Data stored in RT index should survive clean shutdown. When binary logging is enabled, it should also survive crash and/or dirty shutdown, and recover on subsequent startup.

4.2. Known caveats with RT indexes

As of 1.10-beta, RT indexes are a beta quality feature: while no major, showstopper-class issues are known, there still are a few known usage quirks. Those quirks are listed in this section.

  • Prefix and infix indexing are not supported yet.

  • MVAs are not supported yet.

  • Disk chunks optimization routine is not implemented yet.

  • On initial index creation, attributes are reordered by type, in the following order: uint, bigint, float, timestamp, string. So when using INSERT without an explicit column names list, specify all uint column values first, then bigint, etc.

  • Default conservative RAM chunk limit (rt_mem_limit) of 32M can lead to poor performance on bigger indexes, you should raise it to 256..1024M if you're planning to index gigabytes.

  • High DELETE/REPLACE rate can lead to kill-list fragmentation and impact searching performance.

  • No transaction size limits are currently imposed; too many concurrent INSERT/REPLACE transactions might therefore consume a lot of RAM.

  • In case of a damaged binlog, recovery will stop on the first damaged transaction, even though it's technically possible to keep looking further for subsequent undamaged transactions, and recover those. This mid-file damage case (due to flaky HDD/CDD/tape?) is supposed to be extremely rare, though.

  • Multiple INSERTs grouped in a single transaction perform better than equivalent single-row transactions and are recommended for batch loading of data.

4.3. RT index internals

RT index is internally chunked. It keeps a so-called RAM chunk that stores all the most recent changes. RAM chunk memory usage is rather strictly limited with per-index rt_mem_limit directive. Once RAM chunk grows over this limit, a new disk chunk is created from its data, and RAM chunk is reset. Thus, while most changes on the RT index will be performed in RAM only and complete instantly (in milliseconds), those changes that overflow the RAM chunk will stall for the duration of disk chunk creation (a few seconds).

Disk chunks are, in fact, just regular disk-based indexes. But they're a part of an RT index and automatically managed by it, so you need not configure nor manage them manually. Because a new disk chunk is created every time RT chunk overflows the limit, and because in-memory chunk format is close to on-disk format, the disk chunks will be approximately rt_mem_limit bytes in size each.

Generally, it is better to set the limit bigger, to minimize both the frequency of flushes, and the index fragmentation (number of disk chunks). For instance, on a dedicated search server that handles a big RT index, it can be advised to set rt_mem_limit to 1-2 GB. A global limit on all indexes is also planned, but not yet implemented yet as of 1.10-beta.

Disk chunk full-text index data can not be actually modified, so the full-text field changes (ie. row deletions and updates) suppress a previous row version from a disk chunk using a kill-list, but do not actually physically purge the data. Therefore, on workloads with high full-text updates ratio index might eventually get polluted by these previous row versions, and searching performance would degrade. Physical index purging that would improve the performance is planned, but not yet implemented as of 1.10-beta.

Data in RAM chunk gets saved to disk on clean daemon shutdown, and then loaded back on startup. However, on daemon or server crash, updates from RAM chunk might be lost. To prevent that, binary logging of transactions can be used; see Section 4.4, “Binary logging” for details.

Full-text changes in RT index are transactional. They are stored in a per-thread accumulator until COMMIT, then applied at once. Bigger batches per single COMMIT should result in faster indexing.

4.4. Binary logging

Binary logs are essentially a recovery mechanism. With binary logs enabled, searchd writes every given transaction to the binlog file, and uses that for recovery after an unclean shutdown. On clean shutdown, RAM chunks are saved to disk, and then all the binlog files are unlinked.

During normal operation, a new binlog file will be opened every time when binlog_max_log_size limit is reached. Older, already closed binlog files are kept until all of the transactions stored in them (from all indexes) are flushed as a disk chunk. Setting the limit to 0 pretty much prevents binlog from being unlinked at all while searchd is running; however, it will still be unlinked on clean shutdown. (This is the default case as of 2.0.3-release, binlog_max_log_size defaults to 0.)

There are 3 different binlog flushing strategies, controlled by binlog_flush directive which takes the values of 0, 1, or 2. 0 means to flush the log to OS and sync it to disk every second; 1 means flush and sync every transaction; and 2 (the default mode) means flush every transaction but sync every second. Sync is relatively slow because it has to perform physical disk writes, so mode 1 is the safest (every committed transaction is guaranteed to be written on disk) but the slowest. Flushing log to OS prevents from data loss on searchd crashes but not system crashes. Mode 2 is the default.

On recovery after an unclean shutdown, binlogs are replayed and all logged transactions since the last good on-disk state are restored. Transactions are checksummed so in case of binlog file corruption garbage data will not be replayed; such a broken transaction will be detected and, currently, will stop replay. Transactions also start with a magic marker and timestamped, so in case of binlog damage in the middle of the file, it's technically possible to skip broken transactions and keep replaying from the next good one, and/or it's possible to replay transactions until a given timestamp (point-in-time recovery), but none of that is implemented yet as of 1.10-beta.

One unwanted side effect of binlogs is that actively updating a small RT index that fully fits into a RAM chunk part will lead to an ever-growing binlog that can never be unlinked until clean shutdown. Binlogs are essentially append-only deltas against the last known good saved state on disk, and unless RAM chunk gets saved, they can not be unlinked. An ever-growing binlog is not very good for disk use and crash recovery time. Starting with 2.0.1-beta you can configure searchd to perform a periodic RAM chunk flush to fix that problem using a rt_flush_period directive. With periodic flushes enabled, searchd will keep a separate thread, checking whether RT indexes RAM chunks need to be written back to disk. Once that happens, the respective binlogs can be (and are) safely unlinked.

Note that rt_flush_period only controls the frequency at which the checks happen. There are no guarantees that the particular RAM chunk will get saved. For instance, it does not make sense to regularly re-save a huge RAM chunk that only gets a few rows worh of updates. The search daemon determine whether to actually perform the flush with a few heuristics.

Chapter 5. Searching

5.1. Matching modes

So-called matching modes are a legacy feature that used to provide (very) limited query syntax and ranking support. Currently, they are deprecated in favor of full-text query language and so-called rankers. Starting with version 0.9.9-release, it is thus strongly recommended to use SPH_MATCH_EXTENDED and proper query syntax rather than any other legacy mode. All those other modes are actually internally converted to extended syntax anyway. SphinxAPI still defaults to SPH_MATCH_ALL but that is for compatibility reasons only.

There are the following matching modes available:

  • SPH_MATCH_ALL, matches all query words (default mode);

  • SPH_MATCH_ANY, matches any of the query words;

  • SPH_MATCH_PHRASE, matches query as a phrase, requiring perfect match;

  • SPH_MATCH_BOOLEAN, matches query as a boolean expression (see Section 5.2, “Boolean query syntax”);

  • SPH_MATCH_EXTENDED, matches query as an expression in Sphinx internal query language (see Section 5.3, “Extended query syntax”);


  • SPH_MATCH_FULLSCAN, matches query, forcibly using the "full scan" mode as below. NB, any query terms will be ignored, such that filters, filter-ranges and grouping will still be applied, but no text-matching.

SPH_MATCH_EXTENDED2 was used during 0.9.8 and 0.9.9 development cycle, when the internal matching engine was being rewritten (for the sake of additional functionality and better performance). By 0.9.9-release, the older version was removed, and SPH_MATCH_EXTENDED and SPH_MATCH_EXTENDED2 are now just aliases.

The SPH_MATCH_FULLSCAN mode will be automatically activated in place of the specified matching mode when the following conditions are met:

  1. The query string is empty (ie. its length is zero).

  2. docinfo storage is set to extern.

In full scan mode, all the indexed documents will be considered as matching. Such queries will still apply filters, sorting, and group by, but will not perform any full-text searching. This can be useful to unify full-text and non-full-text searching code, or to offload SQL server (there are cases when Sphinx scans will perform better than analogous MySQL queries). An example of using the full scan mode might be to find posts in a forum. By selecting the forum's user ID via SetFilter() but not actually providing any search text, Sphinx will match every document (i.e. every post) where SetFilter() would match - in this case providing every post from that user. By default this will be ordered by relevancy, followed by Sphinx document ID in ascending order (earliest first).

5.2. Boolean query syntax

Boolean queries allow the following special operators to be used:

  • explicit operator AND:

    hello & world
  • operator OR:

    hello | world
  • operator NOT:

    hello -world
    hello !world

  • grouping:

    ( hello world )

Here's an example query which uses all these operators:

Example 5.1. Boolean query example

( cat -dog ) | ( cat -mouse)

There always is implicit AND operator, so "hello world" query actually means "hello & world".

OR operator precedence is higher than AND, so "looking for cat | dog | mouse" means "looking for ( cat | dog | mouse )" and not "(looking for cat) | dog | mouse".

Queries like "-dog", which implicitly include all documents from the collection, can not be evaluated. This is both for technical and performance reasons. Technically, Sphinx does not always keep a list of all IDs. Performance-wise, when the collection is huge (ie. 10-100M documents), evaluating such queries could take very long.

5.3. Extended query syntax

The following special operators and modifiers can be used when using the extended matching mode:

  • operator OR:

    hello | world
  • operator NOT:

    hello -world
    hello !world

  • field search operator:

    @title hello @body world
  • field position limit modifier (introduced in version 0.9.9-rc1):

    @body[50] hello
  • multiple-field search operator:

    @(title,body) hello world
  • all-field search operator:

    @* hello
  • phrase search operator:

    "hello world"
  • proximity search operator:

    "hello world"~10
  • quorum matching operator:

    "the world is a wonderful place"/3
  • strict order operator (aka operator "before"):

    aaa << bbb << ccc
  • exact form modifier (introduced in version 0.9.9-rc1):

    raining =cats and =dogs
  • field-start and field-end modifier (introduced in version 0.9.9-rc2):

    ^hello world$
  • NEAR, generalized proximity operator (introduced in version 2.0.1-beta):

    hello NEAR/3 world NEAR/4 "my test"
  • SENTENCE operator (introduced in version 2.0.1-beta):

    all SENTENCE words SENTENCE "in one sentence"
  • PARAGRAPH operator (introduced in version 2.0.1-beta):

    "Bill Gates" PARAGRAPH "Steve Jobs"
  • zone limit operator:

    ZONE:(h3,h4) only in these titles

Here's an example query that uses some of these operators:

Example 5.2. Extended matching mode: query example

"hello world" @title "example program"~5 @body python -(php|perl) @* code

The full meaning of this search is:

  • Find the words 'hello' and 'world' adjacently in any field in a document;

  • Additionally, the same document must also contain the words 'example' and 'program' in the title field, with up to, but not including, 5 words between the words in question; (E.g. "example PHP program" would be matched however "example script to introduce outside data into the correct context for your program" would not because two terms have 5 or more words between them)

  • Additionally, the same document must contain the word 'python' in the body field, but not contain either 'php' or 'perl';

  • Additionally, the same document must contain the word 'code' in any field.

There always is implicit AND operator, so "hello world" means that both "hello" and "world" must be present in matching document.

OR operator precedence is higher than AND, so "looking for cat | dog | mouse" means "looking for ( cat | dog | mouse )" and not "(looking for cat) | dog | mouse".

Field limit operator limits subsequent searching to a given field. Normally, query will fail with an error message if given field name does not exist in the searched index. However, that can be suppressed by specifying "@@relaxed" option at the very beginning of the query:

@@relaxed @nosuchfield my query

This can be helpful when searching through heterogeneous indexes with different schemas.

Field position limit, introduced in version 0.9.9-rc1, additionaly restricts the searching to first N position within given field (or fields). For example, "@body[50] hello" will not match the documents where the keyword 'hello' occurs at position 51 and below in the body.

Proximity distance is specified in words, adjusted for word count, and applies to all words within quotes. For instance, "cat dog mouse"~5 query means that there must be less than 8-word span which contains all 3 words, ie. "CAT aaa bbb ccc DOG eee fff MOUSE" document will not match this query, because this span is exactly 8 words long.

Quorum matching operator introduces a kind of fuzzy matching. It will only match those documents that pass a given threshold of given words. The example above ("the world is a wonderful place"/3) will match all documents that have at least 3 of the 6 specified words.

Strict order operator (aka operator "before"), introduced in version 0.9.9-rc2, will match the document only if its argument keywords occur in the document exactly in the query order. For instance, "black << cat" query (without quotes) will match the document "black and white cat" but not the "that cat was black" document. Order operator has the lowest priority. It can be applied both to just keywords and more complex expressions, ie. this is a valid query:

(bag of words) << "exact phrase" << red|green|blue

Exact form keyword modifier, introduced in version 0.9.9-rc1, will match the document only if the keyword occurred in exactly the specified form. The default behaviour is to match the document if the stemmed keyword matches. For instance, "runs" query will match both the document that contains "runs" and the document that contains "running", because both forms stem to just "run" - while "=runs" query will only match the first document. Exact form operator requires index_exact_words option to be enabled. This is a modifier that affects the keyword and thus can be used within operators such as phrase, proximity, and quorum operators.

Field-start and field-end keyword modifiers, introduced in version 0.9.9-rc2, will make the keyword match only if it occurred at the very start or the very end of a fulltext field, respectively. For instance, the query "^hello world$" (with quotes and thus combining phrase operator and start/end modifiers) will only match documents that contain at least one field that has exactly these two keywords.

Starting with 0.9.9-rc1, arbitrarily nested brackets and negations are allowed. However, the query must be possible to compute without involving an implicit list of all documents:

// correct query
aaa -(bbb -(ccc ddd))

// queries that are non-computable
aaa | -bbb

NEAR operator, added in 2.0.1-beta, is a generalized version of a proximity operator. The syntax is NEAR/N, it is case-sensitive, and no spaces are allowed beetwen the NEAR keyword, the slash sign, and the distance value.

The original proximity operator only worked on sets of keywords. NEAR is more generic and can accept arbitrary subexpressions as its two arguments, matching the document when both subexpressions are found within N words of each other, no matter in which order. NEAR is left associative and has the same (lowest) precedence as BEFORE.

You should also note how a (one NEAR/7 two NEAR/7 three) query using NEAR is not really equivalent to a ("one two three"~7) one using keyword proximity operator. The difference here is that the proximity operator allows for up to 6 non-matching words between all the 3 matching words, but the version with NEAR is less restrictive: it would allow for up to 6 words between 'one' and 'two' and then for up to 6 more between that two-word matching and a 'three' keyword.

SENTENCE and PARAGRAPH operators, added in 2.0.1-beta, matches the document when both its arguments are within the same sentence or the same paragraph of text, respectively. The arguments can be either keywords, or phrases, or the instances of the same operator. Here are a few examples:

one SENTENCE two
one SENTENCE "two three"
one SENTENCE "two three" SENTENCE four

The order of the arguments within the sentence or paragraph does not matter. These operators only work on indexes built with index_sp (sentence and paragraph indexing feature) enabled, and revert to a mere AND otherwise. Refer to the index_sp directive documentation for the notes on what's considered a sentence and a paragraph.

ZONE limit operator, added in 2.0.1-beta, is quite similar to field limit operator, but restricts matching to a given in-field zone or a list of zones. Note that the subsequent subexpressions are not required to match in a single contiguous span of a given zone, and may match in multiple spans. For instance, (ZONE:th hello world) query will match this example document:

<th>Table 1. Local awareness of Hello Kitty brand.</th>
.. some table data goes here ..
<th>Table 2. World-wide brand awareness.</th>

ZONE operator affects the query until the next field or ZONE limit operator, or the closing parenthesis. It only works on the indexes built with zones support (see Section 11.2.9, “index_zones”) and will be ignored otherwise.

5.4. Search results ranking

Ranking overview

Ranking (aka weighting) of the search results can be defined as a process of computing a so-called relevance (aka weight) for every given matched document with regards to a given query that matched it. So relevance is in the end just a number attached to every document that estimates how relevant the document is to the query. Search results can then be sorted based on this number and/or some additional parameters, so that the most sought after results would come up higher on the results page.

There is no single standard one-size-fits-all way to rank any document in any scenario. Moreover, there can not ever be such a way, because relevance is subjective. As in, what seems relevant to you might not seem relevant to me. Hence, in general case it's not just hard to compute, it's theoretically impossible.

So ranking in Sphinx is configurable. It has a notion of a so-called ranker. A ranker can formally be defined as a function that takes document and query as its input and produces a relevance value as output. In layman's terms, a ranker controls exactly how (using which specific algorithm) will Sphinx assign weights to the document.

Previously, this ranking function was rigidly bound to the matching mode. So in the legacy matching modes (that is, SPH_MATCH_ALL, SPH_MATCH_ANY, SPH_MATCH_PHRASE, and SPH_MATCH_BOOLEAN) you can not choose the ranker. You can only do that in the SPH_MATCH_EXTENDED mode. (Which is the only mode in SphinxQL and the suggested mode in SphinxAPI anyway.) To choose a non-default ranker you can either use SetRankingMode() with SphinxAPI, or OPTION ranker clause in SELECT statement when using SphinxQL.

As a sidenote, legacy matching modes are internally implemented via the unified syntax anyway. When you use one of those modes, Sphinx just internally adjusts the query and sets the associated ranker, then executes the query using the very same unified code path.

Available rankers

Sphinx ships with a number of built-in rankers suited for different purposes. A number of them uses two factors, phrase proximity (aka LCS) and BM25. Phrase proximity works on the keyword positions, while BM25 works on the keyword frequencies. Basically, the better the degree of the phrase match between the document body and the query, the higher is the phrase proximity (it maxes out when the document contains the entire query as a verbatim quote). And BM25 is higher when the document contains more rare words. We'll save the detailed discussion for later.

Currently implemented rankers are:

  • SPH_RANK_PROXIMITY_BM25, the default ranking mode that uses and combines both phrase proximity and BM25 ranking.

  • SPH_RANK_BM25, statistical ranking mode which uses BM25 ranking only (similar to most other full-text engines). This mode is faster but may result in worse quality on queries which contain more than 1 keyword.

  • SPH_RANK_NONE, no ranking mode. This mode is obviously the fastest. A weight of 1 is assigned to all matches. This is sometimes called boolean searching that just matches the documents but does not rank them.

  • SPH_RANK_WORDCOUNT, ranking by the keyword occurrences count. This ranker computes the per-field keyword occurrence counts, then multiplies them by field weights, and sums the resulting values.

  • SPH_RANK_PROXIMITY, added in version 0.9.9-rc1, returns raw phrase proximity value as a result. This mode is internally used to emulate SPH_MATCH_ALL queries.

  • SPH_RANK_MATCHANY, added in version 0.9.9-rc1, returns rank as it was computed in SPH_MATCH_ANY mode ealier, and is internally used to emulate SPH_MATCH_ANY queries.

  • SPH_RANK_FIELDMASK, added in version 0.9.9-rc2, returns a 32-bit mask with N-th bit corresponding to N-th fulltext field, numbering from 0. The bit will only be set when the respective field has any keyword occurences satisfiying the query.

  • SPH_RANK_SPH04, added in version 1.10-beta, is generally based on the default SPH_RANK_PROXIMITY_BM25 ranker, but additionally boosts the matches when they occur in the very beginning or the very end of a text field. Thus, if a field equals the exact query, SPH04 should rank it higher than a field that contains the exact query but is not equal to it. (For instance, when the query is "Hyde Park", a document entitled "Hyde Park" should be ranked higher than a one entitled "Hyde Park, London" or "The Hyde Park Cafe".)

  • SPH_RANK_EXPR, added in version 2.0.2-beta, lets you specify the ranking formula in run time. It exposes a number of internal text factors and lets you define how the final weight should be computed from those factors. You can find more details about its syntax and a reference available factors in a subsection below.

You should specify the SPH_RANK_ prefix and use capital letters only when using the SetRankingMode() call from the SphinxAPI. The API ports expose these as global constants. Using SphinxQL syntax, the prefix should be omitted and the ranker name is case insensitive. Example:

// SphinxAPI
$client->SetRankingMode ( SPH_RANK_SPH04 );

// SphinxQL
mysql_query ( "SELECT ... OPTION ranker=sph04" );

Legacy matching modes rankers

Legacy matching modes automatically select a ranker as follows:





Expression based ranker (SPH_RANK_EXPR)

Expression ranker, added in version 2.0.2-beta, lets you change the ranking formula on the fly, on a per-query basis. For a quick kickoff, this is how you emulate PROXIMITY_BM25 ranker using the expression based one:

SELECT *, WEIGHT() FROM myindex WHERE MATCH('hello world')
OPTION ranker=expr('sum(lcs*user_weight)*1000+bm25')

The output of this query must not change if you omit the OPTION clause, because the default ranker (PROXIMITY_BM25) behaves exactly like specified in the ranker formula above. But the expression ranker is somewhat more flexible than just that and provides access to many more factors.

The ranking formula is an arbitrary arithmetic expression that can use constants, document attributes, built-in functions and operators (described in Section 5.5, “Expressions, functions, and operators”), and also a few ranking-specific things that are only accessible in a ranking formula. Namely, those are field aggregation functions, field-level, and document-level ranking factors.

A document-level factor is a numeric value computed by the ranking engine for every matched document with regards to the current query. (So it differs from a plain document attribute in that the attribute do not depend on the full text query, while factors might.) Those factors can be used anywhere in the ranking expression. Currently implemented document-level factors are:

  • bm25 (integer), a document-level BM25 estimate (computed without keyword occurrence filtering).

  • max_lcs (integer), a query-level maximum possible value that the sum(lcs*user_weight) expression can ever take. This can be useful for weight boost scaling. For instance, MATCHANY ranker formula uses this to guarantee that a full phrase match in any field rankes higher than any combination of partial matches in all fields.

  • field_mask (integer), a document-level 32-bit mask of matched fields.

  • query_word_count (integer), the number of unique keywords in a query, adjusted for a number of excluded keywords. For instance, both (one one one one) and (one !two) queries should assign a value of 1 to this factor, because there is just one unique non-excluded keyword.

  • doc_word_count (integer), the number of unique keywords matched in the entire document.

A field-level factor is a numeric value computed by the ranking engine for every matched in-document text field with regards to the current query. As more than one field can be matched by a query, but the final weight needs to be a single integer value, these values need to be folded into a single one. To achieve that, field-level factors can only be used within a field aggregation function, they can not be used anywhere in the expression. For example, you can not use (lcs+bm25) as your ranking expression, as lcs takes multiple values (one in every matched field). You should use (sum(lcs)+bm25) instead, that expression sums lcs over all matching fields, and then adds bm25 to that per-field sum. Currently implemented field-level factors are:

  • lcs (integer), the length of a maximum verbatim match between the document and the query, coutned in words. LCS stands for Longest Common Subsequence (or Subset). Takes a minimum value of 1 when only stray keywords were matched in a field, and a maximum value of query keywords count when the entire query was matched in a field verbatim (in the exact query keywords order). For example, if the query is 'hello world' and the field contains these two words quoted from the query (that is, adjacent to each other, and exaclty in the query order), lcs will be 2. For example, if the query is 'hello world program' and the field contains 'hello world', lcs will be 2. Note that any subset of the query keyword works, not just a subset of adjacent keywords. For example, if the query is 'hello world program' and the field contains 'hello (test program)', lcs will be 2 just as well, because both 'hello' and 'program' matched in the same respective positions as they were in the query. Finally, if the query is 'hello world program' and the field contains 'hello world program', lcs will be 3. (Hopefully that is unsurpising at this point.)

  • user_weight (integer), the user specified per-field weight (refer to SetFieldWeights() in SphinxAPI and OPTION field_weights in SphinxQL respectively). The weights default to 1 if not specified explicitly.

  • hit_count (integer), the number of keyword occurrences that matched in the field. Note that a single keyword may occur multiple times. For example, if 'hello' occurs 3 times in a field and 'world' occurs 5 times, hit_count will be 8.

  • word_count (integer), the number of unique keywords matched in the field. For example, if 'hello' and 'world' occur anywhere in a field, word_count will be 2, irregardless of how many times do both keywords occur.

  • tf_idf (float), the sum of TF*IDF over all the keywords matched in the field. IDF is the Inverse Document Frequency, a floating point value between 0 and 1 that describes how frequent is the keywords (basically, 0 for a keyword that occurs in every document indexed, and 1 for a unique keyword that occurs in just a single document). TF is the Term Frequency, the number of matched keyword occurrences in the field. As a side note, tf_idf is actually computed by summing IDF over all matched occurences. That's by construction equivalent to summing TF*IDF over all matched keywords.

  • min_hit_pos (integer), the position of the first matched keyword occurrence, counted in words. Indexing begins from position 1.

  • min_best_span_pos (integer), the position of the first maximum LCS occurrences span. For example, assume that our query was 'hello world program' and 'hello world' subphrase was matched twice in the field, in positions 13 and 21. Assume that 'hello' and 'world' additionally occurred elsewhere in the field, but never next to each other and thus never as a subphrase match. In that case, min_best_span_pos will be 13. Note how for the single keyword queries min_best_span_pos will always equal min_hit_pos.

  • exact_hit (boolean), whether a query was an exact match of the entire current field. Used in the SPH04 ranker.

A field aggregation function is a single argument function that takes an expression with field-level factors, iterates it over all the matched fields, and computes the final results. Currently implemented field aggregation functions are:

  • sum, sums the argument expression over all matched fields. For instance, sum(1) should return a number of matched fields.

Expressions for the built-in rankers

Most of the other rankers can actually be emulated with the expression based ranker. You just need to pass a proper expression. Such emulation is, of course, going to be slower than using the built-in, compiled ranker but still might be of interest if you want to fine-tune your ranking formula starting with one of the existing ones. Also, the formulas define the nitty gritty ranker details in a nicely readable fashion.

  • SPH_RANK_PROXIMITY_BM25 = sum(lcs*user_weight)*1000+bm25

  • SPH_RANK_BM25 = bm25


  • SPH_RANK_WORDCOUNT = sum(hit_count*user_weight)

  • SPH_RANK_PROXIMITY = sum(lcs*user_weight)

  • SPH_RANK_MATCHANY = sum((word_count+(lcs-1)*max_lcs)*user_weight)

  • SPH_RANK_FIELDMASK = field_mask

  • SPH_RANK_SPH04 = sum((4*lcs+2*(min_hit_pos==1)+exact_hit)*user_weight)*1000+bm25

5.5. Expressions, functions, and operators

Sphinx lets you use arbitrary arithmetic expressions both via SphinxQL and SphinxAPI, involving attribute values, internal attributes (document ID and relevance weight), arithmetic operations, a number of built-in functions, and user-defined functions. This section documents the supported operators and functions. Here's the complete reference list for quick access.

5.5.1. Operators

Arithmetic operators: +, -, *, /, %, DIV, MOD

The standard arithmetic operators. Arithmetic calculations involving those can be performed in three different modes: (a) using single-precision, 32-bit IEEE 754 floating point values (the default), (b) using signed 32-bit integers, (c) using 64-bit signed integers. The expression parser will automatically switch to integer mode if there are no operations the result in a floating point value. Otherwise, it will use the default floating point mode. For instance, a+b will be computed using 32-bit integers if both arguments are 32-bit integers; or using 64-bit integers if both arguments are integers but one of them is 64-bit; or in floats otherwise. However, a/b or sqrt(a) will always be computed in floats, because these operations return a result of non-integer type. To avoid the first, you can either use IDIV(a,b) or a DIV b form. Also, a*b will not be automatically promoted to 64-bit when the arguments are 32-bit. To enforce 64-bit results, you can use BIGINT(). (But note that if there are non-integer operations, BIGINT() will simply be ignored.)

Comparison operators: <, > <=, >=, =, <>

Comparison operators (eg. = or <=) return 1.0 when the condition is true and 0.0 otherwise. For instance, (a=b)+3 will evaluate to 4 when attribute 'a' is equal to attribute 'b', and to 3 when 'a' is not. Unlike MySQL, the equality comparisons (ie. = and <> operators) introduce a small equality threshold (1e-6 by default). If the difference between compared values is within the threshold, they will be considered equal.

Boolean operators: AND, OR, NOT

Boolean operators (AND, OR, NOT) were introduced in 0.9.9-rc2 and behave as usual. They are left-associative and have the least priority compared to other operators. NOT has more priority than AND and OR but nevertheless less than any other operator. AND and OR have the same priority so brackets use is recommended to avoid confusion in complex expressions.

Bitwise operators: &, |

These operators perform bitwise AND and OR respectively. The operands must be of an integer types. Introduced in version 1.10-beta.

5.5.2. Numeric functions


Returns the absolute value of the argument.


Returns the smallest integer value greater or equal to the argument.


Returns the cosine of the argument.


Returns the exponent of the argument (e=2.718... to the power of the argument).


Returns the N-th Fibonacci number, where N is the integer argument. That is, arguments of 0 and up will generate the values 0, 1, 1, 2, 3, 5, 8, 13 and so on. Note that the computations are done using 32-bit integer math and thus numbers 48th and up will be returned modulo 2^32.


Returns the largest integer value lesser or equal to the argument.


Returns the result of an integer division of the first argument by the second argument. Both arguments must be of an integer type.


Returns the natural logarithm of the argument (with the base of e=2.718...).


Returns the common logarithm of the argument (with the base of 10).


Returns the binary logarithm of the argument (with the base of 2).


Returns the bigger of two arguments.


Returns the smaller of two arguments.


Returns the first argument raised to the power of the second argument.


Returns the sine of the argument.


Returns the square root of the argument.

5.5.3. Date and time functions


Returns the integer day of month (in 1..31 range) from a timestamp argument, according to the current timezone. Introduced in version 2.0.1-beta.


Returns the integer month (in 1..12 range) from a timestamp argument, according to the current timezone. Introduced in version 2.0.1-beta.


Returns the current timestamp as an INTEGER. Introduced in version 0.9.9-rc1.


Returns the integer year (in 1969..2038 range) from a timestamp argument, according to the current timezone. Introduced in version 2.0.1-beta.


Returns the integer year and month code (in 196912..203801 range) from a timestamp argument, according to the current timezone. Introduced in version 2.0.1-beta.


Returns the integer year, month, and date code (in 19691231..20380119 range) from a timestamp argument, according to the current timezone. Introduced in version 2.0.1-beta.

5.5.4. Type conversion functions


Forcibly promotes the integer argument to 64-bit type, and does nothing on floating point argument. It's intended to help enforce evaluation of certain expressions (such as a*b) in 64-bit mode even though all the arguments are 32-bit. Introduced in version 0.9.9-rc1.


Forcibly reinterprets its 32-bit unsigned integer argument as signed, and also expands it to 64-bit type (because 32-bit type is unsigned). It's easily illustrated by the following example: 1-2 normally evaluates to 4294967295, but SINT(1-2) evaluates to -1. Introduced in version 1.10-beta.

5.5.5. Comparison functions


IF() behavior is slightly different that that of its MySQL counterpart. It takes 3 arguments, check whether the 1st argument is equal to 0.0, returns the 2nd argument if it is not zero, or the 3rd one when it is. Note that unlike comparison operators, IF() does not use a threshold! Therefore, it's safe to use comparison results as its 1st argument, but arithmetic operators might produce unexpected results. For instance, the following two calls will produce different results even though they are logically equivalent:

IF ( sqrt(3)*sqrt(3)-3<>0, a, b )
IF ( sqrt(3)*sqrt(3)-3, a, b )

In the first case, the comparison operator <> will return 0.0 (false) because of a threshold, and IF() will always return 'b' as a result. In the second one, the same sqrt(3)*sqrt(3)-3 expression will be compared with zero without threshold by the IF() function itself. But its value will be slightly different from zero because of limited floating point calculations precision. Because of that, the comparison with 0.0 done by IF() will not pass, and the second variant will return 'a' as a result.


