Filters

WHERE

WHERE is an SQL clause which works for both fulltext matching and additional filtering. The following operators are available:

MATCH('query') is supported and maps to fulltext query.

{col_name | expr_alias} [NOT] IN @uservar condition syntax is supported. Refer to SET syntax for a description of global user variables.

HTTP JSON

If you prefer HTTP JSON interface you can also do filtering. It looks more complex that in SQL, but can be recommended for the cases when you need to prepare a query in a programmatic manner, for example as a result of a form in your application filled out by the user.

Here's an example of several filters in a bool query.

This is a fulltext query that matches all the documents containing product in any field. These documents must have a price greater or equal than 500 (gte) and less or equal than 1000 (lte). All of these documents must not have a revision less than 15 (lt).

‹›
  • JSON
JSON
📋
POST /search
{
  "index": "test1",
  "query": {
    "bool": {
      "must": [
        { "match" : { "_all" : "product" } },
        { "range": { "price": { "gte": 500, "lte": 1000 } } }
      ],
      "must_not": {
        "range": { "revision": { "lt": 15 } }
      }
    }
  }
}

bool query

bool query matches documents matching boolean combinations of other queries and/or filters. Queries and filters must be specified in must, should or must_not sections and can be nested.

‹›
  • JSON
JSON
📋
POST /search
{
  "index":"test1",
  "query": {
    "bool": {
      "must": [
        { "match": {"_all":"keyword"} },
        { "range": { "revision": { "gte": 14 } } }
      ]
    }
  }
}

must

Queries and filters specified in the must section must match the documents. If several fulltext queries or filters are specified, all of them. This is the equivalent of AND queries in SQL. Note, if you want to match against an array (multi-value attribute) you can specify the attribute multiple times and if only all the queried values are found in the array the result will be positive, e.g.:

"must": [
  {"equals" : { "product_codes": 5 }},
  {"equals" : { "product_codes": 6 }}
]

Note also, it may be better in terms of performance to use:

  {"in" : { "all(product_codes)": [5,6] }}

(see details below).

should

Queries and filters specified in the should section should match the documents. If some queries are specified in must or must_not, should queries are ignored. On the other hand, if there are no queries other than should, then at least one of these queries must match a document for it to match the bool query. This is the equivalent of OR queries. Note, if you want to match against an array (multi-value attribute) you can specify the attribute multiple times, e.g.:

"should": [
  {"equals" : { "product_codes": 7 }},
  {"equals" : { "product_codes": 8 }}
]

Note also, it may be better in terms of performance to use:

  {"in" : { "any(product_codes)": [7,8] }}

(see details below).

must_not

Queries and filters specified in the must_not section must not match the documents. If several queries are specified under must_not, the document matches if none of them match.

‹›
  • JSON
JSON
📋
POST /search
{
  "index":"t",
  "query": {
    "bool": {
      "should": [
        {
          "equals": {
            "b": 1
          }
        },
        {
          "equals": {
            "b": 3
          }
        }
      ],
      "must": [
        {
          "equals": {
            "a": 1
          }
        }
      ],
      "must_not": {
        "equals": {
          "b": 2
        }
      }
    }
  }
}

Nested bool query

A bool query can be nested inside another bool so you can make more complex queries. To make a nested boolean query just use another bool instead of must, should or must_not. Here is how this query:

a = 2 and (a = 10 or b = 0)

should be presented in JSON.

‹›
  • JSON
JSON
📋

a = 2 and (a = 10 or b = 0)

POST /search
{
  "index":"t",
  "query": {
    "bool": {
      "must": [
        {
          "equals": {
            "a": 2
          }
        },
        {
          "bool": {
            "should": [
              {
                "equals": {
                  "a": 10
                }
              },
              {
                "equals": {
                  "b": 0
                }
              }
            ]
          }
        }
      ]
    }
  }
}

More complex query:

(a = 1 and b = 1) or (a = 10 and b = 2) or (b = 0)
‹›
  • JSON
JSON
📋

(a = 1 and b = 1) or (a = 10 and b = 2) or (b = 0)

POST /search
{
  "index":"t",
  "query": {
    "bool": {
      "should": [
        {
          "bool": {
            "must": [
              {
                "equals": {
                  "a": 1
                }
              },
              {
                "equals": {
                  "b": 1
                }
              }
            ]
          }
        },
        {
          "bool": {
            "must": [
              {
                "equals": {
                  "a": 10
                }
              },
              {
                "equals": {
                  "b": 2
                }
              }
            ]
          }
        },
        {
          "bool": {
            "must": [
              {
                "equals": {
                  "b": 0
                }
              }
            ]
          }

        }
      ]
    }
  }
}

Queries in SQL format

Queries in SQL format (query_string) can also be used in bool queries.

