Searching and ranking functions

BM25A()

BM25A(k1,b) returns the exact BM25A() value. Requires the expr ranker and enabled index_field_lengths. Parameters k1 and b must be floats.

BM25F()

BM25F(k1, b, {field=weight, ...}) returns the exact BM25F() value and requires index_field_lengths to be enabled. The expr ranker is also necessary. Parameters k1 and b must be floats.

EXIST()

Substitutes non-existent columns with default values. It returns either the value of an attribute specified by 'attr-name', or the 'default-value' if that attribute does not exist. STRING or MVA attributes are not supported. This function is useful when searching through multiple tables with different schemas.

SELECT *, EXIST('gid', 6) as cnd FROM i1, i2 WHERE cnd>5

MIN_TOP_SORTVAL()

Returns the sort key value of the worst-ranked element in the current top-N matches if the sort key is a float, and 0 otherwise.

MIN_TOP_WEIGHT()

Returns the weight of the worst-ranked element in the current top-N matches.

PACKEDFACTORS()

PACKEDFACTORS() can be used in queries to display all calculated weighting factors during matching or to provide a binary attribute for creating a custom ranking UDF. This function only works if the expression ranker is specified and the query is not a full scan; otherwise, it returns an error. PACKEDFACTORS() can take an optional argument that disables ATC ranking factor calculation: PACKEDFACTORS({no_atc=1}). Calculating ATC significantly slows down query processing, so this option can be useful if you need to see the ranking factors but don't require ATC. PACKEDFACTORS() can also output in JSON format: PACKEDFACTORS({json=1}). The respective outputs in either key-value pair or JSON format are shown below. (Note that the examples below are wrapped for readability; actual returned values would be single-line.)

mysql> SELECT id, PACKEDFACTORS() FROM test1
    -> WHERE MATCH('test one') OPTION ranker=expr('1') \G
*************************** 1\. row ***************************
             id: 1
packedfactors(): bm25=569, bm25a=0.617197, field_mask=2, doc_word_count=2,
    field1=(lcs=1, hit_count=2, word_count=2, tf_idf=0.152356,
        min_idf=-0.062982, max_idf=0.215338, sum_idf=0.152356, min_hit_pos=4,
        min_best_span_pos=4, exact_hit=0, max_window_hits=1, min_gaps=2,
        exact_order=1, lccs=1, wlccs=0.215338, atc=-0.003974),
    word0=(tf=1, idf=-0.062982),
    word1=(tf=1, idf=0.215338)
1 row in set (0.00 sec)
mysql> SELECT id, PACKEDFACTORS({json=1}) FROM test1
    -> WHERE MATCH('test one') OPTION ranker=expr('1') \G
*************************** 1\. row ***************************
                     id: 1
packedfactors({json=1}):
{

    "bm25": 569,
    "bm25a": 0.617197,
    "field_mask": 2,
    "doc_word_count": 2,
    "fields": [
        {
            "lcs": 1,
            "hit_count": 2,
            "word_count": 2,
            "tf_idf": 0.152356,
            "min_idf": -0.062982,
            "max_idf": 0.215338,
            "sum_idf": 0.152356,
            "min_hit_pos": 4,
            "min_best_span_pos": 4,
            "exact_hit": 0,
            "max_window_hits": 1,
            "min_gaps": 2,
            "exact_order": 1,
            "lccs": 1,
            "wlccs": 0.215338,
            "atc": -0.003974
        }
    ],
    "words": [
        {
            "tf": 1,
            "idf": -0.062982
        },
        {
            "tf": 1,
            "idf": 0.215338
        }
    ]

}
1 row in set (0.01 sec)

This function can be used to implement custom ranking functions in UDFs, as in:

SELECT *, CUSTOM_RANK(PACKEDFACTORS()) AS r
FROM my_index
WHERE match('hello')
ORDER BY r DESC
OPTION ranker=expr('1');

Where CUSTOM_RANK() is a function implemented in a UDF. It should declare a SPH_UDF_FACTORS structure (defined in sphinxudf.h), initialize this structure, unpack the factors into it before usage, and deinitialize it afterwards, as follows:

SPH_UDF_FACTORS factors;
sphinx_factors_init(&factors);
sphinx_factors_unpack((DWORD*)args->arg_values[0], &factors);
// ... can use the contents of factors variable here ...
sphinx_factors_deinit(&factors);

PACKEDFACTORS() data is available at all query stages, not just during the initial matching and ranking pass. This enables another particularly interesting application of PACKEDFACTORS(): re-ranking.