IN(expr,val1,val2,...), introduced in version 0.9.9-rc1, takes 2 or more arguments, and returns 1 if 1st argument (expr) is equal to any of the other arguments (val1..valN), or 0 otherwise. Currently, all the checked values (but not the expression itself!) are required to be constant. (Its technically possible to implement arbitrary expressions too, and that might be implemented in the future.) Constants are pre-sorted and then binary search is used, so IN() even against a big arbitrary list of constants will be very quick. Starting with 0.9.9-rc2, first argument can also be a MVA attribute. In that case, IN() will return 1 if any of the MVA values is equal to any of the other arguments. Starting with 2.0.1-beta, IN() also supports IN(expr,@uservar) syntax to check whether the value belongs to the list in the given global user variable.


INTERVAL(expr,point1,point2,point3,...), introduced in version 0.9.9-rc1, takes 2 or more arguments, and returns the index of the argument that is less than the first argument: it returns 0 if expr<point1, 1 if point1<=expr<point2, and so on. It is required that point1<point2<...<pointN for this function to work correctly.

5.5.6. Miscellaneous functions


Returns the CRC32 value of a string argument. Introduced in version 2.0.1-beta.


GEODIST(lat1,long1,lat2,long2) function, introduced in version 0.9.9-rc2, computes geosphere distance between two given points specified by their coordinates. Note that both latitudes and longitudes must be in radians and the result will be in meters. You can use arbitrary expression as any of the four coordinates. An optimized path will be selected when one pair of the arguments refers directly to a pair attributes and the other one is constant.

5.6. Sorting modes

There are the following result sorting modes available:

  • SPH_SORT_RELEVANCE mode, that sorts by relevance in descending order (best matches first);

  • SPH_SORT_ATTR_DESC mode, that sorts by an attribute in descending order (bigger attribute values first);

  • SPH_SORT_ATTR_ASC mode, that sorts by an attribute in ascending order (smaller attribute values first);

  • SPH_SORT_TIME_SEGMENTS mode, that sorts by time segments (last hour/day/week/month) in descending order, and then by relevance in descending order;

  • SPH_SORT_EXTENDED mode, that sorts by SQL-like combination of columns in ASC/DESC order;

  • SPH_SORT_EXPR mode, that sorts by an arithmetic expression.

SPH_SORT_RELEVANCE ignores any additional parameters and always sorts matches by relevance rank. All other modes require an additional sorting clause, with the syntax depending on specific mode. SPH_SORT_ATTR_ASC, SPH_SORT_ATTR_DESC and SPH_SORT_TIME_SEGMENTS modes require simply an attribute name. SPH_SORT_RELEVANCE is equivalent to sorting by "@weight DESC, @id ASC" in extended sorting mode, SPH_SORT_ATTR_ASC is equivalent to "attribute ASC, @weight DESC, @id ASC", and SPH_SORT_ATTR_DESC to "attribute DESC, @weight DESC, @id ASC" respectively.


In SPH_SORT_TIME_SEGMENTS mode, attribute values are split into so-called time segments, and then sorted by time segment first, and by relevance second.

The segments are calculated according to the current timestamp at the time when the search is performed, so the results would change over time. The segments are as follows:

  • last hour,

  • last day,

  • last week,

  • last month,

  • last 3 months,

  • everything else.

These segments are hardcoded, but it is trivial to change them if necessary.

This mode was added to support searching through blogs, news headlines, etc. When using time segments, recent records would be ranked higher because of segment, but withing the same segment, more relevant records would be ranked higher - unlike sorting by just the timestamp attribute, which would not take relevance into account at all.


In SPH_SORT_EXTENDED mode, you can specify an SQL-like sort expression with up to 5 attributes (including internal attributes), eg:

@relevance DESC, price ASC, @id DESC

Both internal attributes (that are computed by the engine on the fly) and user attributes that were configured for this index are allowed. Internal attribute names must start with magic @-symbol; user attribute names can be used as is. In the example above, @relevance and @id are internal attributes and price is user-specified.

Known internal attributes are:

  • @id (match ID)

  • @weight (match weight)

  • @rank (match weight)

  • @relevance (match weight)

  • @random (return results in random order)

@rank and @relevance are just additional aliases to @weight.


Expression sorting mode lets you sort the matches by an arbitrary arithmetic expression, involving attribute values, internal attributes (@id and @weight), arithmetic operations, and a number of built-in functions. Here's an example:

$cl->SetSortMode ( SPH_SORT_EXPR,
	"@weight + ( user_karma + ln(pageviews) )*0.1" );

The operators and functions supported in the expressions are discussed in a separate section, Section 5.5, “Expressions, functions, and operators”.

5.7. Grouping (clustering) search results

Sometimes it could be useful to group (or in other terms, cluster) search results and/or count per-group match counts - for instance, to draw a nice graph of how much maching blog posts were there per each month; or to group Web search results by site; or to group matching forum posts by author; etc.

In theory, this could be performed by doing only the full-text search in Sphinx and then using found IDs to group on SQL server side. However, in practice doing this with a big result set (10K-10M matches) would typically kill performance.

To avoid that, Sphinx offers so-called grouping mode. It is enabled with SetGroupBy() API call. When grouping, all matches are assigned to different groups based on group-by value. This value is computed from specified attribute using one of the following built-in functions:

  • SPH_GROUPBY_DAY, extracts year, month and day in YYYYMMDD format from timestamp;

  • SPH_GROUPBY_WEEK, extracts year and first day of the week number (counting from year start) in YYYYNNN format from timestamp;

  • SPH_GROUPBY_MONTH, extracts month in YYYYMM format from timestamp;

  • SPH_GROUPBY_YEAR, extracts year in YYYY format from timestamp;

  • SPH_GROUPBY_ATTR, uses attribute value itself for grouping.

The final search result set then contains one best match per group. Grouping function value and per-group match count are returned along as "virtual" attributes named @group and @count respectively.

The result set is sorted by group-by sorting clause, with the syntax similar to SPH_SORT_EXTENDED sorting clause syntax. In addition to @id and @weight, group-by sorting clause may also include:

  • @group (groupby function value),

  • @count (amount of matches in group).

The default mode is to sort by groupby value in descending order, ie. by "@group desc".

On completion, total_found result parameter would contain total amount of matching groups over he whole index.

WARNING: grouping is done in fixed memory and thus its results are only approximate; so there might be more groups reported in total_found than actually present. @count might also be underestimated. To reduce inaccuracy, one should raise max_matches. If max_matches allows to store all found groups, results will be 100% correct.

For example, if sorting by relevance and grouping by "published" attribute with SPH_GROUPBY_DAY function, then the result set will contain

  • one most relevant match per each day when there were any matches published,

  • with day number and per-day match count attached,

  • sorted by day number in descending order (ie. recent days first).

Starting with version 0.9.9-rc2, aggregate functions (AVG(), MIN(), MAX(), SUM()) are supported through SetSelect() API call when using GROUP BY.

5.8. Distributed searching

To scale well, Sphinx has distributed searching capabilities. Distributed searching is useful to improve query latency (ie. search time) and throughput (ie. max queries/sec) in multi-server, multi-CPU or multi-core environments. This is essential for applications which need to search through huge amounts data (ie. billions of records and terabytes of text).

The key idea is to horizontally partition (HP) searched data accross search nodes and then process it in parallel.

Partitioning is done manually. You should

  • setup several instances of Sphinx programs (indexer and searchd) on different servers;

  • make the instances index (and search) different parts of data;

  • configure a special distributed index on some of the searchd instances;

  • and query this index.

This index only contains references to other local and remote indexes - so it could not be directly reindexed, and you should reindex those indexes which it references instead.

When searchd receives a query against distributed index, it does the following:

  1. connects to configured remote agents;

  2. issues the query;

  3. sequentially searches configured local indexes (while the remote agents are searching);

  4. retrieves remote agents' search results;

  5. merges all the results together, removing the duplicates;

  6. sends the merged resuls to client.

From the application's point of view, there are no differences between searching through a regular index, or a distributed index at all. That is, distributed indexes are fully transparent to the application, and actually there's no way to tell whether the index you queried was distributed or local. (Even though as of 0.9.9 Sphinx does not allow to combine searching through distributed indexes with anything else, this constraint will be lifted in the future.)

Any searchd instance could serve both as a master (which aggregates the results) and a slave (which only does local searching) at the same time. This has a number of uses:

  1. every machine in a cluster could serve as a master which searches the whole cluster, and search requests could be balanced between masters to achieve a kind of HA (high availability) in case any of the nodes fails;

  2. if running within a single multi-CPU or multi-core machine, there would be only 1 searchd instance quering itself as an agent and thus utilizing all CPUs/core.

It is scheduled to implement better HA support which would allow to specify which agents mirror each other, do health checks, keep track of alive agents, load-balance requests, etc.

5.9. searchd query log formats

In version 2.0.1-beta and above two query log formats are supported. Previous versions only supported a custom plain text format. That format is still the default one. However, while it might be more convenient for manual monitoring and review, but hard to replay for benchmarks, it only logs search queries but not the other types of requests, does not always contain the complete search query data, etc. The default text format is also harder (and sometimes impossible) to replay for benchmarking purposes. The new sphinxql format alleviates that. It aims to be complete and automatable, even though at the cost of brevity and readability.

5.9.1. Plain log format

By default, searchd logs all succesfully executed search queries into a query log file. Here's an example:

[Fri Jun 29 21:17:58 2007] 0.004 sec [all/0/rel 35254 (0,20)] [lj] test
[Fri Jun 29 21:20:34 2007] 0.024 sec [all/0/rel 19886 (0,20) @channel_id] [lj] test

This log format is as follows:

[query-date] query-time [match-mode/filters-count/sort-mode
    total-matches (offset,limit) @groupby-attr] [index-name] query

Match mode can take one of the following values:

  • "all" for SPH_MATCH_ALL mode;

  • "any" for SPH_MATCH_ANY mode;

  • "phr" for SPH_MATCH_PHRASE mode;

  • "bool" for SPH_MATCH_BOOLEAN mode;

  • "ext" for SPH_MATCH_EXTENDED mode;

  • "ext2" for SPH_MATCH_EXTENDED2 mode;

  • "scan" if the full scan mode was used, either by being specified with SPH_MATCH_FULLSCAN, or if the query was empty (as documented under Matching Modes)

Sort mode can take one of the following values:

  • "rel" for SPH_SORT_RELEVANCE mode;

  • "attr-" for SPH_SORT_ATTR_DESC mode;

  • "attr+" for SPH_SORT_ATTR_ASC mode;

  • "tsegs" for SPH_SORT_TIME_SEGMENTS mode;

  • "ext" for SPH_SORT_EXTENDED mode.

Additionally, if searchd was started with --iostats, there will be a block of data after where the index(es) searched are listed.

A query log entry might take the form of:

[Fri Jun 29 21:17:58 2007] 0.004 sec [all/0/rel 35254 (0,20)] [lj]
   [ios=6 kb=111.1 ms=0.5] test

This additional block is information regarding I/O operations in performing the search: the number of file I/O operations carried out, the amount of data in kilobytes read from the index files and time spent on I/O operations (although there is a background processing component, the bulk of this time is the I/O operation time).

5.9.2. SphinxQL log format

This is a new log format introduced in 2.0.1-beta, with the goals begin logging everything and then some, and in a format easy to automate (for insance, automatically replay). New format can either be enabled via the query_log_format directive in the configuration file, or switched back and forth on the fly with the SET GLOBAL query_log_format=... statement via SphinxQL. In the new format, the example from the previous section would look as follows. (Wrapped below for readability, but with just one query per line in the actual log.)

/* Fri Jun 29 21:17:58.609 2007 2011 conn 2 wall 0.004 found 35254 */
SELECT * FROM lj WHERE MATCH('test') OPTION ranker=proximity;

/* Fri Jun 29 21:20:34 2007.555 conn 3 wall 0.024 found 19886 */
SELECT * FROM lj WHERE MATCH('test') GROUP BY channel_id
OPTION ranker=proximity;

Note that all requests would be logged in this format, including those sent via SphinxAPI and SphinxSE, not just those sent via SphinxQL. Also note, that this kind of logging works only with plain log files and will not work if you use 'syslog' for logging.

The features of SphinxQL log format compared to the default text one are as follows.

  • All request types should be logged. (This is still work in progress.)

  • Full statement data will be logged where possible.

  • Errors and warnings are logged.

  • The log should be automatically replayable via SphinxQL.

  • Additional performance counters (currently, per-agent distributed query times) are logged.

Every request (including both SphinxAPI and SphinxQL) request must result in exactly one log line. All request types, including INSERT, CALL SNIPPETS, etc will eventually get logged, though as of time of this writing, that is a work in progress). Every log line must be a valid SphinxQL statement that reconstructs the full request, except if the logged request is too big and needs shortening for performance reasons. Additional messages, counters, etc can be logged in the comments section after the request.

5.10. MySQL protocol support and SphinxQL

Starting with version 0.9.9-rc2, Sphinx searchd daemon supports MySQL binary network protocol and can be accessed with regular MySQL API. For instance, 'mysql' CLI client program works well. Here's an example of querying Sphinx using MySQL client:

$ mysql -P 9306
Welcome to the MySQL monitor.  Commands end with ; or \g.
Your MySQL connection id is 1
Server version: 0.9.9-dev (r1734)

Type 'help;' or '\h' for help. Type '\c' to clear the buffer.

mysql> SELECT * FROM test1 WHERE MATCH('test') 
    -> ORDER BY group_id ASC OPTION ranker=bm25;
| id   | weight | group_id | date_added |
|    4 |   1442 |        2 | 1231721236 |
|    2 |   2421 |      123 | 1231721236 |
|    1 |   2421 |      456 | 1231721236 |
3 rows in set (0.00 sec)

Note that mysqld was not even running on the test machine. Everything was handled by searchd itself.

The new access method is supported in addition to native APIs which all still work perfectly well. In fact, both access methods can be used at the same time. Also, native API is still the default access method. MySQL protocol support needs to be additionally configured. This is a matter of 1-line config change, adding a new listener with mysql41 specified as a protocol:

listen = localhost:9306:mysql41

Just supporting the protocol and not the SQL syntax would be useless so Sphinx now also supports a subset of SQL that we dubbed SphinxQL. It supports the standard querying all the index types with SELECT, modifying RT indexes with INSERT, REPLACE, and DELETE, and much more. Full SphinxQL reference is available in Chapter 7, SphinxQL reference.

5.11. Multi-queries

Multi-queries, or query batches, let you send multiple queries to Sphinx in one go (more formally, one network request).

Two API methods that implement multi-query mechanism are AddQuery() and RunQueries(). You can also run multiple queries with SphinxQL, see Section 7.21, “Multi-statement queries”. (In fact, regular Query() call is internally implemented as a single AddQuery() call immediately followed by RunQueries() call.) AddQuery() captures the current state of all the query settings set by previous API calls, and memorizes the query. RunQueries() actually sends all the memorized queries, and returns multiple result sets. There are no restrictions on the queries at all, except just a sanity check on a number of queries in a single batch (see Section 11.4.25, “max_batch_queries”).

Why use multi-queries? Generally, it all boils down to performance. First, by sending requests to searchd in a batch instead of one by one, you always save a bit by doing less network roundtrips. Second, and somewhat more important, sending queries in a batch enables searchd to perform certain internal optimizations. As new types of optimizations are being added over time, it generally makes sense to pack all the queries into batches where possible, so that simply upgrading Sphinx to a new version would automatically enable new optimizations. In the case when there aren't any possible batch optimizations to apply, queries will be processed one by one internally.

Why (or rather when) not use multi-queries? Multi-queries requires all the queries in a batch to be independent, and sometimes they aren't. That is, sometimes query B is based on query A results, and so can only be set up after executing query A. For instance, you might want to display results from a secondary index if and only if there were no results found in a primary index. Or maybe just specify offset into 2nd result set based on the amount of matches in the 1st result set. In that case, you will have to use separate queries (or separate batches).

As of 0.9.10, there are two major optimizations to be aware of: common query optimization (available since 0.9.8); and common subtree optimization (available since 0.9.10).

Common query optimization means that searchd will identify all those queries in a batch where only the sorting and group-by settings differ, and only perform searching once. For instance, if a batch consists of 3 queries, all of them are for "ipod nano", but 1st query requests top-10 results sorted by price, 2nd query groups by vendor ID and requests top-5 vendors sorted by rating, and 3rd query requests max price, full-text search for "ipod nano" will only be performed once, and its results will be reused to build 3 different result sets.

So-called faceted searching is a particularly important case that benefits from this optimization. Indeed, faceted searching can be implemented by running a number of queries, one to retrieve search results themselves, and a few other ones with same full-text query but different group-by settings to retrieve all the required groups of results (top-3 authors, top-5 vendors, etc). And as long as full-text query and filtering settings stay the same, common query optimization will trigger, and greatly improve performance.

Common subtree optimization is even more interesting. It lets searchd exploit similarities between batched full-text queries. It identifies common full-text query parts (subtress) in all queries, and caches them between queries. For instance, look at the following query batch:

barack obama president
barack obama john mccain
barack obama speech

There's a common two-word part ("barack obama") that can be computed only once, then cached and shared across the queries. And common subtree optimization does just that. Per-query cache size is strictly controlled by subtree_docs_cache and subtree_hits_cache directives (so that caching all sxiteen gazillions of documents that match "i am" does not exhaust the RAM and instantly kill your server).

Here's a code sample (in PHP) that fire the same query in 3 different sorting modes:

require ( "sphinxapi.php" );
$cl = new SphinxClient ();
$cl->SetMatchMode ( SPH_MATCH_EXTENDED );

$cl->SetSortMode ( SPH_SORT_RELEVANCE );
$cl->AddQuery ( "the", "lj" );
$cl->SetSortMode ( SPH_SORT_EXTENDED, "published desc" );
$cl->AddQuery ( "the", "lj" );
$cl->SetSortMode ( SPH_SORT_EXTENDED, "published asc" );
$cl->AddQuery ( "the", "lj" );
$res = $cl->RunQueries();

How to tell whether the queries in the batch were actually optimized? If they were, respective query log will have a "multiplier" field that specifies how many queries were processed together:

[Sun Jul 12 15:18:17.000 2009] 0.040 sec x3 [ext/0/rel 747541 (0,20)] [lj] the
[Sun Jul 12 15:18:17.000 2009] 0.040 sec x3 [ext/0/ext 747541 (0,20)] [lj] the
[Sun Jul 12 15:18:17.000 2009] 0.040 sec x3 [ext/0/ext 747541 (0,20)] [lj] the

Note the "x3" field. It means that this query was optimized and processed in a sub-batch of 3 queries. For reference, this is how the regular log would look like if the queries were not batched:

[Sun Jul 12 15:18:17.062 2009] 0.059 sec [ext/0/rel 747541 (0,20)] [lj] the
[Sun Jul 12 15:18:17.156 2009] 0.091 sec [ext/0/ext 747541 (0,20)] [lj] the
[Sun Jul 12 15:18:17.250 2009] 0.092 sec [ext/0/ext 747541 (0,20)] [lj] the

Note how per-query time in multi-query case was improved by a factor of 1.5x to 2.3x, depending on a particular sorting mode. In fact, for both common query and common subtree optimizations, there were reports of 3x and even more improvements, and that's from production instances, not just synthetic tests.

5.12. Collations

Introduced to Sphinx in version 2.0.1-beta to supplement string sorting, collations essentially affect the string attribute comparisons. They specify both the character set encoding and the strategy that Sphinx uses to compare strings when doing ORDER BY or GROUP BY with a string attribute involved.

String attributes are stored as is when indexing, and no character set or language information is attached to them. That's okay as long as Sphinx only needs to store and return the strings to the calling application verbatim. But when you ask Sphinx to sort by a string value, that request immediately becomes quite ambiguous.

First, single-byte (ASCII, or ISO-8859-1, or Windows-1251) strings need to be processed differently that the UTF-8 ones that may encode every character with a variable number of bytes. So we need to know what is the character set type to interepret the raw bytes as meaningful characters properly.

Second, we additionally need to know the language-specific string sorting rules. For instance, when sorting according to US rules in en_US locale, the accented character 'ï' (small letter i with diaeresis) should be placed somewhere after 'z'. However, when sorting with French rules and fr_FR locale in mind, it should be placed between 'i' and 'j'. And some other set of rules might choose to ignore accents at all, allowing 'ï' and 'i' to be mixed arbitrarily.

Third, but not least, we might need case-sensitive sorting in some scenarios and case-insensitive sorting in some others.

Collations combine all of the above: the character set, the lanugage rules, and the case sensitivity. Sphinx currently provides the following four collations.

  1. libc_ci

  2. libc_cs

  3. utf8_general_ci

  4. binary

The first two collations rely on several standard C library (libc) calls and can thus support any locale that is installed on your system. They provide case-insensitive (_ci) and case-sensitive (_cs) comparisons respectively. By default they will use C locale, effectively resorting to bytewise comparisons. To change that, you need to specify a different available locale using collation_libc_locale directive. The list of locales available on your system can usually be obtained with the locale command:

$ locale -a

The specific list of the system locales may vary. Consult your OS documentation to install additional needed locales.

utf8_general_ci and binary locales are built-in into Sphinx. The first one is a generic collation for UTF-8 data (without any so-called language tailoring); it should behave similar to utf8_general_ci collation in MySQL. The second one is a simple bytewise comparison.

Collation can be overriden via SphinxQL on a per-session basis using SET collation_connection statement. All subsequent SphinxQL queries will use this collation. SphinxAPI and SphinxSE queries will use the server default collation, as specified in collation_server configuration directive. Sphinx currently defaults to libc_ci collation.

Collations should affect all string attribute comparisons, including those within ORDER BY and GROUP BY, so differently ordered or grouped results can be returned depending on the collation chosen.

5.13. User-defined functions (UDF)

Starting with 2.0.1-beta, Sphinx supports User-Defined Functions, or UDF for short. They can be loaded and unloaded dynamically into searchd without having to restart the daemon, and used in expressions when searching. UDF features at a glance are as follows.

  • Functions can take integer (both 32-bit and 64-bit), float, string, or MVA arguments.

  • Functions can return integer or float values.

  • Functions can check the argument number, types, and names and raise errors.

  • Only simple functions (that is, non-aggregate ones) are currently supported.

User-defined functions need your OS to support dynamically loadable libraries (aka shared objects). Most of the modern OSes are eligible, including Linux, Windows, MacOS, Solaris, BSD and others. (The internal testing has been done on Linux and Windows.) The UDF libraries must reside in a directory specified by plugin_dir directive, and the server must be configured to use workers = threads mode. Relative paths to the library files are not allowed. Once the library is succesfully built and copied to the trusted location, you can then dynamically install and deinstall the functions using CREATE FUNCTION and DROP FUNCTION statements respectively. A single library can contain multiple functions. A library gets loaded when you first install a function from it, and unloaded when you deinstall all the functions from that library.

The library functions that will implement a UDF visible to SQL statements need to follow C calling convention, and a simple naming convention. Sphinx source distribution provides a sample file, src/udfexample.c, that defines a few simple functions showing how to work with integer, string, and MVA arguments; you can use that one as a foundation for your new functions. It includes the UDF interface header file, src/sphinxudf.h, that defines the required types and structures. sphinxudf.h header is standalone, that is, does not require any other parts of Sphinx source to compile.

Every function that you intend to use in your SELECT statements requires at least two corresponding C/C++ functions: the initialization call, and the function call itself. You can also optionally define the deinitialization call if your function requires any post-query cleanup. (For instance, if you were allocating any memory in either the initialization call or the function calls.) Function names in SQL are case insensitive, C function names are not. They need to be all lower-case. Mistakes in function name prevent UDFs from loading. You also have to pay special attention to the calling convention used when compiling, the list and the types of arguments, and the return type of the main function call. Mistakes in either are likely to crash the server, or result in unexpected results in the best case. Last but not least, all functions need to be thread-safe.

Let's assume for the sake of example that your UDF name in SphinxQL will be MYFUNC. The initialization, main, and deinitialization functions would then need to be named as follows and take the following arguments:

/// initialization function
/// called once during query initialization
/// returns 0 on success
/// returns non-zero and fills error_message buffer on failure
int myfunc_init ( SPH_UDF_INIT * init, SPH_UDF_ARGS * args,
                  char * error_message );

/// main call function
/// returns the computed value
/// writes non-zero value into error_flag to indicate errors
RETURN_TYPE myfunc ( SPH_UDF_INIT * init, SPH_UDF_ARGS * args,
                     char * error_flag );

/// optional deinitialization function
/// called once to cleanup once query processing is done
void myfunc_deinit ( SPH_UDF_INIT * init );

The two mentioned structures, SPH_UDF_INIT and SPH_UDF_ARGS, are defined in the src/sphinxudf.h interface header and documented there. RETURN_TYPE of the main function must be one of the following:

  • int for the functions that return INT.

  • sphinx_int64_t for the functions that return BIGINT.

  • float for the functions that return FLOAT.

The calling sequence is as follows. myfunc_init() is called once when initializing the query. It can return a non-zero code to indicate a failure; in that case query is not executed, and the error message from the error_message buffer is returned. Otherwise, myfunc() is be called for every row, and a myfunc_deinit() is then called when the query ends. myfunc() can indicate an error by writing a non-zero byte value to error_flag, in that case, it will no more be called for subsequent rows, and a default value of 0 will be substituted. Sphinx might or might not choose to terminate such queries early, neither behavior is currently guaranteed.

Chapter 6. Command line tools reference

As mentioned elsewhere, Sphinx is not a single program called 'sphinx', but a collection of 4 separate programs which collectively form Sphinx. This section covers these tools and how to use them.

6.1. indexer command reference

indexer is the first of the two principle tools as part of Sphinx. Invoked from either the command line directly, or as part of a larger script, indexer is solely responsible for gathering the data that will be searchable.

The calling syntax for indexer is as follows:

indexer [OPTIONS] [indexname1 [indexname2 [...]]]

Essentially you would list the different possible indexes (that you would later make available to search) in sphinx.conf, so when calling indexer, as a minimum you need to be telling it what index (or indexes) you want to index.

If sphinx.conf contained details on 2 indexes, mybigindex and mysmallindex, you could do the following:

$ indexer mybigindex
$ indexer mysmallindex mybigindex

As part of the configuration file, sphinx.conf, you specify one or more indexes for your data. You might call indexer to reindex one of them, ad-hoc, or you can tell it to process all indexes - you are not limited to calling just one, or all at once, you can always pick some combination of the available indexes.

The majority of the options for indexer are given in the configuration file, however there are some options you might need to specify on the command line as well, as they can affect how the indexing operation is performed. These options are:

  • --config <file> (-c <file> for short) tells indexer to use the given file as its configuration. Normally, it will look for sphinx.conf in the installation directory (e.g. /usr/local/sphinx/etc/sphinx.conf if installed into /usr/local/sphinx), followed by the current directory you are in when calling indexer from the shell. This is most of use in shared environments where the binary files are installed somewhere like /usr/local/sphinx/ but you want to provide users with the ability to make their own custom Sphinx set-ups, or if you want to run multiple instances on a single server. In cases like those you could allow them to create their own sphinx.conf files and pass them to indexer with this option. For example:

    $ indexer --config /home/myuser/sphinx.conf myindex

  • --all tells indexer to update every index listed in sphinx.conf, instead of listing individual indexes. This would be useful in small configurations, or cron-type or maintenance jobs where the entire index set will get rebuilt each day, or week, or whatever period is best. Example usage:

    $ indexer --config /home/myuser/sphinx.conf --all

  • --rotate is used for rotating indexes. Unless you have the situation where you can take the search function offline without troubling users, you will almost certainly need to keep search running whilst indexing new documents. --rotate creates a second index, parallel to the first (in the same place, simply including .new in the filenames). Once complete, indexer notifies searchd via sending the SIGHUP signal, and searchd will attempt to rename the indexes (renaming the existing ones to include .old and renaming the .new to replace them), and then start serving from the newer files. Depending on the setting of seamless_rotate, there may be a slight delay in being able to search the newer indexes. Example usage:

    $ indexer --rotate --all

  • --quiet tells indexer not to output anything, unless there is an error. Again, most used for cron-type, or other script jobs where the output is irrelevant or unnecessary, except in the event of some kind of error. Example usage:

    $ indexer --rotate --all --quiet

  • --noprogress does not display progress details as they occur; instead, the final status details (such as documents indexed, speed of indexing and so on are only reported at completion of indexing. In instances where the script is not being run on a console (or 'tty'), this will be on by default. Example usage:

    $ indexer --rotate --all --noprogress

  • --buildstops <outputfile.text> <N> reviews the index source, as if it were indexing the data, and produces a list of the terms that are being indexed. In other words, it produces a list of all the searchable terms that are becoming part of the index. Note; it does not update the index in question, it simply processes the data 'as if' it were indexing, including running queries defined with sql_query_pre or sql_query_post. outputfile.txt will contain the list of words, one per line, sorted by frequency with most frequent first, and N specifies the maximum number of words that will be listed; if sufficiently large to encompass every word in the index, only that many words will be returned. Such a dictionary list could be used for client application features around "Did you mean..." functionality, usually in conjunction with --buildfreqs, below. Example:

    $ indexer myindex --buildstops word_freq.txt 1000

    This would produce a document in the current directory, word_freq.txt with the 1,000 most common words in 'myindex', ordered by most common first. Note that the file will pertain to the last index indexed when specified with multiple indexes or --all (i.e. the last one listed in the configuration file)

  • --buildfreqs works with --buildstops (and is ignored if --buildstops is not specified). As --buildstops provides the list of words used within the index, --buildfreqs adds the quantity present in the index, which would be useful in establishing whether certain words should be considered stopwords if they are too prevalent. It will also help with developing "Did you mean..." features where you can how much more common a given word compared to another, similar one. Example:

    $ indexer myindex --buildstops word_freq.txt 1000 --buildfreqs

    This would produce the word_freq.txt as above, however after each word would be the number of times it occurred in the index in question.

  • --merge <dst-index> <src-index> is used for physically merging indexes together, for example if you have a main+delta scheme, where the main index rarely changes, but the delta index is rebuilt frequently, and --merge would be used to combine the two. The operation moves from right to left - the contents of src-index get examined and physically combined with the contents of dst-index and the result is left in dst-index. In pseudo-code, it might be expressed as: dst-index += src-index An example:

    $ indexer --merge main delta --rotate

    In the above example, where the main is the master, rarely modified index, and delta is the less frequently modified one, you might use the above to call indexer to combine the contents of the delta into the main index and rotate the indexes.

  • --merge-dst-range <attr> <min> <max> runs the filter range given upon merging. Specifically, as the merge is applied to the destination index (as part of --merge, and is ignored if --merge is not specified), indexer will also filter the documents ending up in the destination index, and only documents will pass through the filter given will end up in the final index. This could be used for example, in an index where there is a 'deleted' attribute, where 0 means 'not deleted'. Such an index could be merged with:

    $ indexer --merge main delta --merge-dst-range deleted 0 0

    Any documents marked as deleted (value 1) would be removed from the newly-merged destination index. It can be added several times to the command line, to add successive filters to the merge, all of which must be met in order for a document to become part of the final index.

  • --dump-rows <FILE> dumps rows fetched by SQL source(s) into the specified file, in a MySQL compatible syntax. Resulting dumps are the exact representation of data as received by indexer and help to repeat indexing-time issues.

  • --verbose guarantees that every row that caused problems indexing (duplicate, zero, or missing document ID; or file field IO issues; etc) will be reported. By default, this option is off, and problem summaries may be reported instead.

  • --sighup-each is useful when you are rebuilding many big indexes, and want each one rotated into searchd as soon as possible. With --sighup-each, indexer will send a SIGHUP signal to searchd after succesfully completing the work on each index. (The default behavior is to send a single SIGHUP after all the indexes were built.)

  • --print-queries prints out SQL queries that indexer sends to the database, along with SQL connection and disconnection events. That is useful to diagnose and fix problems with SQL sources.

6.2. searchd command reference

searchd is the second of the two principle tools as part of Sphinx. searchd is the part of the system which actually handles searches; it functions as a server and is responsible for receiving queries, processing them and returning a dataset back to the different APIs for client applications.