‹›
  • JSON
JSON
📋
POST /search
{
  "index": "test1",
  "query": {
    "bool": {
      "must": [
        { "query_string" : "product" },
        { "query_string" : "good" }
      ]
    }
  }
}

Various filters

Equality filters

Equality filters are the simplest filters that work with integer, float and string attributes.

‹›
  • JSON
JSON
📋
POST /search
{
  "index":"test1",
  "query": {
    "equals": { "price": 500 }
  }
}

Filter equals can be applied to a multi-value attribute and you can use:

  • any() which will be positive if the attribute has at least one value which equals to the queried value;
  • all() which will be positive if the attribute has a single value and it equals to the queried value
‹›
  • JSON
JSON
📋
POST /search
{
  "index":"test1",
  "query": {
    "equals": { "any(price)": 100 }
  }
}

Set filters

Set filters check if attribute value is equal to any of the values in the specified set.

Set filters support integer, string and multi-value attributes.

‹›
  • JSON
JSON
📋
POST /search
{
  "index":"test1",
  "query": {
    "in": {
      "price": [1,10,100]
    }
  }
}

When applied to a multi-value attribute you can use:

  • any() (equivalent to no function) which will be positive if there's at least one match between the attribute values and the queried values;
  • all() which will be positive if all the attribute values are in the queried set
‹›
  • JSON
JSON
📋
POST /search
{
  "index":"test1",
  "query": {
    "in": {
      "all(price)": [1,10]
    }
  }
}

Range filters

Range filters match documents that have attribute values within a specified range.

Range filters support the following properties:

  • gte: greater than or equal to
  • gt: greater than
  • lte: less than or equal to
  • lt: less than
‹›
  • JSON
JSON
📋
POST /search
{
  "index":"test1",
  "query": {
    "range": {
      "price": {
        "gte": 500,
        "lte": 1000
      }
    }
  }
}

Geo distance filters

geo_distance filters are used to filter the documents that are within a specific distance from a geo location.

location_anchor

Specifies the pin location, in degrees. Distances are calculated from this point.

location_source

Specifies the attributes that contain latitude and longitude.

distance_type

Specifies distance calculation function. Can be either adaptive or haversine. adaptive is faster and more precise, for more details see GEODIST(). Optional, defaults to adaptive.

distance

Specifies the maximum distance from the pin locations. All documents within this distance match. The distance can be specified in various units. If no unit is specified, the distance is assumed to be in meters. Here is a list of supported distance units:

  • Meter: m or meters
  • Kilometer: km or kilometers
  • Centimeter: cm or centimeters
  • Millimeter: mm or millimeters
  • Mile: mi or miles
  • Yard: yd or yards
  • Feet: ft or feet
  • Inch: in or inch
  • Nautical mile: NM, nmi or nauticalmiles

location_anchor and location_source properties accept the following latitude/longitude formats:

  • an object with lat and lon keys: { "lat": "attr_lat", "lon": "attr_lon" }
  • a string of the following structure: "attr_lat, attr_lon"
  • an array with the latitude and longitude in the following order: [attr_lon, attr_lat]

Latitude and longitude are specified in degrees.

‹›
  • Basic example
  • Advanced example
📋
POST /search
{
  "index":"test",
  "query": {
    "geo_distance": {
      "location_anchor": {"lat":49, "lon":15},
      "location_source": {"attr_lat, attr_lon"},
      "distance_type": "adaptive",
      "distance":"100 km"
    }
  }
}

Expressions in search

Manticore lets you use arbitrary arithmetic expressions both via SQL and HTTP, involving attribute values, internal attributes (document ID and relevance weight), arithmetic operations, a number of built-in functions, and user-defined functions. Here’s the complete reference list for quick access.

Arithmetic operators

+, -, *, /, %, DIV, MOD

The standard arithmetic operators. Arithmetic calculations involving those can be performed in three different modes:

  1. using single-precision, 32-bit IEEE 754 floating point values (the default),
  2. using signed 32-bit integers
  3. 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 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. BETWEEN and IN operators in case of multi-value attribute return true if at least one value matches the condition(same as ANY()). IN doesn't support JSON attributes. IS (NOT) NULL is supported only for JSON attributes.

Boolean operators

AND, OR, NOT

Boolean operators (AND, OR, NOT) 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.

Functions:

Expressions in HTTP JSON

In HTTP JSON interface expressions are supported via script_fields and expressions

script_fields

{
    "index": "test",
    "query": { 
        "match_all": {} 
    }, "script_fields": {
        "add_all": { 
            "script": { 
                "inline": "( gid * 10 ) | crc32(title)" 
            } 
        },
        "title_len": { 
            "script": { 
                "inline": "crc32(title)" 
            } 
        }
    }
}

In this example two expressions are created: add_all and title_len. First expression calculates ( gid * 10 ) | crc32(title) and stores the result in the add_all attribute. Second expression calculates crc32(title) and stores the result in the title_len attribute.

Only inline expressions are supported for now. The value of inline property (the expression to compute) has the same syntax as SQL expressions.

The expression name can be used in filtering or sorting.

‹›
  • script_fields
script_fields
📋
{
    "index":"movies_rt",
    "script_fields":{
        "cond1":{
            "script":{
                "inline":"actor_2_facebook_likes =296 OR movie_facebook_likes =37000"
            }
        },
        "cond2":{
            "script":{
                "inline":"IF (IN (content_rating,'TV-PG','PG'),2, IF(IN(content_rating,'TV-14','PG-13'),1,0))"
            }
        }
    },
    "limit":10,
    "sort":[
        {
            "cond2":"desc"
        },
        {
            "actor_1_name":"asc"
        },
        {
            "actor_2_name":"desc"
        }
    ],
    "profile":true,
    "query":{
        "bool":{
            "must":[
                {
                    "match":{
                        "*":"star"
                    }
                },
                {
                    "equals":{
                        "cond1":1
                    }
                }
            ],
            "must_not":[
                {
                    "equals":{
                        "content_rating":"R"
                    }
                }
            ]
        }
    }
}

The expression values are by default included in the _source array of the result set. If the source is selective (see Source selection) the expressions name can be added to the _source parameter in the request.

expressions

expressions is an alternative to script_fields with a simpler syntax. Example request adds two expressions and stores the results into add_all and title_len attributes.

‹›
  • expressions
expressions
📋
{
  "index": "test",
  "query": { "match_all": {} },
  "expressions":
  {
      "add_all": "( gid * 10 ) | crc32(title)",
      "title_len": "crc32(title)"
  }
}

Search options

SQL SELECT clause and HTTP /search endpoint support a number of options that can be used to fine-tune search behaviour.

OPTION

General syntax

SQL:

SELECT ... [OPTION <optionname>=<value> [ , ... ]] [/*+ [NO_][ColumnarScan|DocidIndex|SecondaryIndex(<attribute>[,...])]] /*]

HTTP:

POST /search
{   
    "index" : "index_name",
    "options":   
    {
        "optionname": "value",
        "optionname2": <value2>
    }
}
‹›
  • SQL
  • JSON
📋
SELECT * FROM test WHERE MATCH('@title hello @body world')
OPTION ranker=bm25, max_matches=3000,
field_weights=(title=10, body=3), agent_query_timeout=10000
‹›
Response
+------+-------+-------+
| id   | title | body  |
+------+-------+-------+
|    1 | hello | world |
+------+-------+-------+
1 row in set (0.00 sec)

Supported options are:

accurate_aggregation

Integer. Enables or disables guaranteed aggregate accuracy when running groupby queries in multiple threads. Default is 0.

When running a groupby query, it can be run in parallel on plain index on several pseudo shards (if pseudo_sharding is on). A similar approach works on RT indexes. Each shard/chunk executes the query, but the number of groups is limited by max_matches. If the result sets from different shards/chunks have different groups, the group counts and aggregates may be inaccurate. Note that Manticore tries to increase max_matches up to max_matches_increase_threshold based on the number of unique values of the groupby attribute (retrieved from secondary indexes). If it succeeds, there will be no loss in accuracy.