Unlike indexer, searchd is not designed to be run either from a regular script or command-line calling, but instead either as a daemon to be called from init.d (on Unix/Linux type systems) or to be called as a service (on Windows-type systems), so not all of the command line options will always apply, and so will be build-dependent.

Calling searchd is simply a case of:

$ searchd [OPTIONS]

The options available to searchd on all builds are:

  • --help (-h for short) lists all of the parameters that can be called in your particular build of searchd.

  • --config <file> (-c <file> for short) tells searchd to use the given file as its configuration, just as with indexer above.

  • --stop is used to asynchronously stop searchd, using the details of the PID file as specified in the sphinx.conf file, so you may also need to confirm to searchd which configuration file to use with the --config option. NB, calling --stop will also make sure any changes applied to the indexes with UpdateAttributes() will be applied to the index files themselves. Example:

    $ searchd --config /home/myuser/sphinx.conf --stop

  • --stopwait is used to synchronously stop searchd. --stop essentially tells the running instance to exit (by sending it a SIGTERM) and then immediately returns. --stopwait will also attempt to wait until the running searchd instance actually finishes the shutdown (eg. saves all the pending attribute changes) and exits. Example:

    $ searchd --config /home/myuser/sphinx.conf --stopwait

    Possible exit codes are as follows:

    • 0 on success;

    • 1 if connection to running searchd daemon failed;

    • 2 if daemon reported an error during shutdown;

    • 3 if daemon crashed during shutdown.

  • --status command is used to query running searchd instance status, using the connection details from the (optionally) provided configuration file. It will try to connect to the running instance using the first configured UNIX socket or TCP port. On success, it will query for a number of status and performance counter values and print them. You can use Status() API call to access the very same counters from your application. Examples:

    $ searchd --status
    $ searchd --config /home/myuser/sphinx.conf --status

  • --pidfile is used to explicitly state a PID file, where the process information is stored regarding searchd, used for inter-process communications (for example, indexer will need to know the PID to contact searchd for rotating indexes). Normally, searchd would use a PID if running in regular mode (i.e. not with --console), but it is possible that you will be running it in console mode whilst the index is being updated and rotated, for which a PID file will be needed.

    $ searchd --config /home/myuser/sphinx.conf --pidfile /home/myuser/sphinx.pid

  • --console is used to force searchd into console mode; typically it will be running as a conventional server application, and will aim to dump information into the log files (as specified in sphinx.conf). Sometimes though, when debugging issues in the configuration or the daemon itself, or trying to diagnose hard-to-track-down problems, it may be easier to force it to dump information directly to the console/command line from which it is being called. Running in console mode also means that the process will not be forked (so searches are done in sequence) and logs will not be written to. (It should be noted that console mode is not the intended method for running searchd.) You can invoke it as such:

    $ searchd --config /home/myuser/sphinx.conf --console

  • --logdebug enables additional debug output in the daemon log. Should only be needed rarely, to assist with debugging issues that could not be easily reproduced on request.

  • --iostats is used in conjuction with the logging options (the query_log will need to have been activated in sphinx.conf) to provide more detailed information on a per-query basis as to the input/output operations carried out in the course of that query, with a slight performance hit and of course bigger logs. Further details are available under the query log format section. You might start searchd thus:

    $ searchd --config /home/myuser/sphinx.conf --iostats

  • --cpustats is used to provide actual CPU time report (in addition to wall time) in both query log file (for every given query) and status report (aggregated). It depends on clock_gettime() system call and might therefore be unavailable on certain systems. You might start searchd thus:

    $ searchd --config /home/myuser/sphinx.conf --cpustats

  • --port portnumber (-p for short) is used to specify the port that searchd should listen on, usually for debugging purposes. This will usually default to 9312, but sometimes you need to run it on a different port. Specifying it on the command line will override anything specified in the configuration file. The valid range is 0 to 65535, but ports numbered 1024 and below usually require a privileged account in order to run. An example of usage:

    $ searchd --port 9313

  • --listen ( address ":" port | port | path ) [ ":" protocol ] (or -l for short) Works as --port, but allow you to specify not only the port, but full path, as IP address and port, or Unix-domain socket path, that searchd will listen on. Otherwords, you can specify either an IP address (or hostname) and port number, or just a port number, or Unix socket path. If you specify port number but not the address, searchd will listen on all network interfaces. Unix path is identified by a leading slash. As the last param you can also specify a protocol handler (listener) to be used for connections on this socket. Supported protocol values are 'sphinx' (Sphinx 0.9.x API protocol) and 'mysql41' (MySQL protocol used since 4.1 upto at least 5.1).

  • --index <index> (or -i <index> for short) forces this instance of searchd only to serve the specified index. Like --port, above, this is usually for debugging purposes; more long-term changes would generally be applied to the configuration file itself. Example usage:

    $ searchd --index myindex

  • --strip-path strips the path names from all the file names referenced from the index (stopwords, wordforms, exceptions, etc). This is useful for picking up indexes built on another machine with possibly different path layouts.

  • --replay-flags=<OPTIONS> switch, added in version 2.0.2-beta, can be used to specify a list of extra binary log replay options. The supported options are:

    • accept-desc-timestamp, ignore descending transaction timestamps and replay such transactions anyway (the default behavior is to exit with an error).


    $ searchd --replay-flags=accept-desc-timestamp

There are some options for searchd that are specific to Windows platforms, concerning handling as a service, are only be available on Windows binaries.

Note that on Windows searchd will default to --console mode, unless you install it as a service.

  • --install installs searchd as a service into the Microsoft Management Console (Control Panel / Administrative Tools / Services). Any other parameters specified on the command line, where --install is specified will also become part of the command line on future starts of the service. For example, as part of calling searchd, you will likely also need to specify the configuration file with --config, and you would do that as well as specifying --install. Once called, the usual start/stop facilities will become available via the management console, so any methods you could use for starting, stopping and restarting services would also apply to searchd. Example:

    C:\WINDOWS\system32> C:\Sphinx\bin\searchd.exe --install
       --config C:\Sphinx\sphinx.conf

    If you wanted to have the I/O stats every time you started searchd, you would specify its option on the same line as the --install command thus:

    C:\WINDOWS\system32> C:\Sphinx\bin\searchd.exe --install
       --config C:\Sphinx\sphinx.conf --iostats

  • --delete removes the service from the Microsoft Management Console and other places where services are registered, after previously installed with --install. Note, this does not uninstall the software or delete the indexes. It means the service will not be called from the services systems, and will not be started on the machine's next start. If currently running as a service, the current instance will not be terminated (until the next reboot, or searchd is called with --stop). If the service was installed with a custom name (with --servicename), the same name will need to be specified with --servicename when calling to uninstall. Example:

    C:\WINDOWS\system32> C:\Sphinx\bin\searchd.exe --delete

  • --servicename <name> applies the given name to searchd when installing or deleting the service, as would appear in the Management Console; this will default to searchd, but if being deployed on servers where multiple administrators may log into the system, or a system with multiple searchd instances, a more descriptive name may be applicable. Note that unless combined with --install or --delete, this option does not do anything. Example:

    C:\WINDOWS\system32> C:\Sphinx\bin\searchd.exe --install
       --config C:\Sphinx\sphinx.conf --servicename SphinxSearch

  • --ntservice is the option that is passed by the Management Console to searchd to invoke it as a service on Windows platforms. It would not normally be necessary to call this directly; this would normally be called by Windows when the service would be started, although if you wanted to call this as a regular service from the command-line (as the complement to --console) you could do so in theory.

Last but not least, as every other daemon, searchd supports a number of signals.


Initiates a clean shutdown. New queries will not be handled; but queries that are already started will not be forcibly interrupted.


Initiates index rotation. Depending on the value of seamless_rotate setting, new queries might be shortly stalled; clients will receive temporary errors.


Forces reopen of searchd log and query log files, letting you implement log file rotation.

6.3. search command reference

search is one of the helper tools within the Sphinx package. Whereas searchd is responsible for searches in a server-type environment, search is aimed at testing the index from the command line, and testing the index quickly without building a framework to make the connection to the server and process its response.

Note: search is not intended to be deployed as part of a client application; it is strongly recommended you do not write an interface to search instead of searchd, and none of the bundled client APIs support this method. (In any event, search will reload files each time, whereas searchd will cache them in memory for performance.)

That said, many types of query that you could build in the APIs could also be made with search, however for very complex searches it may be easier to construct them using a small script and the corresponding API. Additionally, some newer features may be available in the searchd system that have not yet been brought into search.

The calling syntax for search is as follows:

search [OPTIONS] word1 [word2 [word3 [...]]]

When calling search, it is not necessary to have searchd running; simply make sure that the account running the search program has read access to the configuration file and the index files.

The default behaviour is to apply a search for word1 (AND word2 AND word3... as specified) to all fields in all indexes as given in the configuration file. If constructing the equivalent in the API, this would be the equivalent to passing SPH_MATCH_ALL to SetMatchMode, and specifying * as the indexes to query as part of Query.

There are many options available to search. Firstly, the general options:

  • --config <file> (-c <file> for short) tells search to use the given file as its configuration, just as with indexer above.

  • --index <index> (-i <index> for short) tells search to limit searching to the specified index only; normally it would attempt to search all of the physical indexes listed in sphinx.conf, not any distributed ones.

  • --stdin tells search to accept the query from the standard input, rather than the command line. This can be useful for testing purposes whereby you could feed input via pipes and from scripts.

Options for setting matches:

  • --any (-a for short) changes the matching mode to match any of the words as part of the query (word1 OR word2 OR word3). In the API this would be equivalent to passing SPH_MATCH_ANY to SetMatchMode.

  • --phrase (-p for short) changes the matching mode to match all of the words as part of the query, and do so in the phrase given (not including punctuation). In the API this would be equivalent to passing SPH_MATCH_PHRASE to SetMatchMode.

  • --boolean (-b for short) changes the matching mode to Boolean matching. Note if using Boolean syntax matching on the command line, you may need to escape the symbols (with a backslash) to avoid the shell/command line processor applying them, such as ampersands being escaped on a Unix/Linux system to avoid it forking to the search process, although this can be resolved by using --stdin, as below. In the API this would be equivalent to passing SPH_MATCH_BOOLEAN to SetMatchMode.

  • --ext (-e for short) changes the matching mode to extended matching which provides various text querying operators. In the API this would be equivalent to passing SPH_MATCH_EXTENDED to SetMatchMode.

  • --filter <attr> <v> (-f <attr> <v> for short) filters the results such that only documents where the attribute given (attr) matches the value given (v). For example, --filter deleted 0 only matches documents with an attribute called 'deleted' where its value is 0. You can also add multiple filters on the command line, by specifying multiple --filter multiple times, however if you apply a second filter to an attribute it will override the first defined filter.

Options for handling the results:

  • --limit <count> (-l count for short) limits the total number of matches back to the number given. If a 'group' is specified, this will be the number of grouped results. This defaults to 20 results if not specified (as do the APIs)

  • --offset <count> (-o <count> for short) offsets the result list by the number of places set by the count; this would be used for pagination through results, where if you have 20 results per 'page', the second page would begin at offset 20, the third page at offset 40, etc.

  • --group <attr> (-g <attr> for short) specifies that results should be grouped together based on the attribute specified. Like the GROUP BY clause in SQL, it will combine all results where the attribute given matches, and returns a set of results where each returned result is the best from each group. Unless otherwise specified, this will be the best match on relevance.

  • --groupsort <expr> (-gs <expr> for short) instructs that when results are grouped with --group, the expression given in <expr> shall determine the order of the groups. Note, this does not specify which is the best item within the group, only the order in which the groups themselves shall be returned.

  • --sortby <clause> (-s <clause> for short) specifies that results should be sorted in the order listed in <clause>. This allows you to specify the order you wish results to be presented in, ordering by different columns. For example, you could say --sortby "@weight DESC entrytime DESC" to sort entries first by weight (or relevance) and where two or more entries have the same weight, to then sort by the time with the highest time (newest) first. You will usually need to put the items in quotes (--sortby "@weight DESC") or use commas (--sortby @weight,DESC) to avoid the items being treated separately. Additionally, like the regular sorting modes, if --group (grouping) is being used, this will state how to establish the best match within each group.

  • --sortexpr expr (-S expr for short) specifies that the search results should be presented in an order determined by an arithmetic expression, stated in expr. For example: --sortexpr "@weight + ( user_karma + ln(pageviews) )*0.1" (again noting that this will have to be quoted to avoid the shell dealing with the asterisk). Extended sort mode is discussed in more detail under the SPH_SORT_EXTENDED entry under the Sorting modes section of the manual.

  • --sort=date specifies that the results should be sorted by descending (i.e. most recent first) date. This requires that there is an attribute in the index that is set as a timestamp.

  • --rsort=date specifies that the results should be sorted by ascending (i.e. oldest first) date. This requires that there is an attribute in the index that is set as a timestamp.

  • --sort=ts specifies that the results should be sorted by timestamp in groups; it will return all of the documents whose timestamp is within the last hour, then sorted within that bracket for relevance. After, it would return the documents from the last day, sorted by relevance, then the last week and then the last month. It is discussed in more detail under the SPH_SORT_TIME_SEGMENTS entry under the Sorting modes section of the manual.

Other options:

  • --noinfo (-q for short) instructs search not to look-up data in your SQL database. Specifically, for debugging with MySQL and search, you can provide it with a query to look up the full article based on the returned document ID. It is explained in more detail under the sql_query_info directive.

6.4. spelldump command reference

spelldump is one of the helper tools within the Sphinx package.

It is used to extract the contents of a dictionary file that uses ispell or MySpell format, which can help build word lists for wordforms - all of the possible forms are pre-built for you.

Its general usage is:

spelldump [options] <dictionary> <affix> [result] [locale-name]

The two main parameters are the dictionary's main file and its affix file; usually these are named as [language-prefix].dict and [language-prefix].aff and will be available with most common Linux distributions, as well as various places online.

[result] specifies where the dictionary data should be output to, and [locale-name] additionally specifies the locale details you wish to use.

There is an additional option, -c [file], which specifies a file for case conversion details.

Examples of its usage are:

spelldump en.dict en.aff
spelldump ru.dict ru.aff ru.txt ru_RU.CP1251
spelldump ru.dict ru.aff ru.txt .1251

The results file will contain a list of all the words in the dictionary in alphabetical order, output in the format of a wordforms file, which you can use to customise for your specific circumstances. An example of the result file:

zone > zone
zoned > zoned
zoning > zoning

6.5. indextool command reference

indextool is one of the helper tools within the Sphinx package, introduced in version 0.9.9-rc2. It is used to dump miscellaneous debug information about the physical index. (Additional functionality such as index verification is planned in the future, hence the indextool name rather than just indexdump.) Its general usage is:

indextool <command> [options]

The only currently available option applies to all commands and lets you specify the configuration file:

  • --config <file> (-c <file> for short) overrides the built-in config file names.

The commands are as follows:

  • --dumpheader FILENAME.sph quickly dumps the provided index header file without touching any other index files or even the configuration file. The report provides a breakdown of all the index settings, in particular the entire attribute and field list. Prior to 0.9.9-rc2, this command was present in CLI search utility.

  • --dumpconfig FILENAME.sph dumps the index definition from the given index header file in (almost) compliant sphinx.conf file format. Added in version 2.0.1-beta.

  • --dumpheader INDEXNAME dumps index header by index name with looking up the header path in the configuration file.

  • --dumpdocids INDEXNAME dumps document IDs by index name. It takes the data from attribute (.spa) file and therefore requires docinfo=extern to work.

  • --dumphitlist INDEXNAME KEYWORD dumps all the hits (occurences) of a given keyword in a given index, with keyword specified as text.

  • --dumphitlist INDEXNAME --wordid ID dumps all the hits (occurences) of a given keyword in a given index, with keyword specified as internal numeric ID.

  • --htmlstrip INDEXNAME filters stdin using HTML stripper settings for a given index, and prints the filtering results to stdout. Note that the settings will be taken from sphinx.conf, and not the index header.

  • --check INDEXNAME checks the index data files for consistency errors that might be introduced either by bugs in indexer and/or hardware faults. Starting with version 2.0.2-beta, --check also works on RT indexes, but checks disk chunks only.

  • --strip-path strips the path names from all the file names referenced from the index (stopwords, wordforms, exceptions, etc). This is useful for checking indexes built on another machine with possibly different path layouts.

Chapter 7. SphinxQL reference

SphinxQL is our SQL dialect that exposes all of the search daemon functionality using a standard SQL syntax with a few Sphinx-specific extensions. Everything available via the SphinxAPI is also available SphinxQL but not vice versa; for instance, writes into RT indexes are only available via SphinxQL. This chapter documents supported SphinxQL statements syntax.

7.1. SELECT syntax

    select_expr [, select_expr ...]
    FROM index [, index2 ...]
    [WHERE where_condition]
    [GROUP BY {col_name | expr_alias}]
    [ORDER BY {col_name | expr_alias} {ASC | DESC} [, ...]]
    [WITHIN GROUP ORDER BY {col_name | expr_alias} {ASC | DESC}]
    [LIMIT offset, row_count]
    [OPTION opt_name = opt_value [, ...]]

SELECT statement was introduced in version 0.9.9-rc2. It's syntax is based upon regular SQL but adds several Sphinx-specific extensions and has a few omissions (such as (currently) missing support for JOINs). Specifically,

  • Column list clause. Column names, arbitrary expressions, and star ('*') are all allowed (ie. SELECT @id, group_id*123+456 AS expr1 FROM test1 will work). Unlike in regular SQL, all computed expressions must be aliased with a valid identifier. Starting with version 2.0.1-beta, AS is optional. Special names such as @id and @weight should currently be used with leading at-sign. This at-sign requirement will be lifted in the future.

  • FROM clause. FROM clause should contain the list of indexes to search through. Unlike in regular SQL, comma means enumeration of full-text indexes as in Query() API call rather than JOIN.

  • WHERE clause. This clause will map both to fulltext query and filters. Comparison operators (=, !=, <, >, <=, >=), IN, AND, NOT, and BETWEEN are all supported and map directly to filters. OR is not supported yet but will be in the future. MATCH('query') is supported and maps to fulltext query. Query will be interpreted according to full-text query language rules. There must be at most one MATCH() in the clause. Starting with version 2.0.1-beta, {col_name | expr_alias} [NOT] IN @uservar condition syntax is supported. (Refer to Section 7.7, “SET syntax” for a discussion of global user variables.)

  • GROUP BY clause. Currently only supports grouping by a single column. The column however can be a computed expression:

    SELECT *, group_id*1000+article_type AS gkey FROM example GROUP BY gkey

    Aggregate functions (AVG(), MIN(), MAX(), SUM()) in column list clause are supported. Arguments to aggregate functions can be either plain attributes or arbitrary expressions. COUNT(*) is implicitly supported as using GROUP BY will add @count column to result set. Explicit support might be added in the future. COUNT(DISTINCT attr) is supported. Currently there can be at most one COUNT(DISTINCT) per query and an argument needs to be an attribute. Both current restrictions on COUNT(DISTINCT) might be lifted in the future.

    SELECT *, AVG(price) AS avgprice, COUNT(DISTINCT storeid)
    FROM products
    WHERE MATCH('ipod')
    GROUP BY vendorid

    Starting with 2.0.1-beta, GROUP BY on a string attribute is supported, with respect for current collation (see Section 5.12, “Collations”).

  • WITHIN GROUP ORDER BY clause. This is a Sphinx specific extension that lets you control how the best row within a group will to be selected. The syntax matches that of regular ORDER BY clause:

    SELECT *, INTERVAL(posted,NOW()-7*86400,NOW()-86400) AS timeseg
    FROM example WHERE MATCH('my search query')
    GROUP BY siteid
    ORDER BY timeseg DESC, @weight DESC

    Starting with 2.0.1-beta, WITHIN GROUP ORDER BY on a string attribute is supported, with respect for current collation (see Section 5.12, “Collations”).

  • ORDER BY clause. Unlike in regular SQL, only column names (not expressions) are allowed and explicit ASC and DESC are required. The columns however can be computed expressions:

    SELECT *, @weight*10+docboost AS skey FROM example ORDER BY skey

    Starting with 2.0.1-beta, ORDER BY on a string attribute is supported, with respect for current collation (see Section 5.12, “Collations”).

    Starting with 2.0.2-beta, ORDER BY RAND() syntax is supported. Note that this syntax is actually going to randomize the weight values and then order matches by those randomized weights.

  • LIMIT clause. Both LIMIT N and LIMIT M,N forms are supported. Unlike in regular SQL (but like in Sphinx API), an implicit LIMIT 0,20 is present by default.

  • OPTION clause. This is a Sphinx specific extension that lets you control a number of per-query options. The syntax is:

    OPTION <optionname>=<value> [ , ... ]

    Supported options and respectively allowed values are:

    • 'ranker' - any of 'proximity_bm25', 'bm25', 'none', 'wordcount', 'proximity', 'matchany', or 'fieldmask'

    • 'max_matches' - integer (per-query max matches value)

    • 'cutoff' - integer (max found matches threshold)

    • 'max_query_time' - integer (max search time threshold, msec)

    • 'retry_count' - integer (distributed retries count)

    • 'retry_delay' - integer (distributed retry delay, msec)

    • 'field_weights' - a named integer list (per-field user weights for ranking)

    • 'index_weights' - a named integer list (per-index user weights for ranking)

    • 'reverse_scan' - 0 or 1, lets you control the order in which full-scan query processes the rows

    • 'comment' - string, user comment that gets copied to a query log file


    SELECT * FROM test WHERE MATCH('@title hello @body world')
    OPTION ranker=bm25, max_matches=3000,
        field_weights=(title=10, body=3)

7.2. SHOW META syntax


SHOW META shows additional meta-information about the latest query such as query time and keyword statistics:

mysql> SELECT * FROM test1 WHERE MATCH('test|one|two');
| id   | weight | group_id | date_added |
|    1 |   3563 |      456 | 1231721236 |
|    2 |   2563 |      123 | 1231721236 |
|    4 |   1480 |        2 | 1231721236 |
3 rows in set (0.01 sec)

mysql> SHOW META;
| Variable_name | Value |
| total         | 3     |
| total_found   | 3     |
| time          | 0.005 |
| keyword[0]    | test  |
| docs[0]       | 3     |
| hits[0]       | 5     |
| keyword[1]    | one   |
| docs[1]       | 1     |
| hits[1]       | 2     |
| keyword[2]    | two   |
| docs[2]       | 1     |
| hits[2]       | 2     |
12 rows in set (0.00 sec)

7.3. SHOW WARNINGS syntax


SHOW WARNINGS statement, introduced in version 0.9.9-rc2, can be used to retrieve the warning produced by the latest query. The error message will be returned along with the query itself:

mysql> SELECT * FROM test1 WHERE MATCH('@@title hello') \G
ERROR 1064 (42000): index test1: syntax error, unexpected TOK_FIELDLIMIT
near '@title hello'

mysql> SELECT * FROM test1 WHERE MATCH('@title -hello') \G
ERROR 1064 (42000): index test1: query is non-computable (single NOT operator)

mysql> SELECT * FROM test1 WHERE MATCH('"test doc"/3') \G
*************************** 1. row ***************************
        id: 4
    weight: 2500
  group_id: 2
date_added: 1231721236
1 row in set, 1 warning (0.00 sec)

*************************** 1. row ***************************
  Level: warning
   Code: 1000
Message: quorum threshold too high (words=2, thresh=3); replacing quorum operator
         with AND operator
1 row in set (0.00 sec)

7.4. SHOW STATUS syntax

SHOW STATUS, introduced in version 0.9.9-rc2, displays a number of useful performance counters. IO and CPU counters will only be available if searchd was started with --iostats and --cpustats switches respectively.

| Variable_name      | Value |
| uptime             | 216   |
| connections        | 3     |
| maxed_out          | 0     |
| command_search     | 0     |
| command_excerpt    | 0     |
| command_update     | 0     |
| command_keywords   | 0     |
| command_persist    | 0     |
| command_status     | 0     |
| agent_connect      | 0     |
| agent_retry        | 0     |
| queries            | 10    |
| dist_queries       | 0     |
| query_wall         | 0.075 |
| query_cpu          | OFF   |
| dist_wall          | 0.000 |
| dist_local         | 0.000 |
| dist_wait          | 0.000 |
| query_reads        | OFF   |
| query_readkb       | OFF   |
| query_readtime     | OFF   |
| avg_query_wall     | 0.007 |
| avg_query_cpu      | OFF   |
| avg_dist_wall      | 0.000 |
| avg_dist_local     | 0.000 |
| avg_dist_wait      | 0.000 |
| avg_query_reads    | OFF   |
| avg_query_readkb   | OFF   |
| avg_query_readtime | OFF   |
29 rows in set (0.00 sec)

7.5. INSERT and REPLACE syntax

{INSERT | REPLACE} INTO index [(column, ...)]
	VALUES (value, ...)
	[, (...)]

INSERT statement, introduced in version 1.10-beta, is only supported for RT indexes. It inserts new rows (documents) into an existing index, with the provided column values.

ID column must be present in all cases. Rows with duplicate IDs will not be overwritten by INSERT; use REPLACE to do that.

index is the name of RT index into which the new row(s) should be inserted. The optional column names list lets you only explicitly specify values for some of the columns present in the index. All the other columns will be filled with their default values (0 for scalar types, empty string for text types).

Expressions are not currently supported in INSERT and values should be explicitly specified.

Multiple rows can be inserted using a single INSERT statement by providing several comma-separated, parens-enclosed lists of rows values.

7.6. DELETE syntax

DELETE FROM index WHERE {id = value | id IN (val1 [, val2 [, ...]])}

DELETE statement, introduced in version 1.10-beta, is only supported for RT indexes. It deletes existing rows (documents) from an existing index based on ID.

index is the name of RT index from which the row should be deleted. value is the row ID to be deleted. Support for batch id IN (2,3,5) syntax was added in version 2.0.1-beta.

Additional types of WHERE conditions (such as conditions on attributes, etc) are planned, but not supported yet as of 1.10-beta.

7.7. SET syntax

SET [GLOBAL] server_variable_name = value
SET GLOBAL @user_variable_name = (int_val1 [, int_val2, ...])
SET @@dummy_variable = ignored_value

SET statement, introduced in version 1.10-beta, modifies a variable value. The variable names are case-insensitive. No variable value changes survive server restart.

SET NAMES statement and SET @@variable_name syntax, both introduced in version 2.0.2-beta, do nothing. They were implemented to maintain compatibility with 3rd party MySQL client libraries, connectors, and frameworks that may need to run this statement when connecting.

There are the following classes of the variables:

  1. per-session server variable (1.10-beta and above)

  2. global server variable (2.0.1-beta and above)

  3. global user variable (2.0.1-beta and above)

Global user variables are shared between concurrent sessions. Currently, the only supported value type is the list of BIGINTs, and these variables can only be used along with IN() for filtering purpose. The intended usage scenario is uploading huge lists of values to searchd (once) and reusing them (many times) later, saving on network overheads. Example:

// in session 1
mysql> SET GLOBAL @myfilter=(2,3,5,7,11,13);
Query OK, 0 rows affected (0.00 sec)

// later in session 2
mysql> SELECT * FROM test1 WHERE group_id IN @myfilter;
| id   | weight | group_id | date_added | title           | tag  |
|    3 |      1 |        2 | 1299338153 | another doc     | 15   |
|    4 |      1 |        2 | 1299338153 | doc number four | 7,40 |
2 rows in set (0.02 sec)

Per-session and global server variables affect certain server settings in the respective scope. Known per-session server variables are:

AUTOCOMMIT = {0 | 1}

Whether any data modification statement should be implicitly wrapped by BEGIN and COMMIT. Introduced in version 1.10-beta.


Selects the collation to be used for ORDER BY or GROUP BY on string values in the subsequent queries. Refer to Section 5.12, “Collations” for a list of known collation names. Introduced in version 2.0.1-beta.


Does nothing; a placeholder to support frameworks, clients, and connectors that attempt to automatically enforce a charset when connecting to a Sphinx server. Introduced in version 2.0.1-beta.


Does nothing; a placeholder to support frameworks, clients, and connectors that attempt to automatically enforce a charset when connecting to a Sphinx server. Introduced in version 2.0.2-beta.

SQL_MODE = value

Does nothing; a placeholder to support frameworks, clients, and connectors that attempt to automatically enforce a charset when connecting to a Sphinx server. Introduced in version 2.0.2-beta.

Known global server variables are:

QUERY_LOG_FORMAT = {plain | sphinxql}

Changes the current log format. Introduced in version 2.0.1-beta.

LOG_LEVEL = {info | debug | debugv | debugvv}

Changes the current log verboseness level. Introduced in version 2.0.1-beta.


mysql> SET autocommit=0;
Query OK, 0 rows affected (0.00 sec)

mysql> SET GLOBAL query_log_format=sphinxql;
Query OK, 0 rows affected (0.00 sec)



SET TRANSACTION statement, introduced in version 2.0.2-beta, does nothing. It was implemented to maintain compatibility with 3rd party MySQL client libraries, connectors, and frameworks that may need to run this statement when connecting.


Query OK, 0 rows affected (0.00 sec)

7.9. BEGIN, COMMIT, and ROLLBACK syntax


BEGIN, COMMIT, and ROLLBACK statements were introduced in version 1.10-beta. BEGIN statement (or its START TRANSACTION alias) forcibly commits pending transaction, if any, and begins a new one. COMMIT statement commits the current transaction, making all its changes permanent. ROLLBACK statement rolls back the current transaction, canceling all its changes. SET AUTOCOMMIT controls the autocommit mode in the active session.

AUTOCOMMIT is set to 1 by default, meaning that every statement that perfoms any changes on any index is implicitly wrapped in BEGIN and COMMIT.

Transactions are limited to a single RT index, and also limited in size. They are atomic, consistent, overly isolated, and durable. Overly isolated means that the changes are not only invisible to the concurrent transactions but even to the current session itself.

7.10. CALL SNIPPETS syntax

CALL SNIPPETS(data, index, query[, opt_value AS opt_name[, ...]])

CALL SNIPPETS statement, introduced in version 1.10-beta, builds a snippet from provided data and query, using specified index settings.

data is the source data to extract a snippet from. It could be a single string, or the list of the strings enclosed in curly brackets. index is the name of the index from which to take the text processing settings. query is the full-text query to build snippets for. Additional options are documented in Section 8.7.1, “BuildExcerpts”. Usage example:

CALL SNIPPETS('this is my document text', 'test1', 'hello world',
    5 AS around, 200 AS limit);
CALL SNIPPETS(('this is my document text','this is my another text'), 'test1', 'hello world',
    5 AS around, 200 AS limit);
CALL SNIPPETS(('data/doc1.txt','data/doc2.txt','/home/sphinx/doc3.txt'), 'test1', 'hello world',
    5 AS around, 200 AS limit, 1 AS load_files);

7.11. CALL KEYWORDS syntax

CALL KEYWORDS(text, index, [hits])

CALL KEYWORDS statement, introduced in version 1.10-beta, splits text into particular keywords. It returns tokenized and normalized forms of the keywords, and, optionally, keyword statistics.

text is the text to break down to keywords. index is the name of the index from which to take the text processing settings. hits is an optional boolean parameter that specifies whether to return document and hit occurrence statistics.

7.12. SHOW TABLES syntax


SHOW TABLES statement, introduced in version 2.0.1-beta, enumerates all currently active indexes along with their types. As of 2.0.1-beta, existing index types are local, distributed, and rt respectively. Example:

| Index | Type        |
| dist1 | distributed |
| rt    | rt          |
| test1 | local       |
| test2 | local       |
4 rows in set (0.00 sec)

7.13. DESCRIBE syntax


DESCRIBE statement, introduced in version 2.0.1-beta, lists index columns and their associated types. Columns are document ID, full-text fields, and attributes. The order matches that in which fields and attributes are expected by INSERT and REPLACE statements. As of 2.0.1-beta, column types are field, integer, timestamp, ordinal, bool, float, bigint, string, and mva. ID column will be typed either integer or bigint based on whether the binaries were built with 32-bit or 64-bit document ID support. Example:

mysql> DESC rt;
| Field   | Type    |
| id      | integer |
| title   | field   |
| content | field   |
| gid     | integer |
4 rows in set (0.00 sec)

7.14. CREATE FUNCTION syntax

	SONAME 'udf_lib_file'

CREATE FUNCTION statement, introduced in version 2.0.1-beta, installs a user-defined function (UDF) with the given name and type from the given library file. The library file must reside in a trusted plugin_dir directory. On success, the function is available for use in all subsequent queries that the server receives. Example:

mysql> CREATE FUNCTION avgmva RETURNS INT SONAME 'udfexample.dll';
Query OK, 0 rows affected (0.03 sec)

mysql> SELECT *, AVGMVA(tag) AS q from test1;
| id   | weight | tag     | q         |
|    1 |      1 | 1,3,5,7 | 4.000000  |
|    2 |      1 | 2,4,6   | 4.000000  |
|    3 |      1 | 15      | 15.000000 |
|    4 |      1 | 7,40    | 23.500000 |

7.15. DROP FUNCTION syntax


DROP FUNCTION statement, introduced in version 2.0.1-beta, deinstalls a user-defined function (UDF) with the given name. On success, the function is no longer available for use in subsequent queries. Pending concurrent queries will not be affected and the library unload, if necessary, will be postponed until those queries complete. Example:

mysql> DROP FUNCTION avgmva;
Query OK, 0 rows affected (0.00 sec)

7.16. SHOW VARIABLES syntax


SHOW VARIABLES statement was added in version 2.0.1-beta to improve compatibility with 3rd party MySQL connectors and frameworks that automatically execute this statement.

In version 2.0.1-beta, it did nothing.

Starting from version 2.0.2-beta, it returns the current values of a few server-wide variables. Also, support for GLOBAL and SESSION clauses was added.

| Variable_name        | Value    |
| autocommit           | 1        |
| collation_connection | libc_ci  |
| query_log_format     | sphinxql |
| log_level            | info     |
4 rows in set (0.00 sec)

7.17. SHOW COLLATION syntax


Added in version 2.0.1-beta, this is currently a placeholder query that does nothing and reports success. That is in order to keep compatibility with frameworks and connectors that automatically execute this statement.

Query OK, 0 rows affected (0.00 sec)

7.18. UPDATE syntax

UPDATE index SET col1 = newval1 [, ...] WHERE where_condition

UPDATE statement was added in version 2.0.1-beta. Multiple attributes and values can be specified in a single statement. Both RT and disk indexes are supported.

As of version 2.0.2-beta, all atributes types (int, bigint, float, MVA) except for strings can be updated. Previously, some of the types were not supported.

where_condition (also added in 2.0.2-beta) has the same syntax as in the SELECT statement (see Section 7.1, “SELECT syntax” for details).

When assigning the out-of-range values to 32-bit attributes, they will be trimmed to their lower 32 bits without a prompt. For example, if you try to update the 32-bit unsigned int with a value of 4294967297, the value of 1 will actually be stored, because the lower 32 bits of 4294967297 (0x100000001 in hex) amount to 1 (0x00000001 in hex).

MVA values sets for updating (and also for INSERT or REPLACE, refer to Section 7.5, “INSERT and REPLACE syntax”) must be specificed as comma-separated lists in parentheses. To erase the MVA value, just assign () to it.

mysql> UPDATE myindex SET enabled=0 WHERE id=123;
Query OK, 1 rows affected (0.00 sec)

mysql> UPDATE myindex
  SET bigattr=-100000000000,
  WHERE MATCH('hehe') AND enabled=1;
Query OK, 148 rows affected (0.01 sec)

7.19. ATTACH INDEX syntax

ATTACH INDEX diskindex TO RTINDEX rtindex

ATTACH INDEX statement, added in version 2.0.2-beta, lets you move data from a regular disk index to a RT index.

After a successful ATTACH, the data originally stored in the source disk index becomes a part of the target RT index, and the source disk index becomes unavailable (until the next rebuild). ATTACH does not result in any index data changes. Basically, it just renames the files (making the source index a new disk chunk of the target RT index), and updates the metadata. So it is a generally quick operation which might (frequently) complete as fast as under a second.

Note that when an index is attached to an empty RT index, the fields, attributes, and text processing settings (tokenizer, wordforms, etc) from the source index are copied over and take effect. The respective parts of the RT index definition from the configuration file will be ignored.

As of 2.0.2-beta, ATTACH INDEX comes with a number of restrictions. Most notably, the target RT index is currently required to be empty, making ATTACH INDEX a one-time conversion operation only. Those restrictions may be lifted in future releases, as we add the needed functionality to the RT indexes. The complete list is as follows.

  • Target RT index needs to be empty.

  • Source disk index needs to have index_sp=0, boundary_step=0, stopword_step=1, dict=crc settings.

  • Source disk index needs to have an empty index_zones setting.

mysql> DESC rt;
| Field     | Type    |
| id        | integer |
| testfield | field   |
| testattr  | uint    |
3 rows in set (0.00 sec)

mysql> SELECT * FROM rt;
Empty set (0.00 sec)

mysql> SELECT * FROM disk WHERE MATCH('test');
| id   | weight | group_id | date_added |
|    1 |   1304 |        1 | 1313643256 |
|    2 |   1304 |        1 | 1313643256 |
|    3 |   1304 |        1 | 1313643256 |
|    4 |   1304 |        1 | 1313643256 |
4 rows in set (0.00 sec)

Query OK, 0 rows affected (0.00 sec)

mysql> DESC rt;
| Field      | Type      |
| id         | integer   |
| title      | field     |
| content    | field     |
| group_id   | uint      |
| date_added | timestamp |
5 rows in set (0.00 sec)

mysql> SELECT * FROM rt WHERE MATCH('test');
| id   | weight | group_id | date_added |
|    1 |   1304 |        1 | 1313643256 |
|    2 |   1304 |        1 | 1313643256 |
|    3 |   1304 |        1 | 1313643256 |
|    4 |   1304 |        1 | 1313643256 |
4 rows in set (0.00 sec)

mysql> SELECT * FROM disk WHERE MATCH('test');
ERROR 1064 (42000): no enabled local indexes to search

7.20. FLUSH RTINDEX syntax


FLUSH RTINDEX statement, added in version 2.0.2-beta, forcibly flushes RT index RAM chunk contents to disk.

Backing up a RT index is as simple as copying over its data files, followed by the binary log. However, recovering from that backup means that all the transactions in the log since the last successful RAM chunk write would need to be replayed. Those writes normally happen either on a clean shutdown, or periodically with a (big enough!) interval between writes specified in rt_flush_period directive. So such a backup made at an arbitrary point in time just might end up with way too much binary log data to replay.

FLUSH RTINDEX forcibly writes the RAM chunk contents to disk, and also causes the subsequent cleanup of (now-redundant) binary log files. Thus, recovering from a backup made just after FLUSH RTINDEX should be almost instant.

mysql> FLUSH RTINDEX rt;
Query OK, 0 rows affected (0.05 sec)

7.21. Multi-statement queries

Starting version 2.0.1-beta, SphinxQL supports multi-statement queries, or batches. Possible inter-statement optimizations described in Section 5.11, “Multi-queries” do apply to SphinxQL just as well. The batched queries should be separated by a semicolon. Your MySQL client library needs to support MySQL multi-query mechanism and multiple result set. For instance, mysqli interface in PHP and DBI/DBD libraries in Perl are known to work.

Here's a PHP sample showing how to utilize mysqli interface with Sphinx.


$link = mysqli_connect ( "", "root", "", "", 9306 );
if ( mysqli_connect_errno() )
    die ( "connect failed: " . mysqli_connect_error() );

$batch = "SELECT * FROM test1 ORDER BY group_id ASC;";
$batch .= "SELECT * FROM test1 ORDER BY group_id DESC";

if ( !mysqli_multi_query ( $link, $batch ) )
    die ( "query failed" );

    // fetch and print result set
    if ( $result = mysqli_store_result($link) )
        while ( $row = mysqli_fetch_row($result) )
            printf ( "id=%s\n", $row[0] );

    // print divider
    if ( mysqli_more_results($link) )
        printf ( "------\n" );

} while ( mysqli_next_result($link) );

Its output with the sample test1 index included with Sphinx is as follows.

$ php test_multi.php

The following statements can currently be used in a batch: SELECT, SHOW WARNINGS, SHOW STATUS, and SHOW META. Arbitrary sequence of these statements are allowed. The results sets returned should match those that would be returned if the batched queries were sent one by one.

7.22. Comment syntax

Since version 2.0.1-beta, SphinxQL supports C-style comment syntax. Everything from an opening /* sequence to a closing */ sequence is ignored. Comments can span multiple lines, can not nest, and should not get logged. MySQL specific /*! ... */ comments are also currently ignored. (As the comments support was rather added for better compatibility with mysqldump produced dumps, rather than improving generaly query interoperability between Sphinx and MySQL.)


7.23. List of SphinxQL reserved keywords

A complete alphabetical list of keywords that are currently reserved in SphinxQL syntax (and therefore can not be used as identifiers).


7.24. SphinxQL upgrade notes, version 2.0.1-beta

This section only applies to existing applications that use SphinxQL versions prior to 2.0.1-beta.

In previous versions, SphinxQL just wrapped around SphinxAPI and inherited its magic columns and column set quirks. Essentially, SphinxQL queries could return (slightly) different columns and in a (slightly) different order than it was explicitly requested in the query. Namely, weight magic column (which is not a real column in any index) was added at all times, and GROUP BY related @count, @group, and @distinct magic columns were conditionally added when grouping. Also, the order of columns (attributes) in the result set was actually taken from the index rather than the query. (So if you asked for columns C, B, A in your query but they were in the A, B, C order in the index, they would have been returned in the A, B, C order.)

In version 2.0.1-beta, we fixed that. SphinxQL is now more SQL compliant (and will be further brought in as much compliance with standard SQL syntax as possible). That is not yet a breaking change, because searchd now supports compat_sphinxql_magics directive that flips between the old "compatibility" mode and the new "compliance" mode. However, the compatibility mode support is going to be removed in future, so it's strongly advised to update SphinxQL applications and switch to the compliance mode.

The important changes are as follows:

  • @ID magic name is deprecated in favor of ID. Document ID is considered an attribute.

  • WEIGHT is no longer implicitly returned, because it is not actually a column (an index attribute), but rather an internal function computed per each row (a match). You have to explicitly ask for it, using the WEIGHT() function. (The requirement to alias the result will be lifted in the next release.)

    SELECT id, WEIGHT() w FROM myindex WHERE MATCH('test')

  • You can now use quoted reserved keywords as aliases. The quote character is backtick ("`", ASCII code 96 decimal, 60 hex). One particularly useful example would be returning weight column like the old mode:

    SELECT id, WEIGHT() `weight` FROM myindex WHERE MATCH('test')

  • The column order is now different and should now match the one expliclitly defined in the query. So if you are accessing columns based on their position in the result set rather than the name (for instance, by using mysql_fetch_row() rather than mysql_fetch_assoc() in PHP), check and fix the order of columns in your queries.

  • SELECT * return the columns in index order, as it used to, including the ID column. However, SELECT * does not automatically return WEIGHT(). To update such queries in case you access columns by names, simply add it to the query:

    SELECT *, WEIGHT() `weight` FROM myindex WHERE MATCH('test')

    Otherwise, i.e., in case you rely on column order, select ID, weight, and then other columns:

    SELECT id, *, WEIGHT() `weight` FROM myindex WHERE MATCH('test')

  • Magic @count and @distinct attributes are no longer implicitly returned. You now have to explicitly ask for them when using GROUP BY. (Also note that you currently have to alias them; that requirement will be lifted in the future.)

    SELECT gid, COUNT(*) q FROM myindex WHERE MATCH('test')

Chapter 8. API reference

There is a number of native searchd client API implementations for Sphinx. As of time of this writing, we officially support our own PHP, Python, and Java implementations. There also are third party free, open-source API implementations for Perl, Ruby, and C++.

The reference API implementation is in PHP, because (we believe) Sphinx is most widely used with PHP than any other language. This reference documentation is in turn based on reference PHP API, and all code samples in this section will be given in PHP.

However, all other APIs provide the same methods and implement the very same network protocol. Therefore the documentation does apply to them as well. There might be minor differences as to the method naming conventions or specific data structures used. But the provided functionality must not differ across languages.

8.1. General API functions

8.1.1. GetLastError

Prototype: function GetLastError()

Returns last error message, as a string, in human readable format. If there were no errors during the previous API call, empty string is returned.

You should call it when any other function (such as Query()) fails (typically, the failing function returns false). The returned string will contain the error description.

The error message is not reset by this call; so you can safely call it several times if needed.

8.1.2. GetLastWarning

Prototype: function GetLastWarning ()

Returns last warning message, as a string, in human readable format. If there were no warnings during the previous API call, empty string is returned.

You should call it to verify whether your request (such as Query()) was completed but with warnings. For instance, search query against a distributed index might complete succesfully even if several remote agents timed out. In that case, a warning message would be produced.

The warning message is not reset by this call; so you can safely call it several times if needed.

8.1.3. SetServer

Prototype: function SetServer ( $host, $port )

Sets searchd host name and TCP port. All subsequent requests will use the new host and port settings. Default host and port are 'localhost' and 9312, respectively.

8.1.4. SetRetries

Prototype: function SetRetries ( $count, $delay=0 )

Sets distributed retry count and delay.

On temporary failures searchd will attempt up to $count retries per agent. $delay is the delay between the retries, in milliseconds. Retries are disabled by default. Note that this call will not make the API itself retry on temporary failure; it only tells searchd to do so. Currently, the list of temporary failures includes all kinds of connect() failures and maxed out (too busy) remote agents.

8.1.5. SetConnectTimeout

Prototype: function SetConnectTimeout ( $timeout )

Sets the time allowed to spend connecting to the server before giving up.

Under some circumstances, the server can be delayed in responding, either due to network delays, or a query backlog. In either instance, this allows the client application programmer some degree of control over how their program interacts with searchd when not available, and can ensure that the client application does not fail due to exceeding the script execution limits (especially in PHP).

In the event of a failure to connect, an appropriate error code should be returned back to the application in order for application-level error handling to advise the user.

8.1.6. SetArrayResult

Prototype: function SetArrayResult ( $arrayresult )

PHP specific. Controls matches format in the search results set (whether matches should be returned as an array or a hash).

$arrayresult argument must be boolean. If $arrayresult is false (the default mode), matches will returned in PHP hash format with document IDs as keys, and other information (weight, attributes) as values. If $arrayresult is true, matches will be returned as a plain array with complete per-match information including document ID.

Introduced along with GROUP BY support on MVA attributes. Group-by-MVA result sets may contain duplicate document IDs. Thus they need to be returned as plain arrays, because hashes will only keep one entry per document ID.

8.1.7. IsConnectError

Prototype: function IsConnectError ()

Checks whether the last error was a network error on API side, or a remote error reported by searchd. Returns true if the last connection attempt to searchd failed on API side, false otherwise (if the error was remote, or there were no connection attempts at all). Introduced in version 0.9.9-rc1.

8.2. General query settings

8.2.1. SetLimits

Prototype: function SetLimits ( $offset, $limit, $max_matches=0, $cutoff=0 )

Sets offset into server-side result set ($offset) and amount of matches to return to client starting from that offset ($limit). Can additionally control maximum server-side result set size for current query ($max_matches) and the threshold amount of matches to stop searching at ($cutoff). All parameters must be non-negative integers.

First two parameters to SetLimits() are identical in behavior to MySQL LIMIT clause. They instruct searchd to return at most $limit matches starting from match number $offset. The default offset and limit settings are 0 and 20, that is, to return first 20 matches.

max_matches setting controls how much matches searchd will keep in RAM while searching. All matching documents will be normally processed, ranked, filtered, and sorted even if max_matches is set to 1. But only best N documents are stored in memory at any given moment for performance and RAM usage reasons, and this setting controls that N. Note that there are two places where max_matches limit is enforced. Per-query limit is controlled by this API call, but there also is per-server limit controlled by max_matches setting in the config file. To prevent RAM usage abuse, server will not allow to set per-query limit higher than the per-server limit.

You can't retrieve more than max_matches matches to the client application. The default limit is set to 1000. Normally, you must not have to go over this limit. One thousand records is enough to present to the end user. And if you're thinking about pulling the results to application for further sorting or filtering, that would be much more efficient if performed on Sphinx side.

$cutoff setting is intended for advanced performance control. It tells searchd to forcibly stop search query once $cutoff matches had been found and processed.

8.2.2. SetMaxQueryTime

Prototype: function SetMaxQueryTime ( $max_query_time )

Sets maximum search query time, in milliseconds. Parameter must be a non-negative integer. Default valus is 0 which means "do not limit".

Similar to $cutoff setting from SetLimits(), but limits elapsed query time instead of processed matches count. Local search queries will be stopped once that much time has elapsed. Note that if you're performing a search which queries several local indexes, this limit applies to each index separately.

8.2.3. SetOverride

Prototype: function SetOverride ( $attrname, $attrtype, $values )

Sets temporary (per-query) per-document attribute value overrides. Only supports scalar attributes. $values must be a hash that maps document IDs to overridden attribute values. Introduced in version 0.9.9-rc1.

Override feature lets you "temporary" update attribute values for some documents within a single query, leaving all other queries unaffected. This might be useful for personalized data. For example, assume you're implementing a personalized search function that wants to boost the posts that the user's friends recommend. Such data is not just dynamic, but also personal; so you can't simply put it in the index because you don't want everyone's searches affected. Overrides, on the other hand, are local to a single query and invisible to everyone else. So you can, say, setup a "friends_weight" value for every document, defaulting to 0, then temporary override it with 1 for documents 123, 456 and 789 (recommended by exactly the friends of current user), and use that value when ranking.

8.2.4. SetSelect

Prototype: function SetSelect ( $clause )

Sets the select clause, listing specific attributes to fetch, and expressions to compute and fetch. Clause syntax mimics SQL. Introduced in version 0.9.9-rc1.

SetSelect() is very similar to the part of a typical SQL query between SELECT and FROM. It lets you choose what attributes (columns) to fetch, and also what expressions over the columns to compute and fetch. A certain difference from SQL is that expressions must always be aliased to a correct identifier (consisting of letters and digits) using 'AS' keyword. SQL also lets you do that but does not require to. Sphinx enforces aliases so that the computation results can always be returned under a "normal" name in the result set, used in other clauses, etc.

Everything else is basically identical to SQL. Star ('*') is supported. Functions are supported. Arbitrary amount of expressions is supported. Computed expressions can be used for sorting, filtering, and grouping, just as the regular attributes.

Starting with version 0.9.9-rc2, aggregate functions (AVG(), MIN(), MAX(), SUM()) are supported when using GROUP BY.

Expression sorting (Section 5.6, “SPH_SORT_EXPR mode”) and geodistance functions (Section 8.4.5, “SetGeoAnchor”) are now internally implemented using this computed expressions mechanism, using magic names '@expr' and '@geodist' respectively.


$cl->SetSelect ( "*, @weight+(user_karma+ln(pageviews))*0.1 AS myweight" );
$cl->SetSelect ( "exp_years, salary_gbp*{$gbp_usd_rate} AS salary_usd,
   IF(age>40,1,0) AS over40" );
$cl->SetSelect ( "*, AVG(price) AS avgprice" );

8.3. Full-text search query settings

8.3.1. SetMatchMode

Prototype: function SetMatchMode ( $mode )

Sets full-text query matching mode, as described in Section 5.1, “Matching modes”. Parameter must be a constant specifying one of the known modes.

WARNING: (PHP specific) you must not take the matching mode constant name in quotes, that syntax specifies a string and is incorrect:

$cl->SetMatchMode ( "SPH_MATCH_ANY" ); // INCORRECT! will not work as expected
$cl->SetMatchMode ( SPH_MATCH_ANY ); // correct, works OK

8.3.2. SetRankingMode

Prototype: function SetRankingMode ( $ranker, $rankexpr="" )

Sets ranking mode (aka ranker). Only available in SPH_MATCH_EXTENDED matching mode. Parameter must be a constant specifying one of the known rankers.

By default, in the EXTENDED matching mode Sphinx computes two factors which contribute to the final match weight. The major part is a phrase proximity value between the document text and the query. The minor part is so-called BM25 statistical function, which varies from 0 to 1 depending on the keyword frequency within document (more occurrences yield higher weight) and within the whole index (more rare keywords yield higher weight).

However, in some cases you'd want to compute weight differently - or maybe avoid computing it at all for performance reasons because you're sorting the result set by something else anyway. This can be accomplished by setting the appropriate ranking mode. The list of the modes is available in Section 5.4, “Search results ranking”.

$rankexpr argument was added in version 2.0.2-beta. It lets you specify a ranking formula to use with the expression based ranker, that is, when $ranker is set to SPH_RANK_EXPR. In all other cases, $rankexpr is ignored.

8.3.3. SetSortMode

Prototype: function SetSortMode ( $mode, $sortby="" )

Set matches sorting mode, as described in Section 5.6, “Sorting modes”. Parameter must be a constant specifying one of the known modes.

WARNING: (PHP specific) you must not take the matching mode constant name in quotes, that syntax specifies a string and is incorrect:

$cl->SetSortMode ( "SPH_SORT_ATTR_DESC" ); // INCORRECT! will not work as expected
$cl->SetSortMode ( SPH_SORT_ATTR_ASC ); // correct, works OK

8.3.4. SetWeights

Prototype: function SetWeights ( $weights )

Binds per-field weights in the order of appearance in the index. DEPRECATED, use SetFieldWeights() instead.

8.3.5. SetFieldWeights

Prototype: function SetFieldWeights ( $weights )

Binds per-field weights by name. Parameter must be a hash (associative array) mapping string field names to integer weights.

Match ranking can be affected by per-field weights. For instance, see Section 5.4, “Search results ranking” for an explanation how phrase proximity ranking is affected. This call lets you specify what non-default weights to assign to different full-text fields.

The weights must be positive 32-bit integers. The final weight will be a 32-bit integer too. Default weight value is 1. Unknown field names will be silently ignored.

There is no enforced limit on the maximum weight value at the moment. However, beware that if you set it too high you can start hitting 32-bit wraparound issues. For instance, if you set a weight of 10,000,000 and search in extended mode, then maximum possible weight will be equal to 10 million (your weight) by 1 thousand (internal BM25 scaling factor, see Section 5.4, “Search results ranking”) by 1 or more (phrase proximity rank). The result is at least 10 billion that does not fit in 32 bits and will be wrapped around, producing unexpected results.

8.3.6. SetIndexWeights

Prototype: function SetIndexWeights ( $weights )

Sets per-index weights, and enables weighted summing of match weights across different indexes. Parameter must be a hash (associative array) mapping string index names to integer weights. Default is empty array that means to disable weighting summing.

When a match with the same document ID is found in several different local indexes, by default Sphinx simply chooses the match from the index specified last in the query. This is to support searching through partially overlapping index partitions.

However in some cases the indexes are not just partitions, and you might want to sum the weights across the indexes instead of picking one. SetIndexWeights() lets you do that. With summing enabled, final match weight in result set will be computed as a sum of match weight coming from the given index multiplied by respective per-index weight specified in this call. Ie. if the document 123 is found in index A with the weight of 2, and also in index B with the weight of 3, and you called SetIndexWeights ( array ( "A"=>100, "B"=>10 ) ), the final weight return to the client will be 2*100+3*10 = 230.

8.4. Result set filtering settings

8.4.1. SetIDRange

Prototype: function SetIDRange ( $min, $max )

Sets an accepted range of document IDs. Parameters must be integers. Defaults are 0 and 0; that combination means to not limit by range.

After this call, only those records that have document ID between $min and $max (including IDs exactly equal to $min or $max) will be matched.

8.4.2. SetFilter

Prototype: function SetFilter ( $attribute, $values, $exclude=false )

Adds new integer values set filter.

On this call, additional new filter is added to the existing list of filters. $attribute must be a string with attribute name. $values must be a plain array containing integer values. $exclude must be a boolean value; it controls whether to accept the matching documents (default mode, when $exclude is false) or reject them.

Only those documents where $attribute column value stored in the index matches any of the values from $values array will be matched (or rejected, if $exclude is true).

8.4.3. SetFilterRange

Prototype: function SetFilterRange ( $attribute, $min, $max, $exclude=false )

Adds new integer range filter.

On this call, additional new filter is added to the existing list of filters. $attribute must be a string with attribute name. $min and $max must be integers that define the acceptable attribute values range (including the boundaries). $exclude must be a boolean value; it controls whether to accept the matching documents (default mode, when $exclude is false) or reject them.

Only those documents where $attribute column value stored in the index is between $min and $max (including values that are exactly equal to $min or $max) will be matched (or rejected, if $exclude is true).

8.4.4. SetFilterFloatRange

Prototype: function SetFilterFloatRange ( $attribute, $min, $max, $exclude=false )

Adds new float range filter.

On this call, additional new filter is added to the existing list of filters. $attribute must be a string with attribute name. $min and $max must be floats that define the acceptable attribute values range (including the boundaries). $exclude must be a boolean value; it controls whether to accept the matching documents (default mode, when $exclude is false) or reject them.

Only those documents where $attribute column value stored in the index is between $min and $max (including values that are exactly equal to $min or $max) will be matched (or rejected, if $exclude is true).

8.4.5. SetGeoAnchor

Prototype: function SetGeoAnchor ( $attrlat, $attrlong, $lat, $long )

Sets anchor point for and geosphere distance (geodistance) calculations, and enable them.

$attrlat and $attrlong must be strings that contain the names of latitude and longitude attributes, respectively. $lat and $long are floats that specify anchor point latitude and longitude, in radians.

Once an anchor point is set, you can use magic "@geodist" attribute name in your filters and/or sorting expressions. Sphinx will compute geosphere distance between the given anchor point and a point specified by latitude and lognitude attributes from each full-text match, and attach this value to the resulting match. The latitude and longitude values both in SetGeoAnchor and the index attribute data are expected to be in radians. The result will be returned in meters, so geodistance value of 1000.0 means 1 km. 1 mile is approximately 1609.344 meters.

8.5. GROUP BY settings

8.5.1. SetGroupBy

Prototype: function SetGroupBy ( $attribute, $func, $groupsort="@group desc" )

Sets grouping attribute, function, and groups sorting mode; and enables grouping (as described in Section 5.7, “Grouping (clustering) search results ”).

$attribute is a string that contains group-by attribute name. $func is a constant that chooses a function applied to the attribute value in order to compute group-by key. $groupsort is a clause that controls how the groups will be sorted. Its syntax is similar to that described in Section 5.6, “SPH_SORT_EXTENDED mode”.

Grouping feature is very similar in nature to GROUP BY clause from SQL. Results produces by this function call are going to be the same as produced by the following pseudo code:

SELECT ... GROUP BY $func($attribute) ORDER BY $groupsort

Note that it's $groupsort that affects the order of matches in the final result set. Sorting mode (see Section 8.3.3, “SetSortMode”) affect the ordering of matches within group, ie. what match will be selected as the best one from the group. So you can for instance order the groups by matches count and select the most relevant match within each group at the same time.

Starting with version 0.9.9-rc2, aggregate functions (AVG(), MIN(), MAX(), SUM()) are supported through SetSelect() API call when using GROUP BY.

Starting with version 2.0.1-beta, grouping on string attributes is supported, with respect to current collation.

8.5.2. SetGroupDistinct

Prototype: function SetGroupDistinct ( $attribute )

Sets attribute name for per-group distinct values count calculations. Only available for grouping queries.

$attribute is a string that contains the attribute name. For each group, all values of this attribute will be stored (as RAM limits permit), then the amount of distinct values will be calculated and returned to the client. This feature is similar to COUNT(DISTINCT) clause in standard SQL; so these Sphinx calls:

$cl->SetGroupBy ( "category", SPH_GROUPBY_ATTR, "@count desc" );
$cl->SetGroupDistinct ( "vendor" );

can be expressed using the following SQL clauses:

SELECT id, weight, all-attributes,
	COUNT(DISTINCT vendor) AS @distinct,
	COUNT(*) AS @count
FROM products
GROUP BY category

In the sample pseudo code shown just above, SetGroupDistinct() call corresponds to COUNT(DISINCT vendor) clause only. GROUP BY, ORDER BY, and COUNT(*) clauses are all an equivalent of SetGroupBy() settings. Both queries will return one matching row for each category. In addition to indexed attributes, matches will also contain total per-category matches count, and the count of distinct vendor IDs within each category.

8.6. Querying

8.6.1. Query

Prototype: function Query ( $query, $index="*", $comment="" )

Connects to searchd server, runs given search query with current settings, obtains and returns the result set.

$query is a query string. $index is an index name (or names) string. Returns false and sets GetLastError() message on general error. Returns search result set on success. Additionally, the contents of $comment are sent to the query log, marked in square brackets, just before the search terms, which can be very useful for debugging. Currently, the comment is limited to 128 characters.

Default value for $index is "*" that means to query all local indexes. Characters allowed in index names include Latin letters (a-z), numbers (0-9), minus sign (-), and underscore (_); everything else is considered a separator. Therefore, all of the following samples calls are valid and will search the same two indexes:

$cl->Query ( "test query", "main delta" );
$cl->Query ( "test query", "main;delta" );
$cl->Query ( "test query", "main, delta" );

Index specification order matters. If document with identical IDs are found in two or more indexes, weight and attribute values from the very last matching index will be used for sorting and returning to client (unless explicitly overridden with SetIndexWeights()). Therefore, in the example above, matches from "delta" index will always win over matches from "main".

On success, Query() returns a result set that contains some of the found matches (as requested by SetLimits()) and additional general per-query statistics. The result set is a hash (PHP specific; other languages might utilize other structures instead of hash) with the following keys and values:


Hash which maps found document IDs to another small hash containing document weight and attribute values (or an array of the similar small hashes if SetArrayResult() was enabled).


Total amount of matches retrieved on server (ie. to the server side result set) by this query. You can retrieve up to this amount of matches from server for this query text with current query settings.


Total amount of matching documents in index (that were found and procesed on server).


Hash which maps query keywords (case-folded, stemmed, and otherwise processed) to a small hash with per-keyword statitics ("docs", "hits").


Query error message reported by searchd (string, human readable). Empty if there were no errors.


Query warning message reported by searchd (string, human readable). Empty if there were no warnings.

It should be noted that Query() carries out the same actions as AddQuery() and RunQueries() without the intermediate steps; it is analoguous to a single AddQuery() call, followed by a corresponding RunQueries(), then returning the first array element of matches (from the first, and only, query.)

8.6.2. AddQuery

Prototype: function AddQuery ( $query, $index="*", $comment="" )

Adds additional query with current settings to multi-query batch. $query is a query string. $index is an index name (or names) string. Additionally if provided, the contents of $comment are sent to the query log, marked in square brackets, just before the search terms, which can be very useful for debugging. Currently, this is limited to 128 characters. Returns index to results array returned from RunQueries().

Batch queries (or multi-queries) enable searchd to perform internal optimizations if possible. They also reduce network connection overheads and search process creation overheads in all cases. They do not result in any additional overheads compared to simple queries. Thus, if you run several different queries from your web page, you should always consider using multi-queries.

For instance, running the same full-text query but with different sorting or group-by settings will enable searchd to perform expensive full-text search and ranking operation only once, but compute multiple group-by results from its output.

This can be a big saver when you need to display not just plain search results but also some per-category counts, such as the amount of products grouped by vendor. Without multi-query, you would have to run several queries which perform essentially the same search and retrieve the same matches, but create result sets differently. With multi-query, you simply pass all these querys in a single batch and Sphinx optimizes the redundant full-text search internally.