However, if the number of unique values of the groupby attribute is high, further increasing max_matches may not be a good strategy because it can lead to a loss in performance and higher memory usage. Setting accurate_aggregation to 1 forces groupby searches to run in a single thread, which fixes the accuracy issue. Note that running in a single thread is only enforced when max_matches cannot be set high enough; otherwise, searches with accurate_aggregation=1 will still run in multiple threads.

Overall, setting accurate_aggregation to 1 will guarantee group count and aggregate accuracy in RT indexes and plain indexes with pseudo_sharding=1. The downside is that searches will run slower because they will be forced to run in a single thread.

agent_query_timeout

Integer. Max time in milliseconds to wait for remote queries to complete, see this section.

boolean_simplify

0 or 1 (0 by default). boolean_simplify=1 enables simplifying the query to speed it up.

comment

String, user comment that gets copied to a query log file.

cutoff

Integer. Max found matches threshold. The value is selected automatically if not specified.

  • N = 0 disables the threshold
  • N > 0: instructs Manticore to stop looking for results as soon as it finds N documents.
  • not set: Manticore will decide automatically what the value should be.

In case Manticore cannot calculate the exact matching documents count you will see total_relation: gte in the query meta information, which means that the actual count is Greater Than or Equal to the total (total_found in SHOW META via SQL, hits.total in JSON via HTTP). If the total value is precise you'll get total_relation: eq.

expand_keywords

0 or 1 (0 by default). Expands keywords with exact forms and/or stars when possible. Refer to expand_keywords for more details.

field_weights

Named integer list (per-field user weights for ranking).

Example:

SELECT ... OPTION field_weights=(title=10, body=3)

global_idf

Use global statistics (frequencies) from the global_idf file for IDF computations.

idf

Quoted, comma-separated list of IDF computation flags. Known flags are:

  • normalized: BM25 variant, idf = log((N-n+1)/n), as per Robertson et al
  • plain: plain variant, idf = log(N/n), as per Sparck-Jones
  • tfidf_normalized: additionally divide IDF by query word count, so that TF*IDF fits into [0, 1] range
  • tfidf_unnormalized: do not additionally divide IDF by query word count where N is the collection size and n is the number of matched documents

The historically default IDF (Inverse Document Frequency) in Manticore is equivalent to OPTION idf='normalized,tfidf_normalized', and those normalizations may cause several undesired effects.

First, idf=normalized causes keyword penalization. For instance, if you search for the | something and the occurs in more than 50% of the documents, then documents with both keywords the and something will get less weight than documents with just one keyword something. Using OPTION idf=plain avoids this. Plain IDF varies in [0, log(N)] range, and keywords are never penalized; while the normalized IDF varies in [-log(N), log(N)] range, and too frequent keywords are penalized.

Second, idf=tfidf_normalized causes IDF drift over queries. Historically, we additionally divided IDF by query keyword count, so that the entire sum(tf*idf) over all keywords would still fit into [0,1] range. However, that means that queries word1 and word1 | nonmatchingword2 would assign different weights to the exactly same result set, because the IDFs for both word1 and nonmatchingword2 would be divided by 2. OPTION idf='tfidf_unnormalized' fixes that. Note that BM25, BM25A, BM25F() ranking factors will be scale accordingly once you disable this normalization.

IDF flags can be mixed; plain and normalized are mutually exclusive; tfidf_unnormalized and tfidf_normalized are mutually exclusive; and unspecified flags in such a mutually exclusive group take their defaults. That means that OPTION idf=plain is equivalent to a complete OPTION idf='plain,tfidf_normalized' specification.

index_weights

Named integer list. Per-table user weights for ranking.

local_df

0 or 1,automatically sum DFs over all the local parts of a distributed table, so that the IDF is consistent (and precise) over a locally sharded table.

low_priority

0 or 1 (0 by default). low_priority=1 runs the query with low priority in terms of Linux CPU scheduling. Consider also option threads=1 instead, or use that together with low_priority=1, as it might be better in some use cases.

max_matches

Integer. Per-query max matches value.

Maximum amount of matches that the server keeps in RAM for each table and can return to the client. 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 table. 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 table 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.