AddQuery() internally saves full current settings state along with the query, and you can safely change them afterwards for subsequent AddQuery() calls. Already added queries will not be affected; there's actually no way to change them at all. Here's an example:

$cl->SetSortMode ( SPH_SORT_RELEVANCE );
$cl->AddQuery ( "hello world", "documents" );

$cl->SetSortMode ( SPH_SORT_ATTR_DESC, "price" );
$cl->AddQuery ( "ipod", "products" );

$cl->AddQuery ( "harry potter", "books" );

$results = $cl->RunQueries ();

With the code above, 1st query will search for "hello world" in "documents" index and sort results by relevance, 2nd query will search for "ipod" in "products" index and sort results by price, and 3rd query will search for "harry potter" in "books" index while still sorting by price. Note that 2nd SetSortMode() call does not affect the first query (because it's already added) but affects both other subsequent queries.

Additionally, any filters set up before an AddQuery() will fall through to subsequent queries. So, if SetFilter() is called before the first query, the same filter will be in place for the second (and subsequent) queries batched through AddQuery() unless you call ResetFilters() first. Alternatively, you can add additional filters as well.

This would also be true for grouping options and sorting options; no current sorting, filtering, and grouping settings are affected by this call; so subsequent queries will reuse current query settings.

AddQuery() returns an index into an array of results that will be returned from RunQueries() call. It is simply a sequentially increasing 0-based integer, ie. first call will return 0, second will return 1, and so on. Just a small helper so you won't have to track the indexes manualy if you need then.

8.6.3. RunQueries

Prototype: function RunQueries ()

Connect to searchd, runs a batch of all queries added using AddQuery(), obtains and returns the result sets. Returns false and sets GetLastError() message on general error (such as network I/O failure). Returns a plain array of result sets on success.

Each result set in the returned array is exactly the same as the result set returned from Query().

Note that the batch query request itself almost always succeds - unless there's a network error, blocking index rotation in progress, or another general failure which prevents the whole request from being processed.

However individual queries within the batch might very well fail. In this case their respective result sets will contain non-empty "error" message, but no matches or query statistics. In the extreme case all queries within the batch could fail. There still will be no general error reported, because API was able to succesfully connect to searchd, submit the batch, and receive the results - but every result set will have a specific error message.

8.6.4. ResetFilters

Prototype: function ResetFilters ()

Clears all currently set filters.

This call is only normally required when using multi-queries. You might want to set different filters for different queries in the batch. To do that, you should call ResetFilters() and add new filters using the respective calls.

8.6.5. ResetGroupBy

Prototype: function ResetGroupBy ()

Clears all currently group-by settings, and disables group-by.

This call is only normally required when using multi-queries. You can change individual group-by settings using SetGroupBy() and SetGroupDistinct() calls, but you can not disable group-by using those calls. ResetGroupBy() fully resets previous group-by settings and disables group-by mode in the current state, so that subsequent AddQuery() calls can perform non-grouping searches.

8.7. Additional functionality

8.7.1. BuildExcerpts

Prototype: function BuildExcerpts ( $docs, $index, $words, $opts=array() )

Excerpts (snippets) builder function. Connects to searchd, asks it to generate excerpts (snippets) from given documents, and returns the results.

$docs is a plain array of strings that carry the documents' contents. $index is an index name string. Different settings (such as charset, morphology, wordforms) from given index will be used. $words is a string that contains the keywords to highlight. They will be processed with respect to index settings. For instance, if English stemming is enabled in the index, "shoes" will be highlighted even if keyword is "shoe". Starting with version 0.9.9-rc1, keywords can contain wildcards, that work similarly to star-syntax available in queries. $opts is a hash which contains additional optional highlighting parameters:


A string to insert before a keyword match. Starting with version 1.10-beta, a %PASSAGE_ID% macro can be used in this string. The macro is replaced with an incrementing passage number within a current snippet. Numbering starts at 1 by default but can be overridden with "start_passage_id" option. In a multi-document call, %PASSAGE_ID% would restart at every given document. Default is "<b>".


A string to insert after a keyword match. Starting with version 1.10-beta, a %PASSAGE_ID% macro can be used in this string. Default is "</b>".


A string to insert between snippet chunks (passages). Default is " ... ".


Maximum snippet size, in symbols (codepoints). Integer, default is 256.


How much words to pick around each matching keywords block. Integer, default is 5.


Whether to highlight exact query phrase matches only instead of individual keywords. Boolean, default is false.


Whether to extract single best passage only. Boolean, default is false.


Whether to additionaly break passages by phrase boundary characters, as configured in index settings with phrase_boundary directive. Boolean, default is false.


Whether to sort the extracted passages in order of relevance (decreasing weight), or in order of appearance in the document (increasing position). Boolean, default is false.


Added in version 1.10-beta. Whether to handle $words as a query in extended syntax, or as a bag of words (default behavior). For instance, in query mode ("one two" | "three four") will only highlight and include those occurrences "one two" or "three four" when the two words from each pair are adjacent to each other. In default mode, any single occurrence of "one", "two", "three", or "four" would be highlighted. Boolean, default is false.


Added in version 1.10-beta. Ignores the snippet length limit until it includes all the keywords. Boolean, default is false.


Added in version 1.10-beta. Limits the maximum number of passages that can be included into the snippet. Integer, default is 0 (no limit).


Added in version 1.10-beta. Limits the maximum number of keywords that can be included into the snippet. Integer, default is 0 (no limit).


Added in version 1.10-beta. Specifies the starting value of %PASSAGE_ID% macro (that gets detected and expanded in before_match, after_match strings). Integer, default is 1.


Added in version 1.10-beta. Whether to handle $docs as data to extract snippets from (default behavior), or to treat it as file names, and load data from specified files on the server side. Starting with version 2.0.1-beta, up to dist_threads worker threads per request will be created to parallelize the work when this flag is enabled. Boolean, default is false. Starting with version 2.0.2-beta, building of the snippets could be parallelized between remote agents. Just set the 'dist_threads' param in the config to the value greater than 1, and then invoke the snippets generation over the distributed index, which contain only one(!) first(!) local agent and several remotes.


Added in version 2.0.2-beta. It works only with distributed snippets generation with remote agents. The source files for snippets could be distributed among different agents, and the main daemon will merge together all non-erroneous results. So, if one agent of the distributed index has 'file1.txt', another has 'file2.txt' and you call for the snippets with both these files, the sphinx will merge results from the agents together, so you will get the snippets from both 'file1.txt' and 'file2.txt'. Boolean, default is false.

If the "load_files" is also set, the request will return the error in case if any of the files is not available anywhere. Otherwise (if "load_files" is not set) it will just return the empty strings for all absent files. The master instance reset this flag when distributes the snippets among agents. So, for agents the absence of a file is not critical error, but for the master it might be so. If you want to be sure that all snippets are actually created, set both "load_files_scattered" and "load_files". If the absense of some snippets caused by some agents is not critical for you - set just "load_files_scattered", leaving "load_files" not set.


Added in version 1.10-beta. HTML stripping mode setting. Defaults to "index", which means that index settings will be used. The other values are "none" and "strip", that forcibly skip or apply stripping irregardless of index settings; and "retain", that retains HTML markup and protects it from highlighting. The "retain" mode can only be used when highlighting full documents and thus requires that no snippet size limits are set. String, allowed values are "none", "strip", "index", and "retain".


Added in version 1.10-beta. Allows empty string to be returned as highlighting result when a snippet could not be generated (no keywords match, or no passages fit the limit). By default, the beginning of original text would be returned instead of an empty string. Boolean, default is false.


Added in version 2.0.1-beta. Ensures that passages do not cross a sentence, paragraph, or zone boundary (when used with an index that has the respective indexing settings enabled). String, allowed values are "sentence", "paragraph", and "zone".


Added in version 2.0.1-beta. Emits an HTML tag with an enclosing zone name before each passage. Boolean, default is false.

Snippets extraction algorithm currently favors better passages (with closer phrase matches), and then passages with keywords not yet in snippet. Generally, it will try to highlight the best match with the query, and it will also to highlight all the query keywords, as made possible by the limtis. In case the document does not match the query, beginning of the document trimmed down according to the limits will be return by default. Starting with 1.10-beta, you can also return an empty snippet instead case by setting "allow_empty" option to true.

Returns false on failure. Returns a plain array of strings with excerpts (snippets) on success.

8.7.2. UpdateAttributes

Prototype: function UpdateAttributes ( $index, $attrs, $values )

Instantly updates given attribute values in given documents. Returns number of actually updated documents (0 or more) on success, or -1 on failure.

$index is a name of the index (or indexes) to be updated. $attrs is a plain array with string attribute names, listing attributes that are updated. $values is a hash where key is document ID, and value is a plain array of new attribute values.

$index can be either a single index name or a list, like in Query(). Unlike Query(), wildcard is not allowed and all the indexes to update must be specified explicitly. The list of indexes can include distributed index names. Updates on distributed indexes will be pushed to all agents.

The updates only work with docinfo=extern storage strategy. They are very fast because they're working fully in RAM, but they can also be made persistent: updates are saved on disk on clean searchd shutdown initiated by SIGTERM signal. With additional restrictions, updates are also possible on MVA attributes; refer to mva_updates_pool directive for details.

Usage example:

$cl->UpdateAttributes ( "test1", array("group_id"), array(1=>array(456)) );
$cl->UpdateAttributes ( "products", array ( "price", "amount_in_stock" ),
	array ( 1001=>array(123,5), 1002=>array(37,11), 1003=>(25,129) ) );

The first sample statement will update document 1 in index "test1", setting "group_id" to 456. The second one will update documents 1001, 1002 and 1003 in index "products". For document 1001, the new price will be set to 123 and the new amount in stock to 5; for document 1002, the new price will be 37 and the new amount will be 11; etc.

8.7.3. BuildKeywords

Prototype: function BuildKeywords ( $query, $index, $hits )

Extracts keywords from query using tokenizer settings for given index, optionally with per-keyword occurrence statistics. Returns an array of hashes with per-keyword information.

$query is a query to extract keywords from. $index is a name of the index to get tokenizing settings and keyword occurrence statistics from. $hits is a boolean flag that indicates whether keyword occurrence statistics are required.

Usage example:

$keywords = $cl->BuildKeywords ( "this.is.my query", "test1", false );

8.7.4. EscapeString

Prototype: function EscapeString ( $string )

Escapes characters that are treated as special operators by the query language parser. Returns an escaped string.

$string is a string to escape.

This function might seem redundant because it's trivial to implement in any calling application. However, as the set of special characters might change over time, it makes sense to have an API call that is guaranteed to escape all such characters at all times.

Usage example:

$escaped = $cl->EscapeString ( "[email protected]/string" );

8.7.5. Status

Prototype: function Status ()

Queries searchd status, and returns an array of status variable name and value pairs.

Usage example:

$status = $cl->Status ();
foreach ( $status as $row )
	print join ( ": ", $row ) . "\n";

8.7.6. FlushAttributes

Prototype: function FlushAttributes ()

Forces searchd to flush pending attribute updates to disk, and blocks until completion. Returns a non-negative internal "flush tag" on success. Returns -1 and sets an error message on error. Introduced in version 1.10-beta.

Attribute values updated using UpdateAttributes() API call are only kept in RAM until a so-called flush (which writes the current, possibly updated attribute values back to disk). FlushAttributes() call lets you enforce a flush. The call will block until searchd finishes writing the data to disk, which might take seconds or even minutes depending on the total data size (.spa file size). All the currently updated indexes will be flushed.

Flush tag should be treated as an ever growing magic number that does not mean anything. It's guaranteed to be non-negative. It is guaranteed to grow over time, though not necessarily in a sequential fashion; for instance, two calls that return 10 and then 1000 respectively are a valid situation. If two calls to FlushAttrs() return the same tag, it means that there were no actual attribute updates in between them, and therefore current flushed state remained the same (for all indexes).

Usage example:

$status = $cl->FlushAttributes ();
if ( $status<0 )
	print "ERROR: " . $cl->GetLastError(); 

8.8. Persistent connections

Persistent connections allow to use single network connection to run multiple commands that would otherwise require reconnects.

8.8.1. Open

Prototype: function Open ()

Opens persistent connection to the server.

8.8.2. Close

Prototype: function Close ()

Closes previously opened persistent connection.

Chapter 9. MySQL storage engine (SphinxSE)

9.1. SphinxSE overview

SphinxSE is MySQL storage engine which can be compiled into MySQL server 5.x using its pluggable architecure. It is not available for MySQL 4.x series. It also requires MySQL 5.0.22 or higher in 5.0.x series, or MySQL 5.1.12 or higher in 5.1.x series.

Despite the name, SphinxSE does not actually store any data itself. It is actually a built-in client which allows MySQL server to talk to searchd, run search queries, and obtain search results. All indexing and searching happen outside MySQL.

Obvious SphinxSE applications include:

  • easier porting of MySQL FTS applications to Sphinx;

  • allowing Sphinx use with progamming languages for which native APIs are not available yet;

  • optimizations when additional Sphinx result set processing on MySQL side is required (eg. JOINs with original document tables, additional MySQL-side filtering, etc).

9.2. Installing SphinxSE

You will need to obtain a copy of MySQL sources, prepare those, and then recompile MySQL binary. MySQL sources (mysql-5.x.yy.tar.gz) could be obtained from dev.mysql.com Web site.

For some MySQL versions, there are delta tarballs with already prepared source versions available from Sphinx Web site. After unzipping those over original sources MySQL would be ready to be configured and built with Sphinx support.

If such tarball is not available, or does not work for you for any reason, you would have to prepare sources manually. You will need to GNU Autotools framework (autoconf, automake and libtool) installed to do that.

9.2.1. Compiling MySQL 5.0.x with SphinxSE

  1. copy sphinx.5.0.yy.diff patch file into MySQL sources directory and run

    patch -p1 < sphinx.5.0.yy.diff

    If there's no .diff file exactly for the specific version you need to build, try applying .diff with closest version numbers. It is important that the patch should apply with no rejects.

  2. in MySQL sources directory, run

    sh BUILD/autorun.sh

  3. in MySQL sources directory, create sql/sphinx directory in and copy all files in mysqlse directory from Sphinx sources there. Example:

    cp -R /root/builds/sphinx-0.9.7/mysqlse /root/builds/mysql-5.0.24/sql/sphinx

  4. configure MySQL and enable Sphinx engine:

    ./configure --with-sphinx-storage-engine

  5. build and install MySQL:

    make install

9.2.2. Compiling MySQL 5.1.x with SphinxSE

  1. in MySQL sources directory, create storage/sphinx directory in and copy all files in mysqlse directory from Sphinx sources there. Example:

    cp -R /root/builds/sphinx-0.9.7/mysqlse /root/builds/mysql-5.1.14/storage/sphinx

  2. in MySQL sources directory, run

    sh BUILD/autorun.sh

  3. configure MySQL and enable Sphinx engine:

    ./configure --with-plugins=sphinx

  4. build and install MySQL:

    make install

9.2.3. Checking SphinxSE installation

To check whether SphinxSE has been succesfully compiled into MySQL, launch newly built servers, run mysql client and issue SHOW ENGINES query. You should see a list of all available engines. Sphinx should be present and "Support" column should contain "YES":

mysql> show engines;
| Engine     | Support  | Comment                                                     |
| MyISAM     | DEFAULT  | Default engine as of MySQL 3.23 with great performance      |
| SPHINX     | YES      | Sphinx storage engine                                       |
13 rows in set (0.00 sec)

9.3. Using SphinxSE

To search via SphinxSE, you would need to create special ENGINE=SPHINX "search table", and then SELECT from it with full text query put into WHERE clause for query column.

Let's begin with an example create statement and search query:

    weight      INTEGER NOT NULL,
    query       VARCHAR(3072) NOT NULL,
    group_id    INTEGER,
) ENGINE=SPHINX CONNECTION="sphinx://localhost:9312/test";

SELECT * FROM t1 WHERE query='test it;mode=any';

First 3 columns of search table must have a types of INTEGER UNSINGED or BIGINT for the 1st column (document id), INTEGER or BIGINT for the 2nd column (match weight), and VARCHAR or TEXT for the 3rd column (your query), respectively. This mapping is fixed; you can not omit any of these three required columns, or move them around, or change types. Also, query column must be indexed; all the others must be kept unindexed. Columns' names are ignored so you can use arbitrary ones.

Additional columns must be either INTEGER, TIMESTAMP, BIGINT, VARCHAR, or FLOAT. They will be bound to attributes provided in Sphinx result set by name, so their names must match attribute names specified in sphinx.conf. If there's no such attribute name in Sphinx search results, column will have NULL values.

Special "virtual" attributes names can also be bound to SphinxSE columns. _sph_ needs to be used instead of @ for that. For instance, to obtain the values of @groupby, @count, or @distinct virtual attributes, use _sph_groupby, _sph_count or _sph_distinct column names, respectively.

CONNECTION string parameter can be used to specify default searchd host, port and indexes for queries issued using this table. If no connection string is specified in CREATE TABLE, index name "*" (ie. search all indexes) and localhost:9312 are assumed. Connection string syntax is as follows:


You can change the default connection string later:


You can also override all these parameters per-query.

As seen in example, both query text and search options should be put into WHERE clause on search query column (ie. 3rd column); the options are separated by semicolons; and their names from values by equality sign. Any number of options can be specified. Available options are:

  • query - query text;

  • mode - matching mode. Must be one of "all", "any", "phrase", "boolean", or "extended". Default is "all";

  • sort - match sorting mode. Must be one of "relevance", "attr_desc", "attr_asc", "time_segments", or "extended". In all modes besides "relevance" attribute name (or sorting clause for "extended") is also required after a colon:

    ... WHERE query='test;sort=attr_asc:group_id';
    ... WHERE query='test;sort=extended:@weight desc, group_id asc';

  • offset - offset into result set, default is 0;

  • limit - amount of matches to retrieve from result set, default is 20;

  • index - names of the indexes to search:

    ... WHERE query='test;index=test1;';
    ... WHERE query='test;index=test1,test2,test3;';

  • minid, maxid - min and max document ID to match;

  • weights - comma-separated list of weights to be assigned to Sphinx full-text fields:

    ... WHERE query='test;weights=1,2,3;';

  • filter, !filter - comma-separated attribute name and a set of values to match:

    # only include groups 1, 5 and 19
    ... WHERE query='test;filter=group_id,1,5,19;';
    # exclude groups 3 and 11
    ... WHERE query='test;!filter=group_id,3,11;';

  • range, !range - comma-separated attribute name, min and max value to match:

    # include groups from 3 to 7, inclusive
    ... WHERE query='test;range=group_id,3,7;';
    # exclude groups from 5 to 25
    ... WHERE query='test;!range=group_id,5,25;';

  • maxmatches - per-query max matches value, as in max_matches parameter to SetLimits() API call:

    ... WHERE query='test;maxmatches=2000;';

  • cutoff - maximum allowed matches, as in cutoff parameter to SetLimits() API call:

    ... WHERE query='test;cutoff=10000;';

  • maxquerytme - maximum allowed query time (in milliseconds), as in SetMaxQueryTime() API call:

    ... WHERE query='test;maxquerytime=1000;';

  • groupby - group-by function and attribute, corresponding to SetGroupBy() API call:

    ... WHERE query='test;groupby=day:published_ts;';
    ... WHERE query='test;groupby=attr:group_id;';

  • groupsort - group-by sorting clause:

    ... WHERE query='test;[email protected] desc;';

  • distinct - an attribute to compute COUNT(DISTINCT) for when doing group-by, as in SetGroupDistinct() API call:

    ... WHERE query='test;groupby=attr:country_id;distinct=site_id';

  • indexweights - comma-separated list of index names and weights to use when searching through several indexes:

    ... WHERE query='test;indexweights=idx_exact,2,idx_stemmed,1;';

  • comment - a string to mark this query in query log (mapping to $comment parameter in Query() API call):

    ... WHERE query='test;comment=marker001;';

  • select - a string with expressions to compute (mapping to SetSelect() API call):

    ... WHERE query='test;select=2*a+3*b as myexpr;';

  • host, port - remote searchd host name and TCP port, respectively:

    ... WHERE query='test;host=sphinx-test.loc;port=7312;';

  • ranker - a ranking function to use with "extended" matching mode, as in SetRankingMode() API call (the only mode that supports full query syntax). Known values are "proximity_bm25", "bm25", "none", "wordcount", "proximity", "matchany", "fieldmask"; and, starting with 2.0.4-release, "expr:EXPRESSION" syntax to support expression-based ranker (where EXPRESSION should be replaced with your specific ranking formula):

    ... WHERE query='test;mode=extended;ranker=bm25;';
    ... WHERE query='test;mode=extended;ranker=expr:sum(lcs);';

  • geoanchor - geodistance anchor, as in SetGeoAnchor() API call. Takes 4 parameters which are latitude and longiture attribute names, and anchor point coordinates respectively:

    ... WHERE query='test;geoanchor=latattr,lonattr,0.123,0.456';

One very important note that it is much more efficient to allow Sphinx to perform sorting, filtering and slicing the result set than to raise max matches count and use WHERE, ORDER BY and LIMIT clauses on MySQL side. This is for two reasons. First, Sphinx does a number of optimizations and performs better than MySQL on these tasks. Second, less data would need to be packed by searchd, transferred and unpacked by SphinxSE.

Starting with version 0.9.9-rc1, additional query info besides result set could be retrieved with SHOW ENGINE SPHINX STATUS statement:

| Type   | Name  | Status                                          |
| SPHINX | stats | total: 25, total found: 25, time: 126, words: 2 | 
| SPHINX | words | sphinx:591:1256 soft:11076:15945                | 
2 rows in set (0.00 sec)

This information can also be accessed through status variables. Note that this method does not require super-user privileges.

mysql> SHOW STATUS LIKE 'sphinx_%';
| Variable_name      | Value                            |
| sphinx_total       | 25                               | 
| sphinx_total_found | 25                               | 
| sphinx_time        | 126                              | 
| sphinx_word_count  | 2                                | 
| sphinx_words       | sphinx:591:1256 soft:11076:15945 | 
5 rows in set (0.00 sec)

You could perform JOINs on SphinxSE search table and tables using other engines. Here's an example with "documents" from example.sql:

mysql> SELECT content, date_added FROM test.documents docs
-> JOIN t1 ON (docs.id=t1.id) 
-> WHERE query="one document;mode=any";
| content                             | docdate             |
| this is my test document number two | 2006-06-17 14:04:28 | 
| this is my test document number one | 2006-06-17 14:04:28 | 
2 rows in set (0.00 sec)

| Type   | Name  | Status                                      |
| SPHINX | stats | total: 2, total found: 2, time: 0, words: 2 | 
| SPHINX | words | one:1:2 document:2:2                        | 
2 rows in set (0.00 sec)

9.4. Building snippets (excerpts) via MySQL

Starting with version 0.9.9-rc2, SphinxSE also includes a UDF function that lets you create snippets through MySQL. The functionality is fully similar to BuildExcerprts API call but accesible through MySQL+SphinxSE.

The binary that provides the UDF is named sphinx.so and should be automatically built and installed to proper location along with SphinxSE itself. If it does not get installed automatically for some reason, look for sphinx.so in the build directory and copy it to the plugins directory of your MySQL instance. After that, register the UDF using the following statement:

CREATE FUNCTION sphinx_snippets RETURNS STRING SONAME 'sphinx.so';

Function name must be sphinx_snippets, you can not use an arbitrary name. Function arguments are as follows:

Prototype: function sphinx_snippets ( document, index, words, [options] );

Document and words arguments can be either strings or table columns. Options must be specified like this: 'value' AS option_name. For a list of supported options, refer to BuildExcerprts() API call. The only UDF-specific additional option is named 'sphinx' and lets you specify searchd location (host and port).

Usage examples:

SELECT sphinx_snippets('hello world doc', 'main', 'world',
    'sphinx://' AS sphinx, true AS exact_phrase,
    '[b]' AS before_match, '[/b]' AS after_match)
FROM documents;

SELECT title, sphinx_snippets(text, 'index', 'mysql php') AS text
    FROM sphinx, documents
    WHERE query='mysql php' AND sphinx.id=documents.id;

Chapter 10. Reporting bugs

Unfortunately, Sphinx is not yet 100% bug free (even though I'm working hard towards that), so you might occasionally run into some issues.

Reporting as much as possible about each bug is very important - because to fix it, I need to be able either to reproduce and debug the bug, or to deduce what's causing it from the information that you provide. So here are some instructions on how to do that.

Build-time issues

If Sphinx fails to build for some reason, please do the following:

  1. check that headers and libraries for your DBMS are properly installed (for instance, check that mysql-devel package is present);

  2. report Sphinx version and config file (be sure to remove the passwords!), MySQL (or PostgreSQL) configuration info, gcc version, OS version and CPU type (ie. x86, x86-64, PowerPC, etc):

    gcc --version
    uname -a

  3. report the error message which is produced by configure or gcc (it should be to include error message itself only, not the whole build log).

Run-time issues

If Sphinx builds and runs, but there are any problems running it, please do the following:

  1. describe the bug (ie. both the expected behavior and actual behavior) and all the steps necessary to reproduce it;

  2. include Sphinx version and config file (be sure to remove the passwords!), MySQL (or PostgreSQL) version, gcc version, OS version and CPU type (ie. x86, x86-64, PowerPC, etc):

    mysql --version
    gcc --version
    uname -a

  3. build, install and run debug versions of all Sphinx programs (this is to enable a lot of additional internal checks, so-called assertions):

    make distclean
    ./configure --with-debug
    make install
    killall -TERM searchd

  4. reindex to check if any assertions are triggered (in this case, it's likely that the index is corrupted and causing problems);

  5. if the bug does not reproduce with debug versions, revert to non-debug and mention it in your report;

  6. if the bug could be easily reproduced with a small (1-100 record) part of your database, please provide a gzipped dump of that part;

  7. if the problem is related to searchd, include relevant entries from searchd.log and query.log in your bug report;

  8. if the problem is related to searchd, try running it in console mode and check if it dies with an assertion:

    ./searchd --console

  9. if any program dies with an assertion, provide the assertion message.

Debugging assertions, crashes and hangups

If any program dies with an assertion, crashes without an assertion or hangs up, you would additionally need to generate a core dump and examine it.

  1. enable core dumps. On most Linux systems, this is done using ulimit:

    ulimit -c 32768

  2. run the program and try to reproduce the bug;

  3. if the program crashes (either with or without an assertion), find the core file in current directory (it should typically print out "Segmentation fault (core dumped)" message);

  4. if the program hangs, use kill -SEGV from another console to force it to exit and dump core:


  5. use gdb to examine the core file and obtain a backtrace:

    (gdb) bt
    (gdb) quit

Note that HANGED-PROCESS-ID, CRASHED-PROGRAM-FILE-NAME and CORE-DUMP-FILE-NAME must all be replaced with specific numbers and file names. For example, hanged searchd debugging session would look like:

# kill -SEGV 12345
# ls *core*
# gdb ./searchd core.12345
(gdb) bt
(gdb) quit

Note that ulimit is not server-wide and only affects current shell session. This means that you will not have to restore any server-wide limits - but if you relogin, you will have to set ulimit again.

Core dumps should be placed in current working directory (and Sphinx programs do not change it), so this is where you would look for them.

Please do not immediately remove the core file because there could be additional helpful information which could be retrieved from it. You do not need to send me this file (as the debug info there is closely tied to your system) but I might need to ask you a few additional questions about it.

Chapter 11. sphinx.conf options reference

Table of Contents

11.1. Data source configuration options
11.1.1. type
11.1.2. sql_host
11.1.3. sql_port
11.1.4. sql_user
11.1.5. sql_pass
11.1.6. sql_db
11.1.7. sql_sock
11.1.8. mysql_connect_flags
11.1.9. mysql_ssl_cert, mysql_ssl_key, mysql_ssl_ca
11.1.10. odbc_dsn
11.1.11. sql_query_pre
11.1.12. sql_query
11.1.13. sql_joined_field
11.1.14. sql_query_range
11.1.15. sql_range_step
11.1.16. sql_query_killlist
11.1.17. sql_attr_uint
11.1.18. sql_attr_bool
11.1.19. sql_attr_bigint
11.1.20. sql_attr_timestamp
11.1.21. sql_attr_str2ordinal
11.1.22. sql_attr_float
11.1.23. sql_attr_multi
11.1.24. sql_attr_string
11.1.25. sql_attr_str2wordcount
11.1.26. sql_column_buffers
11.1.27. sql_field_string
11.1.28. sql_field_str2wordcount
11.1.29. sql_file_field
11.1.30. sql_query_post
11.1.31. sql_query_post_index
11.1.32. sql_ranged_throttle
11.1.33. sql_query_info
11.1.34. xmlpipe_command
11.1.35. xmlpipe_field
11.1.36. xmlpipe_field_string
11.1.37. xmlpipe_field_wordcount
11.1.38. xmlpipe_attr_uint
11.1.39. xmlpipe_attr_bigint
11.1.40. xmlpipe_attr_bool
11.1.41. xmlpipe_attr_timestamp
11.1.42. xmlpipe_attr_str2ordinal
11.1.43. xmlpipe_attr_float
11.1.44. xmlpipe_attr_multi
11.1.45. xmlpipe_attr_multi_64
11.1.46. xmlpipe_attr_string
11.1.47. xmlpipe_fixup_utf8
11.1.48. mssql_winauth
11.1.49. mssql_unicode
11.1.50. unpack_zlib
11.1.51. unpack_mysqlcompress
11.1.52. unpack_mysqlcompress_maxsize
11.2. Index configuration options
11.2.1. type
11.2.2. source
11.2.3. path
11.2.4. docinfo
11.2.5. mlock
11.2.6. morphology
11.2.7. dict
11.2.8. index_sp
11.2.9. index_zones
11.2.10. min_stemming_len
11.2.11. stopwords
11.2.12. wordforms
11.2.13. exceptions
11.2.14. min_word_len
11.2.15. charset_type
11.2.16. charset_table
11.2.17. ignore_chars
11.2.18. min_prefix_len
11.2.19. min_infix_len
11.2.20. prefix_fields
11.2.21. infix_fields
11.2.22. enable_star
11.2.23. ngram_len
11.2.24. ngram_chars
11.2.25. phrase_boundary
11.2.26. phrase_boundary_step
11.2.27. html_strip
11.2.28. html_index_attrs
11.2.29. html_remove_elements
11.2.30. local
11.2.31. agent
11.2.32. agent_blackhole
11.2.33. agent_connect_timeout
11.2.34. agent_query_timeout
11.2.35. preopen
11.2.36. ondisk_dict
11.2.37. inplace_enable
11.2.38. inplace_hit_gap
11.2.39. inplace_docinfo_gap
11.2.40. inplace_reloc_factor
11.2.41. inplace_write_factor
11.2.42. index_exact_words
11.2.43. overshort_step
11.2.44. stopword_step
11.2.45. hitless_words
11.2.46. expand_keywords
11.2.47. blend_chars
11.2.48. blend_mode
11.2.49. rt_mem_limit
11.2.50. rt_field
11.2.51. rt_attr_uint
11.2.52. rt_attr_bigint
11.2.53. rt_attr_float
11.2.54. rt_attr_multi
11.2.55. rt_attr_multi_64
11.2.56. rt_attr_timestamp
11.2.57. rt_attr_string
11.3. indexer program configuration options
11.3.1. mem_limit
11.3.2. max_iops
11.3.3. max_iosize
11.3.4. max_xmlpipe2_field
11.3.5. write_buffer
11.3.6. max_file_field_buffer
11.3.7. on_file_field_error
11.4. searchd program configuration options
11.4.1. listen
11.4.2. address
11.4.3. port
11.4.4. log
11.4.5. query_log
11.4.6. query_log_format
11.4.7. read_timeout
11.4.8. client_timeout
11.4.9. max_children
11.4.10. pid_file
11.4.11. max_matches
11.4.12. seamless_rotate
11.4.13. preopen_indexes
11.4.14. unlink_old
11.4.15. attr_flush_period
11.4.16. ondisk_dict_default
11.4.17. max_packet_size
11.4.18. mva_updates_pool
11.4.19. crash_log_path
11.4.20. max_filters
11.4.21. max_filter_values
11.4.22. listen_backlog
11.4.23. read_buffer
11.4.24. read_unhinted
11.4.25. max_batch_queries
11.4.26. subtree_docs_cache
11.4.27. subtree_hits_cache
11.4.28. workers
11.4.29. dist_threads
11.4.30. binlog_path
11.4.31. binlog_flush
11.4.32. binlog_max_log_size
11.4.33. collation_server
11.4.34. collation_libc_locale
11.4.35. plugin_dir
11.4.36. mysql_version_string
11.4.37. rt_flush_period
11.4.38. thread_stack
11.4.39. expansion_limit
11.4.40. compat_sphinxql_magics
11.4.41. watchdog
11.4.42. prefork_rotation_throttle

11.1. Data source configuration options

11.1.1. type

Data source type. Mandatory, no default value. Known types are mysql, pgsql, mssql, xmlpipe and xmlpipe2, and odbc.

All other per-source options depend on source type selected by this option. Names of the options used for SQL sources (ie. MySQL, PostgreSQL, MS SQL) start with "sql_"; names of the ones used for xmlpipe and xmlpipe2 start with "xmlpipe_". All source types except xmlpipe are conditional; they might or might not be supported depending on your build settings, installed client libraries, etc. mssql type is currently only available on Windows. odbc type is available both on Windows natively and on Linux through UnixODBC library.


type = mysql

11.1.2. sql_host

SQL server host to connect to. Mandatory, no default value. Applies to SQL source types (mysql, pgsql, mssql) only.

In the simplest case when Sphinx resides on the same host with your MySQL or PostgreSQL installation, you would simply specify "localhost". Note that MySQL client library chooses whether to connect over TCP/IP or over UNIX socket based on the host name. Specifically "localhost" will force it to use UNIX socket (this is the default and generally recommended mode) and "" will force TCP/IP usage. Refer to MySQL manual for more details.


sql_host = localhost

11.1.3. sql_port

SQL server IP port to connect to. Optional, default is 3306 for mysql source type and 5432 for pgsql type. Applies to SQL source types (mysql, pgsql, mssql) only. Note that it depends on sql_host setting whether this value will actually be used.


sql_port = 3306

11.1.4. sql_user

SQL user to use when connecting to sql_host. Mandatory, no default value. Applies to SQL source types (mysql, pgsql, mssql) only.


sql_user = test

11.1.5. sql_pass

SQL user password to use when connecting to sql_host. Mandatory, no default value. Applies to SQL source types (mysql, pgsql, mssql) only.


sql_pass = mysecretpassword

11.1.6. sql_db

SQL database (in MySQL terms) to use after the connection and perform further queries within. Mandatory, no default value. Applies to SQL source types (mysql, pgsql, mssql) only.


sql_db = test

11.1.7. sql_sock

UNIX socket name to connect to for local SQL servers. Optional, default value is empty (use client library default settings). Applies to SQL source types (mysql, pgsql, mssql) only.

On Linux, it would typically be /var/lib/mysql/mysql.sock. On FreeBSD, it would typically be /tmp/mysql.sock. Note that it depends on sql_host setting whether this value will actually be used.


sql_sock = /tmp/mysql.sock

11.1.8. mysql_connect_flags

MySQL client connection flags. Optional, default value is 0 (do not set any flags). Applies to mysql source type only.

This option must contain an integer value with the sum of the flags. The value will be passed to mysql_real_connect() verbatim. The flags are enumerated in mysql_com.h include file. Flags that are especially interesting in regard to indexing, with their respective values, are as follows:

  • CLIENT_COMPRESS = 32; can use compression protocol

  • CLIENT_SSL = 2048; switch to SSL after handshake

  • CLIENT_SECURE_CONNECTION = 32768; new 4.1 authentication

For instance, you can specify 2080 (2048+32) to use both compression and SSL, or 32768 to use new authentication only. Initially, this option was introduced to be able to use compression when the indexer and mysqld are on different hosts. Compression on 1 Gbps links is most likely to hurt indexing time though it reduces network traffic, both in theory and in practice. However, enabling compression on 100 Mbps links may improve indexing time significantly (upto 20-30% of the total indexing time improvement was reported). Your mileage may vary.


mysql_connect_flags = 32 # enable compression

11.1.9. mysql_ssl_cert, mysql_ssl_key, mysql_ssl_ca

SSL certificate settings to use for connecting to MySQL server. Optional, default values are empty strings (do not use SSL). Applies to mysql source type only.

These directives let you set up secure SSL connection between indexer and MySQL. The details on creating the certificates and setting up MySQL server can be found in MySQL documentation.


mysql_ssl_cert = /etc/ssl/client-cert.pem
mysql_ssl_key = /etc/ssl/client-key.pem
mysql_ssl_ca = /etc/ssl/cacert.pem

11.1.10. odbc_dsn

ODBC DSN to connect to. Mandatory, no default value. Applies to odbc source type only.

ODBC DSN (Data Source Name) specifies the credentials (host, user, password, etc) to use when connecting to ODBC data source. The format depends on specific ODBC driver used.


odbc_dsn = Driver={Oracle ODBC Driver};Dbq=myDBName;Uid=myUsername;Pwd=myPassword

11.1.11. sql_query_pre

Pre-fetch query, or pre-query. Multi-value, optional, default is empty list of queries. Applies to SQL source types (mysql, pgsql, mssql) only.

Multi-value means that you can specify several pre-queries. They are executed before the main fetch query, and they will be exectued exactly in order of appeareance in the configuration file. Pre-query results are ignored.

Pre-queries are useful in a lot of ways. They are used to setup encoding, mark records that are going to be indexed, update internal counters, set various per-connection SQL server options and variables, and so on.

Perhaps the most frequent pre-query usage is to specify the encoding that the server will use for the rows it returnes. It must match the encoding that Sphinx expects (as specified by charset_type and charset_table options). Two MySQL specific examples of setting the encoding are:

sql_query_pre = SET CHARACTER_SET_RESULTS=cp1251
sql_query_pre = SET NAMES utf8

Also specific to MySQL sources, it is useful to disable query cache (for indexer connection only) in pre-query, because indexing queries are not going to be re-run frequently anyway, and there's no sense in caching their results. That could be achieved with:

sql_query_pre = SET SESSION query_cache_type=OFF


sql_query_pre = SET NAMES utf8
sql_query_pre = SET SESSION query_cache_type=OFF

11.1.12. sql_query

Main document fetch query. Mandatory, no default value. Applies to SQL source types (mysql, pgsql, mssql) only.

There can be only one main query. This is the query which is used to retrieve documents from SQL server. You can specify up to 32 full-text fields (formally, upto SPH_MAX_FIELDS from sphinx.h), and an arbitrary amount of attributes. All of the columns that are neither document ID (the first one) nor attributes will be full-text indexed.

Document ID MUST be the very first field, and it MUST BE UNIQUE UNSIGNED POSITIVE (NON-ZERO, NON-NEGATIVE) INTEGER NUMBER. It can be either 32-bit or 64-bit, depending on how you built Sphinx; by default it builds with 32-bit IDs support but --enable-id64 option to configure allows to build with 64-bit document and word IDs support.


sql_query = \
	SELECT id, group_id, UNIX_TIMESTAMP(date_added) AS date_added, \
		title, content \
	FROM documents

11.1.13. sql_joined_field

Joined/payload field fetch query. Multi-value, optional, default is empty list of queries. Applies to SQL source types (mysql, pgsql, mssql) only.

sql_joined_field lets you use two different features: joined fields, and payloads (payload fields). It's syntax is as follows:

sql_joined_field = FIELD-NAME 'from'  ( 'query' | 'payload-query' ); \


  • FIELD-NAME is a joined/payload field name;

  • QUERY is an SQL query that must fetch values to index.

  • RANGE-QUERY is an optional SQL query that fetches a range of values to index. (Added in version 2.0.1-beta.)

Joined fields let you avoid JOIN and/or GROUP_CONCAT statements in the main document fetch query (sql_query). This can be useful when SQL-side JOIN is slow, or needs to be offloaded on Sphinx side, or simply to emulate MySQL-specific GROUP_CONCAT funcionality in case your database server does not support it.

The query must return exactly 2 columns: document ID, and text to append to a joined field. Document IDs can be duplicate, but they must be in ascending order. All the text rows fetched for a given ID will be concatented together, and the concatenation result will be indexed as the entire contents of a joined field. Rows will be concatenated in the order returned from the query, and separating whitespace will be inserted between them. For instance, if joined field query returns the following rows:

( 1, 'red' )
( 1, 'right' )
( 1, 'hand' )
( 2, 'mysql' )
( 2, 'sphinx' )

then the indexing results would be equivalent to that of adding a new text field with a value of 'red right hand' to document 1 and 'mysql sphinx' to document 2.

Joined fields are only indexed differently. There are no other differences between joined fields and regular text fields.

Starting with 2.0.1-beta, ranged queries can be used when a single query is not efficient enough or does not work because of the database driver limitations. It works similar to the ranged queries in the main indexing loop, see Section 3.8, “Ranged queries”. The range will be queried for and fetched upfront once, then multiple queries with different $start and $end substitutions will be run to fetch the actual data.

Payloads let you create a special field in which, instead of keyword positions, so-called user payloads are stored. Payloads are custom integer values attached to every keyword. They can then be used in search time to affect the ranking.

The payload query must return exactly 3 columns: document ID; keyword; and integer payload value. Document IDs can be duplicate, but they must be in ascending order. Payloads must be unsigned integers within 24-bit range, ie. from 0 to 16777215. For reference, payloads are currently internally stored as in-field keyword positions, but that is not guaranteed and might change in the future.

Currently, the only method to account for payloads is to use SPH_RANK_PROXIMITY_BM25 ranker. On indexes with payload fields, it will automatically switch to a variant that matches keywords in those fields, computes a sum of matched payloads multiplied by field wieghts, and adds that sum to the final rank.


sql_joined_field = \
	tagstext from query; \
	SELECT docid, CONCAT('tag',tagid) FROM tags ORDER BY docid ASC

11.1.14. sql_query_range

Range query setup. Optional, default is empty. Applies to SQL source types (mysql, pgsql, mssql) only.

Setting this option enables ranged document fetch queries (see Section 3.8, “Ranged queries”). Ranged queries are useful to avoid notorious MyISAM table locks when indexing lots of data. (They also help with other less notorious issues, such as reduced performance caused by big result sets, or additional resources consumed by InnoDB to serialize big read transactions.)

The query specified in this option must fetch min and max document IDs that will be used as range boundaries. It must return exactly two integer fields, min ID first and max ID second; the field names are ignored.

When ranged queries are enabled, sql_query will be required to contain $start and $end macros (because it obviously would be a mistake to index the whole table many times over). Note that the intervals specified by $start..$end will not overlap, so you should not remove document IDs that are exactly equal to $start or $end from your query. The example in Section 3.8, “Ranged queries”) illustrates that; note how it uses greater-or-equal and less-or-equal comparisons.


sql_query_range = SELECT MIN(id),MAX(id) FROM documents

11.1.15. sql_range_step

Range query step. Optional, default is 1024. Applies to SQL source types (mysql, pgsql, mssql) only.

Only used when ranged queries are enabled. The full document IDs interval fetched by sql_query_range will be walked in this big steps. For example, if min and max IDs fetched are 12 and 3456 respectively, and the step is 1000, indexer will call sql_query several times with the following substitutions:

  • $start=12, $end=1011

  • $start=1012, $end=2011

  • $start=2012, $end=3011

  • $start=3012, $end=3456


sql_range_step = 1000

11.1.16. sql_query_killlist

Kill-list query. Optional, default is empty (no query). Applies to SQL source types (mysql, pgsql, mssql) only. Introduced in version 0.9.9-rc1.

This query is expected to return a number of 1-column rows, each containing just the document ID. The returned document IDs are stored within an index. Kill-list for a given index suppresses results from other indexes, depending on index order in the query. The intended use is to help implement deletions and updates on existing indexes without rebuilding (actually even touching them), and especially to fight phantom results problem.

Let us dissect an example. Assume we have two indexes, 'main' and 'delta'. Assume that documents 2, 3, and 5 were deleted since last reindex of 'main', and documents 7 and 11 were updated (ie. their text contents were changed). Assume that a keyword 'test' occurred in all these mentioned documents when we were indexing 'main'; still occurs in document 7 as we index 'delta'; but does not occur in document 11 any more. We now reindex delta and then search through both these indexes in proper (least to most recent) order:

$res = $cl->Query ( "test", "main delta" );

First, we need to properly handle deletions. The result set should not contain documents 2, 3, or 5. Second, we also need to avoid phantom results. Unless we do something about it, document 11 will appear in search results! It will be found in 'main' (but not 'delta'). And it will make it to the final result set unless something stops it.

Kill-list, or K-list for short, is that something. Kill-list attached to 'delta' will suppress the specified rows from all the preceding indexes, in this case just 'main'. So to get the expected results, we should put all the updated and deleted document IDs into it.


sql_query_killlist = \
	SELECT id FROM documents WHERE updated_ts>[email protected]_reindex UNION \
	SELECT id FROM documents_deleted WHERE deleted_ts>[email protected]_reindex

11.1.17. sql_attr_uint

Unsigned integer attribute declaration. Multi-value (there might be multiple attributes declared), optional. Applies to SQL source types (mysql, pgsql, mssql) only.

The column value should fit into 32-bit unsigned integer range. Values outside this range will be accepted but wrapped around. For instance, -1 will be wrapped around to 2^32-1 or 4,294,967,295.

You can specify bit count for integer attributes by appending ':BITCOUNT' to attribute name (see example below). Attributes with less than default 32-bit size, or bitfields, perform slower. But they require less RAM when using extern storage: such bitfields are packed together in 32-bit chunks in .spa attribute data file. Bit size settings are ignored if using inline storage.


sql_attr_uint = group_id
sql_attr_uint = forum_id:9 # 9 bits for forum_id

11.1.18. sql_attr_bool

Boolean attribute declaration. Multi-value (there might be multiple attributes declared), optional. Applies to SQL source types (mysql, pgsql, mssql) only. Equivalent to sql_attr_uint declaration with a bit count of 1.


sql_attr_bool = is_deleted # will be packed to 1 bit

11.1.19. sql_attr_bigint

64-bit signed integer attribute declaration. Multi-value (there might be multiple attributes declared), optional. Applies to SQL source types (mysql, pgsql, mssql) only. Note that unlike sql_attr_uint, these values are signed. Introduced in version 0.9.9-rc1.


sql_attr_bigint = my_bigint_id

11.1.20. sql_attr_timestamp

UNIX timestamp attribute declaration. Multi-value (there might be multiple attributes declared), optional. Applies to SQL source types (mysql, pgsql, mssql) only.

Timestamps can store date and time in the range of Jan 01, 1970 to Jan 19, 2038 with a precision of one second. The expected column value should be a timestamp in UNIX format, ie. 32-bit unsigned integer number of seconds elapsed since midnight, January 01, 1970, GMT. Timestamps are internally stored and handled as integers everywhere. But in addition to working with timestamps as integers, it's also legal to use them along with different date-based functions, such as time segments sorting mode, or day/week/month/year extraction for GROUP BY.

Note that DATE or DATETIME column types in MySQL can not be directly used as timestamp attributes in Sphinx; you need to explicitly convert such columns using UNIX_TIMESTAMP function (if data is in range).

Note timestamps can not represent dates before January 01, 1970, and UNIX_TIMESTAMP() in MySQL will not return anything expected. If you only needs to work with dates, not times, consider TO_DAYS() function in MySQL instead.


# sql_query = ... UNIX_TIMESTAMP(added_datetime) AS added_ts ...
sql_attr_timestamp = added_ts

11.1.21. sql_attr_str2ordinal

Ordinal string number attribute declaration. Multi-value (there might be multiple attributes declared), optional. Applies to SQL source types (mysql, pgsql, mssql) only.

This attribute type (so-called ordinal, for brevity) is intended to allow sorting by string values, but without storing the strings themselves. When indexing ordinals, string values are fetched from database, temporarily stored, sorted, and then replaced by their respective ordinal numbers in the array of sorted strings. So, the ordinal number is an integer such that sorting by it produces the same result as if lexicographically sorting by original strings. by string values lexicographically.

Earlier versions could consume a lot of RAM for indexing ordinals. Starting with revision r1112, ordinals accumulation and sorting also runs in fixed memory (at the cost of using additional temporary disk space), and honors mem_limit settings.

Ideally the strings should be sorted differently, depending on the encoding and locale. For instance, if the strings are known to be Russian text in KOI8R encoding, sorting the bytes 0xE0, 0xE1, and 0xE2 should produce 0xE1, 0xE2 and 0xE0, because in KOI8R value 0xE0 encodes a character that is (noticeably) after characters encoded by 0xE1 and 0xE2. Unfortunately, Sphinx does not support that at the moment and will simply sort the strings bytewise.

Note that the ordinals are by construction local to each index, and it's therefore impossible to merge ordinals while retaining the proper order. The processed strings are replaced by their sequential number in the index they occurred in, but different indexes have different sets of strings. For instance, if 'main' index contains strings "aaa", "bbb", "ccc", and so on up to "zzz", they'll be assigned numbers 1, 2, 3, and so on up to 26, respectively. But then if 'delta' only contains "zzz" the assigned number will be 1. And after the merge, the order will be broken. Unfortunately, this is impossible to workaround without storing the original strings (and once Sphinx supports storing the original strings, ordinals will not be necessary any more).


sql_attr_str2ordinal = author_name

11.1.22. sql_attr_float

Floating point attribute declaration. Multi-value (there might be multiple attributes declared), optional. Applies to SQL source types (mysql, pgsql, mssql) only.

The values will be stored in single precision, 32-bit IEEE 754 format. Represented range is approximately from 1e-38 to 1e+38. The amount of decimal digits that can be stored precisely is approximately 7. One important usage of the float attributes is storing latitude and longitude values (in radians), for further usage in query-time geosphere distance calculations.


sql_attr_float = lat_radians
sql_attr_float = long_radians

11.1.23. sql_attr_multi

Multi-valued attribute (MVA) declaration. Multi-value (ie. there may be more than one such attribute declared), optional. Applies to SQL source types (mysql, pgsql, mssql) only.

Plain attributes only allow to attach 1 value per each document. However, there are cases (such as tags or categories) when it is desired to attach multiple values of the same attribute and be able to apply filtering or grouping to value lists.

The declaration format is as follows (backslashes are for clarity only; everything can be declared in a single line as well):

sql_attr_multi = ATTR-TYPE ATTR-NAME 'from' SOURCE-TYPE \
	[;QUERY] \


  • ATTR-TYPE is 'uint', 'bigint' or 'timestamp'

  • SOURCE-TYPE is 'field', 'query', or 'ranged-query'

  • QUERY is SQL query used to fetch all ( docid, attrvalue ) pairs

  • RANGE-QUERY is SQL query used to fetch min and max ID values, similar to 'sql_query_range'


sql_attr_multi = uint tag from query; SELECT id, tag FROM tags
sql_attr_multi = bigint tag from ranged-query; \
	SELECT id, tag FROM tags WHERE id>=$start AND id<=$end; \
	SELECT MIN(id), MAX(id) FROM tags

11.1.24. sql_attr_string

String attribute declaration. Multi-value (ie. there may be more than one such attribute declared), optional. Applies to SQL source types (mysql, pgsql, mssql) only. Introduced in version 1.10-beta.

String attributes can store arbitrary strings attached to every document. There's a fixed size limit of 4 MB per value. Also, searchd will currently cache all the values in RAM, which is an additional implicit limit.

As of 1.10-beta, strings can only be used for storage and retrieval. They can not participate in expressions, be used for filtering, sorting, or grouping (ie. in WHERE, ORDER or GROUP clauses). Note that attributes declared using sql_attr_string will not be full-text indexed; you can use sql_field_string directive for that.


sql_attr_string = title # will be stored but will not be indexed

11.1.25. sql_attr_str2wordcount

Word-count attribute declaration. Multi-value (ie. there may be more than one such attribute declared), optional. Applies to SQL source types (mysql, pgsql, mssql) only. Introduced in version 1.10-beta.

Word-count attribute takes a string column, tokenizes it according to index settings, and stores the resulting number of tokens in an attribute. This number of tokens ("word count") is a normal integer that can be later used, for instance, in custom ranking expressions (boost shorter titles, help identify exact field matches, etc).


sql_attr_str2wordcount = title_wc

11.1.26. sql_column_buffers

Per-column buffer sizes. Optional, default is empty (deduce the sizes automatically). Applies to odbc, mssql source types only. Introduced in version 2.0.1-beta.

ODBC and MS SQL drivers sometimes can not return the maximum actual column size to be expected. For instance, NVARCHAR(MAX) columns always report their length as 2147483647 bytes to indexer even though the actually used length is likely considerably less. However, the receiving buffers still need to be allocated upfront, and their sizes have to be determined. When the driver does not report the column length at all, Sphinx allocates default 1 KB buffers for each non-char column, and 1 MB buffers for each char column. Driver-reported column length also gets clamped by an upper limie of 8 MB, so in case the driver reports (almost) a 2 GB column length, it will be clamped and a 8 MB buffer will be allocated instead for that column. These hard-coded limits can be overridden using the sql_column_buffers directive, either in order to save memory on actually shorter columns, or overcome the 8 MB limit on actually longer columns. The directive values must be a comma-separated lists of selected column names and sizes:

sql_column_buffers = <colname>=<size>[K|M] [, ...]


sql_query = SELECT id, mytitle, mycontent FROM documents
sql_column_buffers = mytitle=64K, mycontent=10M

11.1.27. sql_field_string

Combined string attribute and full-text field declaration. Multi-value (ie. there may be more than one such attribute declared), optional. Applies to SQL source types (mysql, pgsql, mssql) only. Introduced in version 1.10-beta.

sql_attr_string only stores the column value but does not full-text index it. In some cases it might be desired to both full-text index the column and store it as attribute. sql_field_string lets you do exactly that. Both the field and the attribute will be named the same.


sql_field_string = title # will be both indexed and stored

11.1.28. sql_field_str2wordcount

Combined word-count attribute and full-text field declaration. Multi-value (ie. there may be more than one such attribute declared), optional. Applies to SQL source types (mysql, pgsql, mssql) only. Introduced in version 1.10-beta.

sql_attr_str2wordcount only stores the column word count but does not full-text index it. In some cases it might be desired to both full-text index the column and also have the count. sql_field_str2wordcount lets you do exactly that. Both the field and the attribute will be named the same.


sql_field_str2wordcount = title # will be indexed, and counted/stored

11.1.29. sql_file_field

File based field declaration. Applies to SQL source types (mysql, pgsql, mssql) only. Introduced in version 1.10-beta.

This directive makes indexer interpret field contents as a file name, and load and index the referred file. Files larger than max_file_field_buffer in size are skipped. Any errors during the file loading (IO errors, missed limits, etc) will be reported as indexing warnings and will not early terminate the indexing. No content will be indexed for such files.


sql_file_field = my_file_path # load and index files referred to by my_file_path

11.1.30. sql_query_post

Post-fetch query. Optional, default value is empty. Applies to SQL source types (mysql, pgsql, mssql) only.

This query is executed immediately after sql_query completes successfully. When post-fetch query produces errors, they are reported as warnings, but indexing is not terminated. It's result set is ignored. Note that indexing is not yet completed at the point when this query gets executed, and further indexing still may fail. Therefore, any permanent updates should not be done from here. For instance, updates on helper table that permanently change the last successfully indexed ID should not be run from post-fetch query; they should be run from post-index query instead.


sql_query_post = DROP TABLE my_tmp_table

11.1.31. sql_query_post_index

Post-index query. Optional, default value is empty. Applies to SQL source types (mysql, pgsql, mssql) only.

This query is executed when indexing is fully and succesfully completed. If this query produces errors, they are reported as warnings, but indexing is not terminated. It's result set is ignored. $maxid macro can be used in its text; it will be expanded to maximum document ID which was actually fetched from the database during indexing. If no documents were indexed, $maxid will be expanded to 0.


sql_query_post_index = REPLACE INTO counters ( id, val ) \
    VALUES ( 'max_indexed_id', $maxid )

11.1.32. sql_ranged_throttle

Ranged query throttling period, in milliseconds. Optional, default is 0 (no throttling). Applies to SQL source types (mysql, pgsql, mssql) only.

Throttling can be useful when indexer imposes too much load on the database server. It causes the indexer to sleep for given amount of milliseconds once per each ranged query step. This sleep is unconditional, and is performed before the fetch query.


sql_ranged_throttle = 1000 # sleep for 1 sec before each query step

11.1.33. sql_query_info

Document info query. Optional, default is empty. Applies to mysql source type only.

Only used by CLI search to fetch and display document information, only works with MySQL at the moment, and only intended for debugging purposes. This query fetches the row that will be displayed by CLI search utility for each document ID. It is required to contain $id macro that expands to the queried document ID.


sql_query_info = SELECT * FROM documents WHERE id=$id

11.1.34. xmlpipe_command

Shell command that invokes xmlpipe stream producer. Mandatory. Applies to xmlpipe and xmlpipe2 source types only.

Specifies a command that will be executed and which output will be parsed for documents. Refer to Section 3.9, “xmlpipe data source” or Section 3.10, “xmlpipe2 data source” for specific format description.


xmlpipe_command = cat /home/sphinx/test.xml

11.1.35. xmlpipe_field

xmlpipe field declaration. Multi-value, optional. Applies to xmlpipe2 source type only. Refer to Section 3.10, “xmlpipe2 data source”.


xmlpipe_field = subject
xmlpipe_field = content

11.1.36. xmlpipe_field_string

xmlpipe field and string attribute declaration. Multi-value, optional. Applies to xmlpipe2 source type only. Refer to Section 3.10, “xmlpipe2 data source”. Introduced in version 1.10-beta.

Makes the specified XML element indexed as both a full-text field and a string attribute. Equivalent to <sphinx:field name="field" attr="string"/> declaration within the XML file.


xmlpipe_field_string = subject

11.1.37. xmlpipe_field_wordcount

xmlpipe field and word count attribute declaration. Multi-value, optional. Applies to xmlpipe2 source type only. Refer to Section 3.10, “xmlpipe2 data source”. Introduced in version 1.10-beta.

Makes the specified XML element indexed as both a full-text field and a word count attribute. Equivalent to <sphinx:field name="field" attr="wordcount"/> declaration within the XML file.


xmlpipe_field_wordcount = subject

11.1.38. xmlpipe_attr_uint

xmlpipe integer attribute declaration. Multi-value, optional. Applies to xmlpipe2 source type only. Syntax fully matches that of sql_attr_uint.


xmlpipe_attr_uint = author_id

11.1.39. xmlpipe_attr_bigint

xmlpipe signed 64-bit integer attribute declaration. Multi-value, optional. Applies to xmlpipe2 source type only. Syntax fully matches that of sql_attr_bigint.


xmlpipe_attr_bigint = my_bigint_id

11.1.40. xmlpipe_attr_bool

xmlpipe boolean attribute declaration. Multi-value, optional. Applies to xmlpipe2 source type only. Syntax fully matches that of sql_attr_bool.


xmlpipe_attr_bool = is_deleted # will be packed to 1 bit

11.1.41. xmlpipe_attr_timestamp

xmlpipe UNIX timestamp attribute declaration. Multi-value, optional. Applies to xmlpipe2 source type only. Syntax fully matches that of sql_attr_timestamp.


xmlpipe_attr_timestamp = published

11.1.42. xmlpipe_attr_str2ordinal

xmlpipe string ordinal attribute declaration. Multi-value, optional. Applies to xmlpipe2 source type only. Syntax fully matches that of sql_attr_str2ordinal.


xmlpipe_attr_str2ordinal = author_sort

11.1.43. xmlpipe_attr_float

xmlpipe floating point attribute declaration. Multi-value, optional. Applies to xmlpipe2 source type only. Syntax fully matches that of sql_attr_float.


xmlpipe_attr_float = lat_radians
xmlpipe_attr_float = long_radians

11.1.44. xmlpipe_attr_multi

xmlpipe MVA attribute declaration. Multi-value, optional. Applies to xmlpipe2 source type only.

This setting declares an MVA attribute tag in xmlpipe2 stream. The contents of the specified tag will be parsed and a list of integers that will constitute the MVA will be extracted, similar to how sql_attr_multi parses SQL column contents when 'field' MVA source type is specified.


xmlpipe_attr_multi = taglist

11.1.45. xmlpipe_attr_multi_64

xmlpipe MVA attribute declaration. Declares the BIGINT (signed 64-bit integer) MVA attribute. Multi-value, optional. Applies to xmlpipe2 source type only.

This setting declares an MVA attribute tag in xmlpipe2 stream. The contents of the specified tag will be parsed and a list of integers that will constitute the MVA will be extracted, similar to how sql_attr_multi parses SQL column contents when 'field' MVA source type is specified.


xmlpipe_attr_multi_64 = taglist

11.1.46. xmlpipe_attr_string

xmlpipe string declaration. Multi-value, optional. Applies to xmlpipe2 source type only. Introduced in version 1.10-beta.

This setting declares a string attribute tag in xmlpipe2 stream. The contents of the specified tag will be parsed and stored as a string value.


xmlpipe_attr_string = subject

11.1.47. xmlpipe_fixup_utf8

Perform Sphinx-side UTF-8 validation and filtering to prevent XML parser from choking on non-UTF-8 documents. Optional, default is 0. Applies to xmlpipe2 source type only.

Under certain occasions it might be hard or even impossible to guarantee that the incoming XMLpipe2 document bodies are in perfectly valid and conforming UTF-8 encoding. For instance, documents with national single-byte encodings could sneak into the stream. libexpat XML parser is fragile, meaning that it will stop processing in such cases. UTF8 fixup feature lets you avoid that. When fixup is enabled, Sphinx will preprocess the incoming stream before passing it to the XML parser and replace invalid UTF-8 sequences with spaces.


xmlpipe_fixup_utf8 = 1

11.1.48. mssql_winauth

MS SQL Windows authentication flag. Boolean, optional, default value is 0 (false). Applies to mssql source type only. Introduced in version 0.9.9-rc1.

Whether to use currently logged in Windows account credentials for authentication when connecting to MS SQL Server. Note that when running searchd as a service, account user can differ from the account you used to install the service.


mssql_winauth = 1

11.1.49. mssql_unicode

MS SQL encoding type flag. Boolean, optional, default value is 0 (false). Applies to mssql source type only. Introduced in version 0.9.9-rc1.

Whether to ask for Unicode or single-byte data when querying MS SQL Server. This flag must be in sync with charset_type directive; that is, to index Unicode data, you must set both charset_type in the index (to 'utf-8') and mssql_unicode in the source (to 1). For reference, MS SQL will actually return data in UCS-2 encoding instead of UTF-8, but Sphinx will automatically handle that.


mssql_unicode = 1

11.1.50. unpack_zlib

Columns to unpack using zlib (aka deflate, aka gunzip). Multi-value, optional, default value is empty list of columns. Applies to SQL source types (mysql, pgsql, mssql) only. Introduced in version 0.9.9-rc1.

Columns specified using this directive will be unpacked by indexer using standard zlib algorithm (called deflate and also implemented by gunzip). When indexing on a different box than the database, this lets you offload the database, and save on network traffic. The feature is only available if zlib and zlib-devel were both available during build time.


unpack_zlib = col1
unpack_zlib = col2

11.1.51. unpack_mysqlcompress

Columns to unpack using MySQL UNCOMPRESS() algorithm. Multi-value, optional, default value is empty list of columns. Applies to SQL source types (mysql, pgsql, mssql) only. Introduced in version 0.9.9-rc1.