See also max_matches_increase_threshold, which can affect the behavior of the max_matches option.

max_matches_increase_threshold

Integer. Sets the threshold that max_matches can be increased to. Default is 16384.

Manticore may increase max_matches to improve groupby and/or aggregation accuracy when pseudo_sharding is enabled and if it detects that the number of unique values of groupby attribute is less than this threshold. Loss of accuracy may occur when pseudo sharding executes the query in several threads or RT table performs parallel searches in disk chunks.

If the number of unique values of groupby attribute is less than the threshold, max_matches will be set to this number. Otherwise, default max_matches will be used.

If max_matches was set explicitly in query options, this threshold has no effect.

Note that if this threshold is set too high, the result will be increased memory consumption and general performance degradation.

You can also force guaranteed groupby/aggregate accuracy mode using accurate_aggregation option.

max_query_time

Sets maximum search query time, in milliseconds. Must be a non-negative integer. Default value is 0 which means "do not limit". Local search queries will be stopped once that much time has elapsed. Note that if you're performing a search which queries several local tables, this limit applies to each table separately. Note it may increase the query's response time a little bit, the overhead is caused by constant tracking if it's time to stop the query.

max_predicted_time

Integer. Max predicted search time, see predicted_time_costs.

morphology

none allows to replace all query terms with their exact forms if table was built with index_exact_words enabled. Useful to prevent stemming or lemmatizing query terms.

not_terms_only_allowed

0 or 1, allows standalone negation for the query. Default is 0. See also corresponding global setting.

‹›
  • SQL
SQL
📋
MySQL [(none)]> select * from tbl where match('-donald');
ERROR 1064 (42000): index t: query error: query is non-computable (single NOT operator)
MySQL [(none)]> select * from t where match('-donald') option not_terms_only_allowed=1;
+---------------------+-----------+
| id                  | field     |
+---------------------+-----------+
| 1658178727135150081 | smth else |
+---------------------+-----------+

ranker

Any of:

  • proximity_bm25
  • bm25
  • none
  • wordcount
  • proximity
  • matchany
  • fieldmask
  • sph04
  • expr
  • or export

Refer to Search results ranking for more details on each ranker.

rand_seed

Lets you specify a specific integer seed value for an ORDER BY RAND() query, for example: ... OPTION rand_seed=1234. By default, a new and different seed value is autogenerated for every query

retry_count

Integer. Distributed retries count.

retry_delay

Integer. Distributed retry delay, msec.

sort_method

  • pq - priority queue, set by default
  • kbuffer - gives faster sorting for already pre-sorted data, e.g. table data sorted by id The result set is in both cases the same; picking one option or the other may just improve (or worsen!) performance.

threads

Limits max number of threads to use for current query processing. Default - no limit (the query can occupy all threads as defined globally). For batch of queries the option must be attached to the very first query in the batch, and it is then applied when working queue is created and then is effective for the whole batch. This option has same meaning as option max_threads_per_query, but applied only to the current query or batch of queries.

token_filter

Quoted, colon-separated of library name:plugin name:optional string of settings. Query-time token filter gets created on search each time full-text invoked by every table involved and let you implement a custom tokenizer that makes tokens according to custom rules.

SELECT * FROM index WHERE MATCH ('yes@no') OPTION token_filter='mylib.so:blend:@'

Query optimizer hints

In rare cases, Manticore's built-in query analyzer may be incorrect in understanding a query and determining whether a docid index, secondary indexes, or columnar scan should be used. To override the query optimizer's decisions, you can use the following hints in your query:

  • /*+ DocidIndex(id) */ to force the use of a docid index, /*+ NO_DocidIndex(id) */ to tell the optimizer to ignore it
  • /*+ SecondaryIndex(<attr_name1>[, <attr_nameN>]) */ to force the use of a secondary index (if available), /*+ NO_SecondaryIndex(id) */ to tell the optimizer to ignore it
  • /*+ ColumnarScan(<attr_name1>[, <attr_nameN>]) */ to force the use of a columnar scan (if the attribute is columnar), /*+ NO_ColumnarScan(id) */ to tell the optimizer to ignore it

For more information on how the query optimizer works, see the Cost based optimizer.

‹›
  • SQL
SQL
📋
SELECT * FROM students where age > 21 /*+ SecondaryIndex(age) */