Columns specified using this directive will be unpacked by indexer using modified zlib algorithm used by MySQL COMPRESS() and UNCOMPRESS() functions. When indexing on a different box than the database, this lets you offload the database, and save on network traffic. The feature is only available if zlib and zlib-devel were both available during build time.


unpack_mysqlcompress = body_compressed
unpack_mysqlcompress = description_compressed

11.1.52. unpack_mysqlcompress_maxsize

Buffer size for UNCOMPRESS()ed data. Optional, default value is 16M. Introduced in version 0.9.9-rc1.

When using unpack_mysqlcompress, due to implementation intrincacies it is not possible to deduce the required buffer size from the compressed data. So the buffer must be preallocated in advance, and unpacked data can not go over the buffer size. This option lets you control the buffer size, both to limit indexer memory use, and to enable unpacking of really long data fields if necessary.


unpack_mysqlcompress_maxsize = 1M

11.2. Index configuration options

11.2.1. type

Index type. Known values are 'plain', 'distributed', and 'rt'. Optional, default is 'plain' (plain local index).

Sphinx supports several different types of indexes. Versions 0.9.x supported two index types: plain local indexes that are stored and processed on the local machine; and distributed indexes, that involve not only local searching but querying remote searchd instances over the network as well (see Section 5.8, “Distributed searching”). Version 1.10-beta also adds support for so-called real-time indexes (or RT indexes for short) that are also stored and processed locally, but additionally allow for on-the-fly updates of the full-text index (see Chapter 4, Real-time indexes). Note that attributes can be updated on-the-fly using either plain local indexes or RT ones.

Index type setting lets you choose the needed type. By default, plain local index type will be assumed.


type = distributed

11.2.2. source

Adds document source to local index. Multi-value, mandatory.

Specifies document source to get documents from when the current index is indexed. There must be at least one source. There may be multiple sources, without any restrictions on the source types: ie. you can pull part of the data from MySQL server, part from PostgreSQL, part from the filesystem using xmlpipe2 wrapper.

However, there are some restrictions on the source data. First, document IDs must be globally unique across all sources. If that condition is not met, you might get unexpected search results. Second, source schemas must be the same in order to be stored within the same index.

No source ID is stored automatically. Therefore, in order to be able to tell what source the matched document came from, you will need to store some additional information yourself. Two typical approaches include:

  1. mangling document ID and encoding source ID in it:

    source src1
    	sql_query = SELECT id*10+1, ... FROM table1
    source src2
    	sql_query = SELECT id*10+2, ... FROM table2

  2. storing source ID simply as an attribute:

    source src1
    	sql_query = SELECT id, 1 AS source_id FROM table1
    	sql_attr_uint = source_id
    source src2
    	sql_query = SELECT id, 2 AS source_id FROM table2
    	sql_attr_uint = source_id


source = srcpart1
source = srcpart2
source = srcpart3

11.2.3. path

Index files path and file name (without extension). Mandatory.

Path specifies both directory and file name, but without extension. indexer will append different extensions to this path when generating final names for both permanent and temporary index files. Permanent data files have several different extensions starting with '.sp'; temporary files' extensions start with '.tmp'. It's safe to remove .tmp* files is if indexer fails to remove them automatically.

For reference, different index files store the following data:

  • .spa stores document attributes (used in extern docinfo storage mode only);

  • .spd stores matching document ID lists for each word ID;

  • .sph stores index header information;

  • .spi stores word lists (word IDs and pointers to .spd file);

  • .spk stores kill-lists;

  • .spm stores MVA data;

  • .spp stores hit (aka posting, aka word occurence) lists for each word ID;

  • .sps stores string attribute data.


path = /var/data/test1

11.2.4. docinfo

Document attribute values (docinfo) storage mode. Optional, default is 'extern'. Known values are 'none', 'extern' and 'inline'.

Docinfo storage mode defines how exactly docinfo will be physically stored on disk and RAM. "none" means that there will be no docinfo at all (ie. no attributes). Normally you need not to set "none" explicitly because Sphinx will automatically select "none" when there are no attributes configured. "inline" means that the docinfo will be stored in the .spd file, along with the document ID lists. "extern" means that the docinfo will be stored separately (externally) from document ID lists, in a special .spa file.

Basically, externally stored docinfo must be kept in RAM when querying. for performance reasons. So in some cases "inline" might be the only option. However, such cases are infrequent, and docinfo defaults to "extern". Refer to Section 3.3, “Attributes” for in-depth discussion and RAM usage estimates.


docinfo = inline

11.2.5. mlock

Memory locking for cached data. Optional, default is 0 (do not call mlock()).

For search performance, searchd preloads a copy of .spa and .spi files in RAM, and keeps that copy in RAM at all times. But if there are no searches on the index for some time, there are no accesses to that cached copy, and OS might decide to swap it out to disk. First queries to such "cooled down" index will cause swap-in and their latency will suffer.

Setting mlock option to 1 makes Sphinx lock physical RAM used for that cached data using mlock(2) system call, and that prevents swapping (see man 2 mlock for details). mlock(2) is a privileged call, so it will require searchd to be either run from root account, or be granted enough privileges otherwise. If mlock() fails, a warning is emitted, but index continues working.


mlock = 1

11.2.6. morphology

A list of morphology preprocessors to apply. Optional, default is empty (do not apply any preprocessor).

Morphology preprocessors can be applied to the words being indexed to replace different forms of the same word with the base, normalized form. For instance, English stemmer will normalize both "dogs" and "dog" to "dog", making search results for both searches the same.

Built-in preprocessors include English stemmer, Russian stemmer (that supports UTF-8 and Windows-1251 encodings), Soundex, and Metaphone. The latter two replace the words with special phonetic codes that are equal is words are phonetically close. Additional stemmers provided by Snowball project libstemmer library can be enabled at compile time using --with-libstemmer configure option. Built-in English and Russian stemmers should be faster than their libstemmer counterparts, but can produce slightly different results, because they are based on an older version. Metaphone implementation is based on Double Metaphone algorithm and indexes the primary code.

Built-in values that are available for use in morphology option are as follows:

  • none - do not perform any morphology processing;

  • stem_en - apply Porter's English stemmer;

  • stem_ru - apply Porter's Russian stemmer;

  • stem_enru - apply Porter's English and Russian stemmers;

  • stem_cz - apply Czech stemmer;

  • soundex - replace keywords with their SOUNDEX code;

  • metaphone - replace keywords with their METAPHONE code.

Additional values provided by libstemmer are in 'libstemmer_XXX' format, where XXX is libstemmer algorithm codename (refer to libstemmer_c/libstemmer/modules.txt for a complete list).

Several stemmers can be specified (comma-separated). They will be applied to incoming words in the order they are listed, and the processing will stop once one of the stemmers actually modifies the word. Also when wordforms feature is enabled the word will be looked up in word forms dictionary first, and if there is a matching entry in the dictionary, stemmers will not be applied at all. Or in other words, wordforms can be used to implement stemming exceptions.


morphology = stem_en, libstemmer_sv

11.2.7. dict

The keywords dictionary type. Known values are 'crc' and 'keywords'. Optional, default is 'crc'. Introduced in version 2.0.1-beta.

CRC dictionary mode (dict=crc) is the default dictionary type in Sphinx, and the only one available until version 2.0.1-beta. Keywords dictionary mode (dict=keywords) was added in 2.0.1-beta, primarly to (greatly) reduce indexing impact and enable substring searches on huge collections. They also eliminate the chance of CRC32 collisions. In 2.0.1-beta, that mode was only supported for disk indexes. Starting with 2.0.2-beta, RT indexes are also supported.

CRC dictionaries never store the original keyword text in the index. Instead, keywords are replaced with their control sum value (either CRC32 or FNV64, depending whether Sphinx was built with --enable-id64) both when searching and indexing, and that value is used internally in the index.

That approach has two drawbacks. First, in CRC32 case there is a chance of control sum collision between several pairs of different keywords, growing quadratically with the number of unique keywords in the index. (FNV64 case is unaffected in practice, as a chance of a single FNV64 collision in a dictionary of 1 billion entries is approximately 1:16, or 6.25 percent. And most dictionaries will be much more compact that a billion keywords, as a typical spoken human language has in the region of 1 to 10 million word forms.) Second, and more importantly, substring searches are not directly possible with control sums. Sphinx alleviated that by pre-indexing all the possible substrings as separate keywords (see Section 11.2.18, “min_prefix_len”, Section 11.2.19, “min_infix_len” directives). That actually has an added benefit of matching substrings in the quickest way possible. But at the same time pre-indexing all substrings grows the index size a lot (factors of 3-10x and even more would not be unusual) and impacts the indexing time respectively, rendering substring searches on big indexes rather impractical.

Keywords dictionary, introduced in 2.0.1-beta, fixes both these drawbacks. It stores the keywords in the index and performs search-time wildcard expansion. For example, a search for a 'test*' prefix could internally expand to 'test|tests|testing' query based on the dictionary contents. That expansion is fully transparent to the application, except that the separate per-keyword statistics for all the actually matched keywords would now also be reported.

Indexing with keywords dictionary should be 1.1x to 1.3x slower compared to regular, non-substring indexing - but times faster compared to substring indexing (either prefix or infix). Index size should only be slightly bigger that than of the regular non-substring index, with a 1..10% percent total difference Regular keyword searching time must be very close or identical across all three discussed index kinds (CRC non-substring, CRC substring, keywords). Substring searching time can vary greatly depending on how many actual keywords match the given substring (in other words, into how many keywords does the search term expand). The maximum number of keywords matched is restricted by the expansion_limit directive.

Essentially, keywords and CRC dictionaries represent the two different trade-off substring searching decisions. You can choose to either sacrifice indexing time and index size in favor of top-speed worst-case searches (CRC dictionary), or only slightly impact indexing time but sacrifice worst-case searching time when the prefix expands into very many keywords (keywords dictionary).


dict = keywords

11.2.8. index_sp

Whether to detect and index sentence and paragraph boundaries. Optional, default is 0 (do not detect and index). Introduced in version 2.0.1-beta.

This directive enables sentence and paragraph boundary indexing. It's required for the SENTENCE and PARAGRAPH operators to work. Sentence boundary detection is based on plain text analysis, so you only need to set index_sp = 1 to enable it. Paragraph detection is however based on HTML markup, and happens in the HTML stripper. So to index paragraph locations you also need to enable the stripper by specifying html_strip = 1. Both types of boundaries are detected based on a few built-in rules enumerated just below.

Sentence boundary detection rules are as follows.

  • Question and excalamation signs (? and !) are always a sentence boundary.

  • Trailing dot (.) is a sentence boundary, except:

    • When followed by a letter. That's considered a part of an abbreviation (as in "S.T.A.L.K.E.R" or "Goldman Sachs S.p.A.").

    • When followed by a comma. That's considered an abbreviation followed by a comma (as in "Telecom Italia S.p.A., founded in 1994").

    • When followed by a space and a small letter. That's considered an abbreviation within a sentence (as in "News Corp. announced in Februrary").

    • When preceded by a space and a capital letter, and followed by a space. That's considered a middle initial (as in "John D. Doe").

Paragraph boundaries are inserted at every block-level HTML tag. Namely, those are (as taken from HTML 4 standard) ADDRESS, BLOCKQUOTE, CAPTION, CENTER, DD, DIV, DL, DT, H1, H2, H3, H4, H5, LI, MENU, OL, P, PRE, TABLE, TBODY, TD, TFOOT, TH, THEAD, TR, and UL.

Both sentences and paragraphs increment the keyword position counter by 1.


index_sp = 1

11.2.9. index_zones

A list of in-field HTML/XML zones to index. Optional, default is empty (do not index zones). Introduced in version 2.0.1-beta.

Zones can be formally defined as follows. Everything between an opening and a matching closing tag is called a span, and the aggregate of all spans corresponding sharing the same tag name is called a zone. For instance, everything between the occurrences of <H1> and </H1> in the document field belongs to H1 zone.

Zone indexing, enabled by index_zones directive, is an optional extension of the HTML stripper. So it will also require that the stripper is enabled (with html_strip = 1). The value of the index_zones should be a comma-separated list of those tag names and wildcards (ending with a star) that should be indexed as zones.

Zones can nest and overlap arbitrarily. The only requirement is that every opening tag has a matching tag. You can also have an arbitrary number of both zones (as in unique zone names, such as H1) and spans (all the occurrences of those H1 tags) in a document. Once indexed, zones can then be used for matching with the ZONE operator, see Section 5.3, “Extended query syntax”.


index_zones = h*, th, title

11.2.10. min_stemming_len

Minimum word length at which to enable stemming. Optional, default is 1 (stem everything). Introduced in version 0.9.9-rc1.

Stemmers are not perfect, and might sometimes produce undesired results. For instance, running "gps" keyword through Porter stemmer for English results in "gp", which is not really the intent. min_stemming_len feature lets you suppress stemming based on the source word length, ie. to avoid stemming too short words. Keywords that are shorter than the given threshold will not be stemmed. Note that keywords that are exactly as long as specified will be stemmed. So in order to avoid stemming 3-character keywords, you should specify 4 for the value. For more finely grained control, refer to wordforms feature.


min_stemming_len = 4

11.2.11. stopwords

Stopword files list (space separated). Optional, default is empty.

Stopwords are the words that will not be indexed. Typically you'd put most frequent words in the stopwords list because they do not add much value to search results but consume a lot of resources to process.

You can specify several file names, separated by spaces. All the files will be loaded. Stopwords file format is simple plain text. The encoding must match index encoding specified in charset_type. File data will be tokenized with respect to charset_table settings, so you can use the same separators as in the indexed data. The stemmers will also be applied when parsing stopwords file.

While stopwords are not indexed, they still do affect the keyword positions. For instance, assume that "the" is a stopword, that document 1 contains the line "in office", and that document 2 contains "in the office". Searching for "in office" as for exact phrase will only return the first document, as expected, even though "the" in the second one is stopped.

Stopwords files can either be created manually, or semi-automatically. indexer provides a mode that creates a frequency dictionary of the index, sorted by the keyword frequency, see --buildstops and --buildfreqs switch in Section 6.1, “indexer command reference”. Top keywords from that dictionary can usually be used as stopwords.


stopwords = /usr/local/sphinx/data/stopwords.txt
stopwords = stopwords-ru.txt stopwords-en.txt

11.2.12. wordforms

Word forms dictionary. Optional, default is empty.

Word forms are applied after tokenizing the incoming text by charset_table rules. They essentialy let you replace one word with another. Normally, that would be used to bring different word forms to a single normal form (eg. to normalize all the variants such as "walks", "walked", "walking" to the normal form "walk"). It can also be used to implement stemming exceptions, because stemming is not applied to words found in the forms list.

Dictionaries are used to normalize incoming words both during indexing and searching. Therefore, to pick up changes in wordforms file it's required to reindex and restart searchd.

Word forms support in Sphinx is designed to support big dictionaries well. They moderately affect indexing speed: for instance, a dictionary with 1 million entries slows down indexing about 1.5 times. Searching speed is not affected at all. Additional RAM impact is roughly equal to the dictionary file size, and dictionaries are shared across indexes: ie. if the very same 50 MB wordforms file is specified for 10 different indexes, additional searchd RAM usage will be about 50 MB.

Dictionary file should be in a simple plain text format. Each line should contain source and destination word forms, in exactly the same encoding as specified in charset_type, separated by "greater" sign. Rules from the charset_table will be applied when the file is loaded. So basically it's as case sensitive as your other full-text indexed data, ie. typically case insensitive. Here's the file contents sample:

walks > walk
walked > walk
walking > walk

There is a bundled spelldump utility that helps you create a dictionary file in the format Sphinx can read from source .dict and .aff dictionary files in ispell or MySpell format (as bundled with OpenOffice).

Starting with version 0.9.9-rc1, you can map several source words to a single destination word. Because the work happens on tokens, not the source text, differences in whitespace and markup are ignored.

core 2 duo > c2d
e6600 > c2d
core 2duo > c2d

Notice however that the destination wordforms are still always interpreted as a single keyword! Having a mapping like "St John > Saint John" will result in not matching "St John" when searching for "Saint" or "John", because the destination keyword will be "Saint John" with a space character in it (and it's barely possible to input a destination keyword with a space).


wordforms = /usr/local/sphinx/data/wordforms.txt

11.2.13. exceptions

Tokenizing exceptions file. Optional, default is empty.

Exceptions allow to map one or more tokens (including tokens with characters that would normally be excluded) to a single keyword. They are similar to wordforms in that they also perform mapping, but have a number of important differences.

Short summary of the differences is as follows:

  • exceptions are case sensitive, wordforms are not;

  • exceptions can use special characters that are not in charset_table, wordforms fully obey charset_table;

  • exceptions can underperform on huge dictionaries, wordforms handle millions of entries well.

The expected file format is also plain text, with one line per exception, and the line format is as follows:

map-from-tokens => map-to-token

Example file:

AT & T => AT&T
AT&T => AT&T
Standarten   Fuehrer => standartenfuhrer
Standarten Fuhrer => standartenfuhrer
MS Windows => ms windows
Microsoft Windows => ms windows
C++ => cplusplus
c++ => cplusplus
C plus plus => cplusplus

All tokens here are case sensitive: they will not be processed by charset_table rules. Thus, with the example exceptions file above, "At&t" text will be tokenized as two keywords "at" and "t", because of lowercase letters. On the other hand, "AT&T" will match exactly and produce single "AT&T" keyword.

Note that this map-to keyword is a) always interpereted as a single word, and b) is both case and space sensitive! In our sample, "ms windows" query will not match the document with "MS Windows" text. The query will be interpreted as a query for two keywords, "ms" and "windows". And what "MS Windows" gets mapped to is a single keyword "ms windows", with a space in the middle. On the other hand, "standartenfuhrer" will retrieve documents with "Standarten Fuhrer" or "Standarten Fuehrer" contents (capitalized exactly like this), or any capitalization variant of the keyword itself, eg. "staNdarTenfUhreR". (It won't catch "standarten fuhrer", however: this text does not match any of the listed exceptions because of case sensitivity, and gets indexed as two separate keywords.)

Whitespace in the map-from tokens list matters, but its amount does not. Any amount of the whitespace in the map-form list will match any other amount of whitespace in the indexed document or query. For instance, "AT & T" map-from token will match "AT    &  T" text, whatever the amount of space in both map-from part and the indexed text. Such text will therefore be indexed as a special "AT&T" keyword, thanks to the very first entry from the sample.

Exceptions also allow to capture special characters (that are exceptions from general charset_table rules; hence the name). Assume that you generally do not want to treat '+' as a valid character, but still want to be able search for some exceptions from this rule such as 'C++'. The sample above will do just that, totally independent of what characters are in the table and what are not.

Exceptions are applied to raw incoming document and query data during indexing and searching respectively. Therefore, to pick up changes in the file it's required to reindex and restart searchd.


exceptions = /usr/local/sphinx/data/exceptions.txt

11.2.14. min_word_len

Minimum indexed word length. Optional, default is 1 (index everything).

Only those words that are not shorter than this minimum will be indexed. For instance, if min_word_len is 4, then 'the' won't be indexed, but 'they' will be.


min_word_len = 4

11.2.15. charset_type

Character set encoding type. Optional, default is 'sbcs'. Known values are 'sbcs' and 'utf-8'.

Different encodings have different methods for mapping their internal characters codes into specific byte sequences. Two most common methods in use today are single-byte encoding and UTF-8. Their corresponding charset_type values are 'sbcs' (stands for Single Byte Character Set) and 'utf-8'. The selected encoding type will be used everywhere where the index is used: when indexing the data, when parsing the query against this index, when generating snippets, etc.

Note that while 'utf-8' implies that the decoded values must be treated as Unicode codepoint numbers, there's a family of 'sbcs' encodings that may in turn treat different byte values differently, and that should be properly reflected in your charset_table settings. For example, the same byte value of 224 (0xE0 hex) maps to different Russian letters depending on whether koi-8r or windows-1251 encoding is used.


charset_type = utf-8

11.2.16. charset_table

Accepted characters table, with case folding rules. Optional, default value depends on charset_type value.

charset_table is the main workhorse of Sphinx tokenizing process, ie. the process of extracting keywords from document text or query txet. It controls what characters are accepted as valid and what are not, and how the accepted characters should be transformed (eg. should the case be removed or not).

You can think of charset_table as of a big table that has a mapping for each and every of 100K+ characters in Unicode (or as of a small 256-character table if you're using SBCS). By default, every character maps to 0, which means that it does not occur within keywords and should be treated as a separator. Once mentioned in the table, character is mapped to some other character (most frequently, either to itself or to a lowercase letter), and is treated as a valid keyword part.

The expected value format is a commas-separated list of mappings. Two simplest mappings simply declare a character as valid, and map a single character to another single character, respectively. But specifying the whole table in such form would result in bloated and barely manageable specifications. So there are several syntax shortcuts that let you map ranges of characters at once. The complete list is as follows:


Single char mapping, declares source char 'A' as allowed to occur within keywords and maps it to destination char 'a' (but does not declare 'a' as allowed).


Range mapping, declares all chars in source range as allowed and maps them to the destination range. Does not declare destination range as allowed. Also checks ranges' lengths (the lengths must be equal).


Stray char mapping, declares a character as allowed and maps it to itself. Equivalent to a->a single char mapping.


Stray range mapping, declares all characters in range as allowed and maps them to themselves. Equivalent to a..z->a..z range mapping.


Checkerboard range map. Maps every pair of chars to the second char. More formally, declares odd characters in range as allowed and maps them to the even ones; also declares even characters as allowed and maps them to themselves. For instance, A..Z/2 is equivalent to A->B, B->B, C->D, D->D, ..., Y->Z, Z->Z. This mapping shortcut is helpful for a number of Unicode blocks where uppercase and lowercase letters go in such interleaved order instead of contiguous chunks.

Control characters with codes from 0 to 31 are always treated as separators. Characters with codes 32 to 127, ie. 7-bit ASCII characters, can be used in the mappings as is. To avoid configuration file encoding issues, 8-bit ASCII characters and Unicode characters must be specified in U+xxx form, where 'xxx' is hexadecimal codepoint number. This form can also be used for 7-bit ASCII characters to encode special ones: eg. use U+20 to encode space, U+2E to encode dot, U+2C to encode comma.


# 'sbcs' defaults for English and Russian
charset_table = 0..9, A..Z->a..z, _, a..z, \
	U+A8->U+B8, U+B8, U+C0..U+DF->U+E0..U+FF, U+E0..U+FF

# 'utf-8' defaults for English and Russian
charset_table = 0..9, A..Z->a..z, _, a..z, \
	U+410..U+42F->U+430..U+44F, U+430..U+44F

11.2.17. ignore_chars

Ignored characters list. Optional, default is empty.

Useful in the cases when some characters, such as soft hyphenation mark (U+00AD), should be not just treated as separators but rather fully ignored. For example, if '-' is simply not in the charset_table, "abc-def" text will be indexed as "abc" and "def" keywords. On the contrary, if '-' is added to ignore_chars list, the same text will be indexed as a single "abcdef" keyword.

The syntax is the same as for charset_table, but it's only allowed to declare characters, and not allowed to map them. Also, the ignored characters must not be present in charset_table.


ignore_chars = U+AD

11.2.18. min_prefix_len

Minimum word prefix length to index. Optional, default is 0 (do not index prefixes).

Prefix indexing allows to implement wildcard searching by 'wordstart*' wildcards (refer to enable_star option for details on wildcard syntax). When mininum prefix length is set to a positive number, indexer will index all the possible keyword prefixes (ie. word beginnings) in addition to the keywords themselves. Too short prefixes (below the minimum allowed length) will not be indexed.

For instance, indexing a keyword "example" with min_prefix_len=3 will result in indexing "exa", "exam", "examp", "exampl" prefixes along with the word itself. Searches against such index for "exam" will match documents that contain "example" word, even if they do not contain "exam" on itself. However, indexing prefixes will make the index grow significantly (because of many more indexed keywords), and will degrade both indexing and searching times.

There's no automatic way to rank perfect word matches higher in a prefix index, but there's a number of tricks to achieve that. First, you can setup two indexes, one with prefix indexing and one without it, search through both, and use SetIndexWeights() call to combine weights. Second, you can enable star-syntax and rewrite your extended-mode queries:

# in sphinx.conf
enable_star = 1

// in query
$cl->Query ( "( keyword | keyword* ) other keywords" );


min_prefix_len = 3

11.2.19. min_infix_len

Minimum infix prefix length to index. Optional, default is 0 (do not index infixes).

Infix indexing allows to implement wildcard searching by 'start*', '*end', and '*middle*' wildcards (refer to enable_star option for details on wildcard syntax). When mininum infix length is set to a positive number, indexer will index all the possible keyword infixes (ie. substrings) in addition to the keywords themselves. Too short infixes (below the minimum allowed length) will not be indexed. For instance, indexing a keyword "test" with min_infix_len=2 will result in indexing "te", "es", "st", "tes", "est" infixes along with the word itself. Searches against such index for "es" will match documents that contain "test" word, even if they do not contain "es" on itself. However, indexing infixes will make the index grow significantly (because of many more indexed keywords), and will degrade both indexing and searching times.

There's no automatic way to rank perfect word matches higher in an infix index, but the same tricks as with prefix indexes can be applied.


min_infix_len = 3

11.2.20. prefix_fields

The list of full-text fields to limit prefix indexing to. Optional, default is empty (index all fields in prefix mode).

Because prefix indexing impacts both indexing and searching performance, it might be desired to limit it to specific full-text fields only: for instance, to provide prefix searching through URLs, but not through page contents. prefix_fields specifies what fields will be prefix-indexed; all other fields will be indexed in normal mode. The value format is a comma-separated list of field names.


prefix_fields = url, domain

11.2.21. infix_fields

The list of full-text fields to limit infix indexing to. Optional, default is empty (index all fields in infix mode).

Similar to prefix_fields, but lets you limit infix-indexing to given fields.


infix_fields = url, domain

11.2.22. enable_star

Enables star-syntax (or wildcard syntax) when searching through prefix/infix indexes. Optional, default is is 0 (do not use wildcard syntax), for compatibility with 0.9.7. Known values are 0 and 1.

This feature enables "star-syntax", or wildcard syntax, when searching through indexes which were created with prefix or infix indexing enabled. It only affects searching; so it can be changed without reindexing by simply restarting searchd.

The default value is 0, that means to disable star-syntax and treat all keywords as prefixes or infixes respectively, depending on indexing-time min_prefix_len and min_infix_len settings. The value of 1 means that star ('*') can be used at the start and/or the end of the keyword. The star will match zero or more characters.

For example, assume that the index was built with infixes and that enable_star is 1. Searching should work as follows:

  1. "abcdef" query will match only those documents that contain the exact "abcdef" word in them.

  2. "abc*" query will match those documents that contain any words starting with "abc" (including the documents which contain the exact "abc" word only);

  3. "*cde*" query will match those documents that contain any words which have "cde" characters in any part of the word (including the documents which contain the exact "cde" word only).

  4. "*def" query will match those documents that contain any words ending with "def" (including the documents that contain the exact "def" word only).


enable_star = 1

11.2.23. ngram_len

N-gram lengths for N-gram indexing. Optional, default is 0 (disable n-gram indexing). Known values are 0 and 1 (other lengths to be implemented).

N-grams provide basic CJK (Chinese, Japanese, Korean) support for unsegmented texts. The issue with CJK searching is that there could be no clear separators between the words. Ideally, the texts would be filtered through a special program called segmenter that would insert separators in proper locations. However, segmenters are slow and error prone, and it's common to index contiguous groups of N characters, or n-grams, instead.

When this feature is enabled, streams of CJK characters are indexed as N-grams. For example, if incoming text is "ABCDEF" (where A to F represent some CJK characters) and length is 1, in will be indexed as if it was "A B C D E F". (With length equal to 2, it would produce "AB BC CD DE EF"; but only 1 is supported at the moment.) Only those characters that are listed in ngram_chars table will be split this way; other ones will not be affected.

Note that if search query is segmented, ie. there are separators between individual words, then wrapping the words in quotes and using extended mode will resut in proper matches being found even if the text was not segmented. For instance, assume that the original query is BC DEF. After wrapping in quotes on the application side, it should look like "BC" "DEF" (with quotes). This query will be passed to Sphinx and internally split into 1-grams too, resulting in "B C" "D E F" query, still with quotes that are the phrase matching operator. And it will match the text even though there were no separators in the text.

Even if the search query is not segmented, Sphinx should still produce good results, thanks to phrase based ranking: it will pull closer phrase matches (which in case of N-gram CJK words can mean closer multi-character word matches) to the top.


ngram_len = 1

11.2.24. ngram_chars

N-gram characters list. Optional, default is empty.

To be used in conjunction with in ngram_len, this list defines characters, sequences of which are subject to N-gram extraction. Words comprised of other characters will not be affected by N-gram indexing feature. The value format is identical to charset_table.


ngram_chars = U+3000..U+2FA1F

11.2.25. phrase_boundary

Phrase boundary characters list. Optional, default is empty.

This list controls what characters will be treated as phrase boundaries, in order to adjust word positions and enable phrase-level search emulation through proximity search. The syntax is similar to charset_table. Mappings are not allowed and the boundary characters must not overlap with anything else.

On phrase boundary, additional word position increment (specified by phrase_boundary_step) will be added to current word position. This enables phrase-level searching through proximity queries: words in different phrases will be guaranteed to be more than phrase_boundary_step distance away from each other; so proximity search within that distance will be equivalent to phrase-level search.

Phrase boundary condition will be raised if and only if such character is followed by a separator; this is to avoid abbreviations such as S.T.A.L.K.E.R or URLs being treated as several phrases.


phrase_boundary = ., ?, !, U+2026 # horizontal ellipsis

11.2.26. phrase_boundary_step

Phrase boundary word position increment. Optional, default is 0.

On phrase boundary, current word position will be additionally incremented by this number. See phrase_boundary for details.


phrase_boundary_step = 100

11.2.27. html_strip

Whether to strip HTML markup from incoming full-text data. Optional, default is 0. Known values are 0 (disable stripping) and 1 (enable stripping).

Both HTML tags and entities and considered markup and get processed.

HTML tags are removed, their contents (i.e., everything between <P> and </P>) are left intact by default. You can choose to keep and index attributes of the tags (e.g., HREF attribute in an A tag, or ALT in an IMG one). Several well-known inline tags are completely removed, all other tags are treated as block level and replaced with whitespace. For example, 'te<B>st</B>' text will be indexed as a single keyword 'test', however, 'te<P>st</P>' will be indexed as two keywords 'te' and 'st'. Known inline tags are as follows: A, B, I, S, U, BASEFONT, BIG, EM, FONT, IMG, LABEL, SMALL, SPAN, STRIKE, STRONG, SUB, SUP, TT.

HTML entities get decoded and replaced with corresponding UTF-8 characters. Stripper supports both numeric forms (such as &#239;) and text forms (such as &oacute; or &nbsp;). All entities as specified by HTML4 standard are supported.

Stripping does not work with xmlpipe source type (it's suggested to upgrade to xmlpipe2 anyway). It should work with properly formed HTML and XHTML, but, just as most browsers, may produce unexpected results on malformed input (such as HTML with stray <'s or unclosed >'s).

Only the tags themselves, and also HTML comments, are stripped. To strip the contents of the tags too (eg. to strip embedded scripts), see html_remove_elements option. There are no restrictions on tag names; ie. everything that looks like a valid tag start, or end, or a comment will be stripped.


html_strip = 1

11.2.28. html_index_attrs

A list of markup attributes to index when stripping HTML. Optional, default is empty (do not index markup attributes).

Specifies HTML markup attributes whose contents should be retained and indexed even though other HTML markup is stripped. The format is per-tag enumeration of indexable attributes, as shown in the example below.


html_index_attrs = img=alt,title; a=title;

11.2.29. html_remove_elements

A list of HTML elements for which to strip contents along with the elements themselves. Optional, default is empty string (do not strip contents of any elements).

This feature allows to strip element contents, ie. everything that is between the opening and the closing tags. It is useful to remove embedded scripts, CSS, etc. Short tag form for empty elements (ie. <br />) is properly supported; ie. the text that follows such tag will not be removed.

The value is a comma-separated list of element (tag) names whose contents should be removed. Tag names are case insensitive.


html_remove_elements = style, script

11.2.30. local

Local index declaration in the distributed index. Multi-value, optional, default is empty.

This setting is used to declare local indexes that will be searched when given distributed index is searched. Many local indexes can be declared per each distributed index. Any local index can also be mentioned several times in different distributed indexes.

Note that by default all local indexes will be searched sequentially, utilizing only 1 CPU or core. To parallelize processing of the local parts in the distributed index, you should use dist_threads directive, see Section 11.4.29, “dist_threads”.

Before dist_threads, there also was a legacy solution to configure searchd to query itself instead of using local indexes (refer to Section 11.2.31, “agent” for the details). However, that creates redundant CPU and network load, and dist_threads is now strongly suggested instead.


local = chunk1
local = chunk2

11.2.31. agent

Remote agent declaration in the distributed index. Multi-value, optional, default is empty.

This setting is used to declare remote agents that will be searched when given distributed index is searched. The agents can be thought of as network pointers that specify host, port, and index names. In the basic case agents would correspond to remote physical machines. More formally, that is not always correct: you can point several agents to the same remote machine; or you can even point agents to the very same single instance of searchd (in order to utilize many CPUs or cores).

The value format is as follows:

agent = specification:remote-indexes-list
specification = hostname ":" port | path

Where 'hostname' is remote host name; 'port' is remote TCP port; 'path' is Unix-domain socket path and 'remote-indexes-list' is a comma-separated list of remote index names.

All agents will be searched in parallel. However, all indexes specified for a given agent will be searched sequentially in this agent. This lets you fine-tune the configuration to the hardware. For instance, if two remote indexes are stored on the same physical HDD, it's better to configure one agent with several sequentially searched indexes to avoid HDD steping. If they are stored on different HDDs, having two agents will be advantageous, because the work will be fully parallelized. The same applies to CPUs; though CPU performance impact caused by two processes stepping on each other is somewhat smaller and frequently can be ignored at all.

On machines with many CPUs and/or HDDs, agents can be pointed to the same machine to utilize all of the hardware in parallel and reduce query latency. There is no need to setup several searchd instances for that; it's legal to configure the instance to contact itself. Here's an example setup, intended for a 4-CPU machine, that will use up to 4 CPUs in parallel to process each query:

index dist
	type = distributed
	local = chunk1
	agent = localhost:9312:chunk2
	agent = localhost:9312:chunk3
	agent = localhost:9312:chunk4

Note how one of the chunks is searched locally and the same instance of searchd queries itself to launch searches through three other ones in parallel.


agent = localhost:9312:chunk2 # contact itself
agent = /var/run/searchd.s:chunk2
agent = searchbox2:9312:chunk3,chunk4 # search remote indexes

11.2.32. agent_blackhole

Remote blackhole agent declaration in the distributed index. Multi-value, optional, default is empty. Introduced in version 0.9.9-rc1.

agent_blackhole lets you fire-and-forget queries to remote agents. That is useful for debugging (or just testing) production clusters: you can setup a separate debugging/testing searchd instance, and forward the requests to this instance from your production master (aggregator) instance without interfering with production work. Master searchd will attempt to connect and query blackhole agent normally, but it will neither wait nor process any responses. Also, all network errors on blackhole agents will be ignored. The value format is completely identical to regular agent directive.


agent_blackhole = testbox:9312:testindex1,testindex2

11.2.33. agent_connect_timeout

Remote agent connection timeout, in milliseconds. Optional, default is 1000 (ie. 1 second).

When connecting to remote agents, searchd will wait at most this much time for connect() call to complete succesfully. If the timeout is reached but connect() does not complete, and retries are enabled, retry will be initiated.


agent_connect_timeout = 300

11.2.34. agent_query_timeout

Remote agent query timeout, in milliseconds. Optional, default is 3000 (ie. 3 seconds).

After connection, searchd will wait at most this much time for remote queries to complete. This timeout is fully separate from connection timeout; so the maximum possible delay caused by a remote agent equals to the sum of agent_connection_timeout and agent_query_timeout. Queries will not be retried if this timeout is reached; a warning will be produced instead.


agent_query_timeout = 10000 # our query can be long, allow up to 10 sec

11.2.35. preopen

Whether to pre-open all index files, or open them per each query. Optional, default is 0 (do not preopen).

This option tells searchd that it should pre-open all index files on startup (or rotation) and keep them open while it runs. Currently, the default mode is not to pre-open the files (this may change in the future). Preopened indexes take a few (currently 2) file descriptors per index. However, they save on per-query open() calls; and also they are invulnerable to subtle race conditions that may happen during index rotation under high load. On the other hand, when serving many indexes (100s to 1000s), it still might be desired to open the on per-query basis in order to save file descriptors.

This directive does not affect indexer in any way, it only affects searchd.


preopen = 1

11.2.36. ondisk_dict

Whether to keep the dictionary file (.spi) for this index on disk, or precache it in RAM. Optional, default is 0 (precache in RAM). Introduced in version 0.9.9-rc1.

The dictionary (.spi) can be either kept on RAM or on disk. The default is to fully cache it in RAM. That improves performance, but might cause too much RAM pressure, especially if prefixes or infixes were used. Enabling ondisk_dict results in 1 additional disk IO per keyword per query, but reduces memory footprint.

This directive does not affect indexer in any way, it only affects searchd.


ondisk_dict = 1

11.2.37. inplace_enable

Whether to enable in-place index inversion. Optional, default is 0 (use separate temporary files). Introduced in version 0.9.9-rc1.

inplace_enable greatly reduces indexing disk footprint, at a cost of slightly slower indexing (it uses around 2x less disk, but yields around 90-95% the original performance).

Indexing involves two major phases. The first phase collects, processes, and partially sorts documents by keyword, and writes the intermediate result to temporary files (.tmp*). The second phase fully sorts the documents, and creates the final index files. Thus, rebuilding a production index on the fly involves around 3x peak disk footprint: 1st copy for the intermediate temporary files, 2nd copy for newly constructed copy, and 3rd copy for the old index that will be serving production queries in the meantime. (Intermediate data is comparable in size to the final index.) That might be too much disk footprint for big data collections, and inplace_enable allows to reduce it. When enabled, it reuses the temporary files, outputs the final data back to them, and renames them on completion. However, this might require additional temporary data chunk relocation, which is where the performance impact comes from.

This directive does not affect searchd in any way, it only affects indexer.


inplace_enable = 1

11.2.38. inplace_hit_gap

In-place inversion fine-tuning option. Controls preallocated hitlist gap size. Optional, default is 0. Introduced in version 0.9.9-rc1.

This directive does not affect searchd in any way, it only affects indexer.


inplace_hit_gap = 1M

11.2.39. inplace_docinfo_gap

In-place inversion fine-tuning option. Controls preallocated docinfo gap size. Optional, default is 0. Introduced in version 0.9.9-rc1.

This directive does not affect searchd in any way, it only affects indexer.


inplace_docinfo_gap = 1M

11.2.40. inplace_reloc_factor

In-place inversion fine-tuning option. Controls relocation buffer size within indexing memory arena. Optional, default is 0.1. Introduced in version 0.9.9-rc1.

This directive does not affect searchd in any way, it only affects indexer.


inplace_reloc_factor = 0.1

11.2.41. inplace_write_factor

In-place inversion fine-tuning option. Controls in-place write buffer size within indexing memory arena. Optional, default is 0.1. Introduced in version 0.9.9-rc1.

This directive does not affect searchd in any way, it only affects indexer.


inplace_write_factor = 0.1

11.2.42. index_exact_words

Whether to index the original keywords along with the stemmed/remapped versions. Optional, default is 0 (do not index). Introduced in version 0.9.9-rc1.

When enabled, index_exact_words forces indexer to put the raw keywords in the index along with the stemmed versions. That, in turn, enables exact form operator in the query language to work. This impacts the index size and the indexing time. However, searching performance is not impacted at all.


index_exact_words = 1

11.2.43. overshort_step

Position increment on overshort (less that min_word_len) keywords. Optional, allowed values are 0 and 1, default is 1. Introduced in version 0.9.9-rc1.

This directive does not affect searchd in any way, it only affects indexer.


overshort_step = 1

11.2.44. stopword_step

Position increment on stopwords. Optional, allowed values are 0 and 1, default is 1. Introduced in version 0.9.9-rc1.

This directive does not affect searchd in any way, it only affects indexer.


stopword_step = 1

11.2.45. hitless_words

Hitless words list. Optional, allowed values are 'all', or a list file name. Introduced in version 1.10-beta.

By default, Sphinx full-text index stores not only a list of matching documents for every given keyword, but also a list of its in-document positions (aka hitlist). Hitlists enables phrase, proximity, strict order and other advanced types of searching, as well as phrase proximity ranking. However, hitlists for specific frequent keywords (that can not be stopped for some reason despite being frequent) can get huge and thus slow to process while querying. Also, in some cases we might only care about boolean keyword matching, and never need position-based searching operators (such as phrase matching) nor phrase ranking.

hitless_words lets you create indexes that either do not have positional information (hitlists) at all, or skip it for specific keywords.

Hitless index will generally use less space than the respective regular index (about 1.5x can be expected). Both indexing and searching should be faster, at a cost of missing positional query and ranking support. When searching, positional queries (eg. phrase queries) will be automatically converted to respective non-positional (document-level) or combined queries. For instance, if keywords "hello" and "world" are hitless, "hello world" phrase query will be converted to (hello & world) bag-of-words query, matching all documents that mention either of the keywords but not necessarily the exact phrase. And if, in addition, keywords "simon" and "says" are not hitless, "simon says hello world" will be converted to ("simon says" & hello & world) query, matching all documents that contain "hello" and "world" anywhere in the document, and also "simon says" as an exact phrase.


hitless_words = all

11.2.46. expand_keywords

Expand keywords with exact forms and/or stars when possible. Optional, default is 0 (do not expand keywords). Introduced in version 1.10-beta.

Queries against indexes with expand_keywords feature enabled are internally expanded as follows. If the index was built with prefix or infix indexing enabled, every keyword gets internally replaced with a disjunction of keyword itself and a respective prefix or infix (keyword with stars). If the index was built with both stemming and index_exact_words enabled, exact form is also added. Here's an example that shows how internal expansion works when all of the above (infixes, stemming, and exact words) are combined:

running -> ( running | *running* | =running )

Expanded queries take naturally longer to complete, but can possibly improve the search quality, as the documents with exact form matches should be ranked generally higher than documents with stemmed or infix matches.

Note that the existing query syntax does not allowe to emulate this kind of expansion, because internal expansion works on keyword level and expands keywords within phrase or quorum operators too (which is not possible through the query syntax).

This directive does not affect indexer in any way, it only affects searchd.


expand_keywords = 1

11.2.47. blend_chars

Blended characters list. Optional, default is empty. Introduced in version 1.10-beta.

Blended characters are indexed both as separators and valid characters. For instance, assume that & is configured as blended and AT&T occurs in an indexed document. Three different keywords will get indexed, namely "at&t", treating blended characters as valid, plus "at" and "t", treating them as separators.

Positions for tokens obtained by replacing blended characters with whitespace are assigned as usual, so regular keywords will be indexed just as if there was no blend_chars specified at all. An additional token that mixes blended and non-blended characters will be put at the starting position. For instance, if the field contents are "AT&T company" occurs in the very beginning of the text field, "at" will be given position 1, "t" position 2, "company" positin 3, and "AT&T" will also be given position 1 ("blending" with the opening regular keyword). Thus, querying for either AT&T or just AT will match that document, and querying for "AT T" as a phrase also match it. Last but not least, phrase query for "AT&T company" will also match it, despite the position

Blended characters can overlap with special characters used in query syntax (think of T-Mobile or @twitter). Where possible, query parser will automatically handle blended character as blended. For instance, "hello @twitter" within quotes (a phrase operator) would handle @-sign as blended, because @-syntax for field operator is not allowed within phrases. Otherwise, the character would be handled as an operator. So you might want to escape the keywords.

Starting with version 2.0.1-beta, blended characters can be remapped, so that multiple different blended characters could be normalized into just one base form. This is useful when indexing multiple alternative Unicode codepoints with equivalent glyphs.


blend_chars = +, &, U+23
blend_chars = +, &->+ # 2.0.1 and above

11.2.48. blend_mode

Blended tokens indexing mode. Optional, default is trim_none. Introduced in version 2.0.1-beta.

By default, tokens that mix blended and non-blended characters get indexed in there entirety. For instance, when both at-sign and an exclamation are in blend_chars, "@dude!" will get result in two tokens indexed: "@dude!" (with all the blended characters) and "dude" (without any). Therefore "@dude" query will not match it.

blend_mode directive adds flexibility to this indexing behavior. It takes a comma-separated list of options.

blend_mode = option [, option [, ...]]
option = trim_none | trim_head | trim_tail | trim_both | skip_pure

Options specify token indexing variants. If multiple options are specified, multiple variants of the same token will be indexed. Regular keywords (resulting from that token by replacing blended with whitespace) are always be indexed.


Index the entire token.


Trim heading blended characters, and index the resulting token.


Trim trailing blended characters, and index the resulting token.


Trim both heading and trailing blended characters, and index the resulting token.


Do not index the token if it's purely blended, that is, consists of blended characters only.

Returning to the "@dude!" example above, setting blend_mode = trim_head, trim_tail will result in two tokens being indexed, "@dude" and "dude!". In this particular example, trim_both would have no effect, because trimming both blended characters results in "dude" which is already indexed as a regular keyword. Indexing "@U.S.A." with trim_both (and assuming that dot is blended two) would result in "U.S.A" being indexed. Last but not least, skip_pure enables you to fully ignore sequences of blended characters only. For example, "one @@@ two" would be indexed exactly as "one two", and match that as a phrase. That is not the case by default because a fully blended token gets indexed and offsets the second keyword position.

Default behavior is to index the entire token, equivalent to blend_mode = trim_none.


blend_mode = trim_tail, skip_pure

11.2.49. rt_mem_limit

RAM chunk size limit. Optional, default is empty. Introduced in version 1.10-beta.

RT index keeps some data in memory (so-called RAM chunk) and also maintains a number of on-disk indexes (so-called disk chunks). This directive lets you control the RAM chunk size. Once there's too much data to keep in RAM, RT index will flush it to disk, activate a newly created disk chunk, and reset the RAM chunk.

The limit is pretty strict; RT index should never allocate more memory than it's limited to. The memory is not preallocated either, hence, specifying 512 MB limit and only inserting 3 MB of data should result in allocating 3 MB, not 512 MB.


rt_mem_limit = 512M

11.2.50. rt_field

Full-text field declaration. Multi-value, mandatory Introduced in version 1.10-beta.

Full-text fields to be indexed are declared using rt_field directive. The names must be unique. The order is preserved; and so field values in INSERT statements without an explicit list of inserted columns will have to be in the same order as configured.


rt_field = author
rt_field = title
rt_field = content

11.2.51. rt_attr_uint

Unsigned integer attribute declaration. Multi-value (an arbitrary number of attributes is allowed), optional. Declares an unsigned 32-bit attribute. Introduced in version 1.10-beta.


rt_attr_uint = gid

11.2.52. rt_attr_bigint

BIGINT attribute declaration. Multi-value (an arbitrary number of attributes is allowed), optional. Declares a signed 64-bit attribute. Introduced in version 1.10-beta.


rt_attr_bigint = guid

11.2.53. rt_attr_float

Floating point attribute declaration. Multi-value (an arbitrary number of attributes is allowed), optional. Declares a single precision, 32-bit IEEE 754 format float attribute. Introduced in version 1.10-beta.


rt_attr_float = gpa

11.2.54. rt_attr_multi

Multi-valued attribute (MVA) declaration. Declares the UNSIGNED INTEGER (unsigned 32-bit) MVA attribute. Multi-value (ie. there may be more than one such attribute declared), optional. Applies to RT indexes only.


rt_attr_multi = my_tags

11.2.55. rt_attr_multi_64

Multi-valued attribute (MVA) declaration. Declares the BIGINT (signed 64-bit) MVA attribute. Multi-value (ie. there may be more than one such attribute declared), optional. Applies to RT indexes only.


rt_attr_multi_64 = my_wide_tags

11.2.56. rt_attr_timestamp

Timestamp attribute declaration. Multi-value (an arbitrary number of attributes is allowed), optional. Introduced in version 1.10-beta.


rt_attr_timestamp = date_added

11.2.57. rt_attr_string

String attribute declaration. Multi-value (an arbitrary number of attributes is allowed), optional. Introduced in version 1.10-beta.


rt_attr_string = author

11.3. indexer program configuration options

11.3.1. mem_limit

Indexing RAM usage limit. Optional, default is 32M.

Enforced memory usage limit that the indexer will not go above. Can be specified in bytes, or kilobytes (using K postfix), or megabytes (using M postfix); see the example. This limit will be automatically raised if set to extremely low value causing I/O buffers to be less than 8 KB; the exact lower bound for that depends on the indexed data size. If the buffers are less than 256 KB, a warning will be produced.

Maximum possible limit is 2047M. Too low values can hurt indexing speed, but 256M to 1024M should be enough for most if not all datasets. Setting this value too high can cause SQL server timeouts. During the document collection phase, there will be periods when the memory buffer is partially sorted and no communication with the database is performed; and the database server can timeout. You can resolve that either by raising timeouts on SQL server side or by lowering mem_limit.


mem_limit = 256M
# mem_limit = 262144K # same, but in KB
# mem_limit = 268435456 # same, but in bytes

11.3.2. max_iops

Maximum I/O operations per second, for I/O throttling. Optional, default is 0 (unlimited).

I/O throttling related option. It limits maximum count of I/O operations (reads or writes) per any given second. A value of 0 means that no limit is imposed.

indexer can cause bursts of intensive disk I/O during indexing, and it might desired to limit its disk activity (and keep something for other programs running on the same machine, such as searchd). I/O throttling helps to do that. It works by enforcing a minimum guaranteed delay between subsequent disk I/O operations performed by indexer. Modern SATA HDDs are able to perform up to 70-100+ I/O operations per second (that's mostly limited by disk heads seek time). Limiting indexing I/O to a fraction of that can help reduce search performance dedgradation caused by indexing.


max_iops = 40

11.3.3. max_iosize

Maximum allowed I/O operation size, in bytes, for I/O throttling. Optional, default is 0 (unlimited).

I/O throttling related option. It limits maximum file I/O operation (read or write) size for all operations performed by indexer. A value of 0 means that no limit is imposed. Reads or writes that are bigger than the limit will be split in several smaller operations, and counted as several operation by max_iops setting. At the time of this writing, all I/O calls should be under 256 KB (default internal buffer size) anyway, so max_iosize values higher than 256 KB must not affect anything.


max_iosize = 1048576

11.3.4. max_xmlpipe2_field

Maximum allowed field size for XMLpipe2 source type, bytes. Optional, default is 2 MB.


max_xmlpipe2_field = 8M

11.3.5. write_buffer

Write buffer size, bytes. Optional, default is 1 MB.

Write buffers are used to write both temporary and final index files when indexing. Larger buffers reduce the number of required disk writes. Memory for the buffers is allocated in addition to mem_limit. Note that several (currently up to 4) buffers for different files will be allocated, proportionally increasing the RAM usage.


write_buffer = 4M

11.3.6. max_file_field_buffer

Maximum file field adaptive buffer size, bytes. Optional, default is 8 MB, minimum is 1 MB.

File field buffer is used to load files referred to from sql_file_field columns. This buffer is adaptive, starting at 1 MB at first allocation, and growing in 2x steps until either file contents can be loaded, or maximum buffer size, specified by max_file_field_buffer directive, is reached.

Thus, if there are no file fields are specified, no buffer is allocated at all. If all files loaded during indexing are under (for example) 2 MB in size, but max_file_field_buffer value is 128 MB, peak buffer usage would still be only 2 MB. However, files over 128 MB would be entirely skipped.


max_file_field_buffer = 128M

11.3.7. on_file_field_error

How to handle IO errors in file fields. Optional, default is ignore_field. Introduced in version 2.0.2-beta.

When there is a problem indexing a file referenced by a file field (Section 11.1.29, “sql_file_field”), indexer can either index the document, assuming empty content in this particular field, or skip the document, or fail indexing entirely. on_file_field_error directive controls that behavior. The values it takes are:

  • ignore_field, index the current document without field;

  • skip_document, skip the current document but continue indexing;

  • fail_index, fail indexing with an error message.

The problems that can arise are: open error, size error (file too big), and data read error. Warning messages on any problem will be given at all times, irregardless of the phase and the on_file_field_error setting.

Note that with on_file_field_error = skip_document documents will only be ignored if problems are detected during an early check phase, and not during the actual file parsing phase. indexer will open every referenced file and check its size before doing any work, and then open it again when doing actual parsing work. So in case a file goes away between these two open attempts, the document will still be indexed.


on_file_field_errors = skip_document

11.4. searchd program configuration options

11.4.1. listen

This setting lets you specify IP address and port, or Unix-domain socket path, that searchd will listen on. Introduced in version 0.9.9-rc1.

The informal grammar for listen setting is:

listen = ( address ":" port | port | path ) [ ":" protocol ]

I.e. you can specify either an IP address (or hostname) and port number, or just a port number, or Unix socket path. If you specify port number but not the address, searchd will listen on all network interfaces. Unix path is identified by a leading slash.

Starting with version 0.9.9-rc2, you can also specify a protocol handler (listener) to be used for connections on this socket. Supported protocol values are 'sphinx' (Sphinx 0.9.x API protocol) and 'mysql41' (MySQL protocol used since 4.1 upto at least 5.1). More details on MySQL protocol support can be found in Section 5.10, “MySQL protocol support and SphinxQL” section.


listen = localhost
listen = localhost:5000
listen =
listen = /var/run/sphinx.s
listen = 9312
listen = localhost:9306:mysql41

There can be multiple listen directives, searchd will listen for client connections on all specified ports and sockets. If no listen directives are found then the server will listen on all available interfaces using the default SphinxAPI port 9312. Starting with 1.10-beta, it will also listen on default SphinxQL port 9306. Both port numbers are assigned by IANA (see http://www.iana.org/assignments/port-numbers for details) and should therefore be available.

Unix-domain sockets are not supported on Windows.

11.4.2. address

Interface IP address to bind on. Optional, default is (ie. listen on all interfaces). DEPRECATED, use listen instead.

address setting lets you specify which network interface searchd will bind to, listen on, and accept incoming network connections on. The default value is which means to listen on all interfaces. At the time, you can not specify multiple interfaces.


address =

11.4.3. port

searchd TCP port number. DEPRECATED, use listen instead. Used to be mandatory. Default port number is 9312.


port = 9312

11.4.4. log

Log file name. Optional, default is 'searchd.log'. All searchd run time events will be logged in this file.

Also you can use the 'syslog' as the file name. In this case the events will be sent to syslog daemon. To use the syslog option the sphinx must be configured '--with-syslog' on building.


log = /var/log/searchd.log

11.4.5. query_log

Query log file name. Optional, default is empty (do not log queries). All search queries will be logged in this file. The format is described in Section 5.9, “searchd query log formats”.

In case of 'plain' format, you can use the 'syslog' as the path to the log file. In this case all search queries will be sent to syslog daemon with LOG_INFO priority, prefixed with '[query]' instead of timestamp. To use the syslog option the sphinx must be configured '--with-syslog' on building.


query_log = /var/log/query.log

11.4.6. query_log_format

Query log format. Optional, allowed values are 'plain' and 'sphinxql', default is 'plain'. Introduced in version 2.0.1-beta.

Starting with version 2.0.1-beta, two different log formats are supported. The default one logs queries in a custom text format. The new one logs valid SphinxQL statements. This directive allows to switch between the two formats on search daemon startup. The log format can also be altered on the fly, using SET GLOBAL query_log_format=sphinxql syntax. Refer to Section 5.9, “searchd query log formats” for more discussion and format details.


query_log_format = sphinxql

11.4.7. read_timeout

Network client request read timeout, in seconds. Optional, default is 5 seconds. searchd will forcibly close the client connections which fail to send a query within this timeout.


read_timeout = 1

11.4.8. client_timeout

Maximum time to wait between requests (in seconds) when using persistent connections. Optional, default is five minutes.


client_timeout = 3600

11.4.9. max_children

Maximum amount of children to fork (or in other words, concurrent searches to run in parallel). Optional, default is 0 (unlimited).

Useful to control server load. There will be no more than this much concurrent searches running, at all times. When the limit is reached, additional incoming clients are dismissed with temporarily failure (SEARCHD_RETRY) status code and a message stating that the server is maxed out.


max_children = 10

11.4.10. pid_file

searchd process ID file name. Mandatory.

PID file will be re-created (and locked) on startup. It will contain head daemon process ID while the daemon is running, and it will be unlinked on daemon shutdown. It's mandatory because Sphinx uses it internally for a number of things: to check whether there already is a running instance of searchd; to stop searchd; to notify it that it should rotate the indexes. Can also be used for different external automation scripts.


pid_file = /var/run/searchd.pid

11.4.11. max_matches

Maximum amount of matches that the daemon keeps in RAM for each index and can return to the client. Optional, default is 1000.

Introduced in order to control and limit RAM usage, max_matches setting defines how much matches will be kept in RAM while searching each index. Every match found will still be processed; but only best N of them will be kept in memory and return to the client in the end. Assume that the index contains 2,000,000 matches for the query. You rarely (if ever) need to retrieve all of them. Rather, you need to scan all of them, but only choose "best" at most, say, 500 by some criteria (ie. sorted by relevance, or price, or anything else), and display those 500 matches to the end user in pages of 20 to 100 matches. And tracking only the best 500 matches is much more RAM and CPU efficient than keeping all 2,000,000 matches, sorting them, and then discarding everything but the first 20 needed to display the search results page. max_matches controls N in that "best N" amount.

This parameter noticeably affects per-query RAM and CPU usage. Values of 1,000 to 10,000 are generally fine, but higher limits must be used with care. Recklessly raising max_matches to 1,000,000 means that searchd will have to allocate and initialize 1-million-entry matches buffer for every query. That will obviously increase per-query RAM usage, and in some cases can also noticeably impact performance.

CAVEAT EMPTOR! Note that there also is another place where this limit is enforced. max_matches can be decreased on the fly through the corresponding API call, and the default value in the API is also set to 1,000. So in order to retrieve more than 1,000 matches to your application, you will have to change the configuration file, restart searchd, and set proper limit in SetLimits() call. Also note that you can not set the value in the API higher than the value in the .conf file. This is prohibited in order to have some protection against malicious and/or malformed requests.


max_matches = 10000

11.4.12. seamless_rotate

Prevents searchd stalls while rotating indexes with huge amounts of data to precache. Optional, default is 1 (enable seamless rotation).

Indexes may contain some data that needs to be precached in RAM. At the moment, .spa, .spi and .spm files are fully precached (they contain attribute data, MVA data, and keyword index, respectively.) Without seamless rotate, rotating an index tries to use as little RAM as possible and works as follows:

  1. new queries are temporarly rejected (with "retry" error code);

  2. searchd waits for all currently running queries to finish;

  3. old index is deallocated and its files are renamed;

  4. new index files are renamed and required RAM is allocated;

  5. new index attribute and dictionary data is preloaded to RAM;

  6. searchd resumes serving queries from new index.

However, if there's a lot of attribute or dictionary data, then preloading step could take noticeble time - up to several minutes in case of preloading 1-5+ GB files.

With seamless rotate enabled, rotation works as follows:

  1. new index RAM storage is allocated;

  2. new index attribute and dictionary data is asynchronously preloaded to RAM;

  3. on success, old index is deallocated and both indexes' files are renamed;

  4. on failure, new index is deallocated;

  5. at any given moment, queries are served either from old or new index copy.

Seamless rotate comes at the cost of higher peak memory usage during the rotation (because both old and new copies of .spa/.spi/.spm data need to be in RAM while preloading new copy). Average usage stays the same.


seamless_rotate = 1

11.4.13. preopen_indexes

Whether to forcibly preopen all indexes on startup. Optional, default is 1 (preopen everything).

Starting with 2.0.1-beta, the default value for this option is now 1 (foribly preopen all indexes). In prior versions, it used to be 0 (use per-index settings).

When set to 1, this directive overrides and enforces preopen on all indexes. They will be preopened, no matter what is the per-index preopen setting. When set to 0, per-index settings can take effect. (And they default to 0.)

Pre-opened indexes avoid races between search queries and rotations that can cause queries to fail occasionally. They also make searchd use more file handles. In most scenarios it's therefore preferred and recommended to preopen indexes.


preopen_indexes = 1

11.4.14. unlink_old

Whether to unlink .old index copies on succesful rotation. Optional, default is 1 (do unlink).


unlink_old = 0

11.4.15. attr_flush_period

When calling UpdateAttributes() to update document attributes in real-time, changes are first written to the in-memory copy of attributes (docinfo must be set to extern). Then, once searchd shuts down normally (via SIGTERM being sent), the changes are written to disk. Introduced in version 0.9.9-rc1.

Starting with 0.9.9-rc1, it is possible to tell searchd to periodically write these changes back to disk, to avoid them being lost. The time between those intervals is set with attr_flush_period, in seconds.

It defaults to 0, which disables the periodic flushing, but flushing will still occur at normal shut-down.


attr_flush_period = 900 # persist updates to disk every 15 minutes

11.4.16. ondisk_dict_default

Instance-wide defaults for ondisk_dict directive. Optional, default it 0 (precache dictionaries in RAM). Introduced in version 0.9.9-rc1.

This directive lets you specify the default value of ondisk_dict for all the indexes served by this copy of searchd. Per-index directive take precedence, and will overwrite this instance-wide default value, allowing for fine-grain control.


ondisk_dict_default = 1 # keep all dictionaries on disk

11.4.17. max_packet_size

Maximum allowed network packet size. Limits both query packets from clients, and response packets from remote agents in distributed environment. Only used for internal sanity checks, does not directly affect RAM use or performance. Optional, default is 8M. Introduced in version 0.9.9-rc1.


max_packet_size = 32M

11.4.18. mva_updates_pool

Shared pool size for in-memory MVA updates storage. Optional, default size is 1M. Introduced in version 0.9.9-rc1.

This setting controls the size of the shared storage pool for updated MVA values. Specifying 0 for the size disable MVA updates at all. Once the pool size limit is hit, MVA update attempts will result in an error. However, updates on regular (scalar) attributes will still work. Due to internal technical difficulties, currently it is not possible to store (flush) any updates on indexes where MVA were updated; though this might be implemented in the future. In the meantime, MVA updates are intended to be used as a measure to quickly catchup with latest changes in the database until the next index rebuild; not as a persistent storage mechanism.


mva_updates_pool = 16M

11.4.19. crash_log_path

Deprecated debugging setting, path (formally prefix) for crash log files. Introduced in version 0.9.9-rc1. Deprecated in version 2.0.1-beta, as crash debugging information now gets logged into searchd.log in text form, and separate binary crash logs are no longer needed.

11.4.20. max_filters

Maximum allowed per-query filter count. Only used for internal sanity checks, does not directly affect RAM use or performance. Optional, default is 256. Introduced in version 0.9.9-rc1.


max_filters = 1024

11.4.21. max_filter_values

Maximum allowed per-filter values count. Only used for internal sanity checks, does not directly affect RAM use or performance. Optional, default is 4096. Introduced in version 0.9.9-rc1.


max_filter_values = 16384

11.4.22. listen_backlog

TCP listen backlog. Optional, default is 5.

Windows builds currently (as of 0.9.9) can only process the requests one by one. Concurrent requests will be enqueued by the TCP stack on OS level, and requests that can not be enqueued will immediately fail with "connection refused" message. listen_backlog directive controls the length of the connection queue. Non-Windows builds should work fine with the default value.