In the example above, we used an expression-based ranker with a dummy expression and sorted the result set by the value computed by our UDF. In other words, we used the UDF to rank all our results. Now, let's assume for the sake of an example that our UDF is extremely expensive to compute, with a throughput of only 10,000 calls per second. If our query matches 1,000,000 documents, we would want to use a much simpler expression to do most of our ranking in order to maintain reasonable performance. Then, we would apply the expensive UDF to only a few top results, say, the top 100 results. In other words, we would build the top 100 results using a simpler ranking function and then re-rank those with a more complex one. This can be done with subselects:

SELECT * FROM (
    SELECT *, CUSTOM_RANK(PACKEDFACTORS()) AS r
    FROM my_index WHERE match('hello')
    OPTION ranker=expr('sum(lcs)*1000+bm25')
    ORDER BY WEIGHT() DESC
    LIMIT 100
) ORDER BY r DESC LIMIT 10

In this example, the expression-based ranker is called for every matched document to compute WEIGHT(), so it gets called 1,000,000 times. However, the UDF computation can be postponed until the outer sort, and it will only be performed for the top 100 matches by WEIGHT(), according to the inner limit. This means the UDF will only be called 100 times. Finally, the top 10 matches by UDF value are selected and returned to the application.

For reference, in a distributed setup, the PACKEDFACTORS() data is sent from the agents to the master node in binary format. This makes it technically feasible to implement additional re-ranking passes on the master node if needed.

When used in SQL but not called from any UDFs, the result of PACKEDFACTORS() is formatted as plain text, which can be used to manually assess the ranking factors. Note that this feature is not currently supported by the Manticore API.

REMOVE_REPEATS()

REMOVE_REPEATS ( result_set, column, offset, limit ) - removes repeated adjusted rows with the same 'column' value.

SELECT REMOVE_REPEATS((SELECT * FROM dist1), gid, 0, 10)

WEIGHT()

The WEIGHT() function returns the calculated matching score. If no ordering is specified, the result is sorted in descending order by the score provided by WEIGHT(). In this example, we order first by weight and then by an integer attribute.

The search above performs a simple matching, where all words need to be present. However, we can do more (and this is just a simple example):

mysql> SELECT *,WEIGHT() FROM testrt WHERE MATCH('"list of business laptops"/3');
+------+------+-------------------------------------+---------------------------+----------+
| id   | gid  | title                               | content                   | weight() |
+------+------+-------------------------------------+---------------------------+----------+
|    1 |   10 | List of HP business laptops         | Elitebook Probook         |     2397 |
|    2 |   10 | List of Dell business laptops       | Latitude Precision Vostro |     2397 |
|    3 |   20 | List of Dell gaming laptops         | Inspirion Alienware       |     2375 |
|    5 |   30 | List of ASUS ultrabooks and laptops | Zenbook Vivobook          |     2375 |
+------+------+-------------------------------------+---------------------------+----------+
4 rows in set (0.00 sec)


mysql> SHOW META;
+----------------+----------+
| Variable_name  | Value    |
+----------------+----------+
| total          | 4        |
| total_found    | 4        |
| total_relation | eq       |
| time           | 0.000    |
| keyword[0]     | list     |
| docs[0]        | 5        |
| hits[0]        | 5        |
| keyword[1]     | of       |
| docs[1]        | 4        |
| hits[1]        | 4        |
| keyword[2]     | business |
| docs[2]        | 2        |
| hits[2]        | 2        |
| keyword[3]     | laptops  |
| docs[3]        | 5        |
| hits[3]        | 5        |
+----------------+----------+
16 rows in set (0.00 sec)

Here, we search for four words, but a match can occur even if only three of the four words are found. The search will rank documents containing all words higher.

ZONESPANLIST()

The ZONESPANLIST() function returns pairs of matched zone spans. Each pair contains the matched zone span identifier, a colon, and the order number of the matched zone span. For example, if a document reads <emphasis role="bold"><i>text</i> the <i>text</i></emphasis>, and you query for 'ZONESPAN:(i,b) text', then ZONESPANLIST() will return the string "1:1 1:2 2:1", meaning that the first zone span matched "text" in spans 1 and 2, and the second zone span in span 1 only.

QUERY()

QUERY() returns the current search query. QUERY() is a postlimit expression and is intended to be used with SNIPPET().

Table functions are a mechanism for post-query result set processing. Table functions take an arbitrary result set as input and return a new, processed set as output. The first argument should be the input result set, but a table function can optionally take and handle more arguments. Table functions can completely change the result set, including the schema. Currently, only built-in table functions are supported. Table functions work for both outer SELECT and nested SELECT.

Type Casting Functions

Type casting comprises three principal actions: conversion, reinterpretation, and promotion.

  • Conversion. This refers to the process of changing the data type of a value to another data type. This involves additional computations and is exclusively performed by the TO_STRING() function.
  • Reinterpretation. This involves treating the binary data representing a value as if it were of a different data type, without actually changing the underlying data. This is handled by SINT(), doesn't involve extra computations; instead, it merely reinterprets existing data.
  • Promotion. This refers to the process of converting a value to a "larger" or more precise data type. It doesn't require extra computation either; it merely requests the argument to deliver a value of a different type. Only JSON fields and a few other functions can promote their values to integers. If an argument cannot yield a value of a different type, the promotion will fail. For instance, the TIMEDIFF() function usually returns a string, but can also return a number. So, BIGINT(TIMEDIFF(1,2)) will execute successfully, compelling TIMEDIFF() to supply an integer value. Conversely, DATE_FORMAT() solely returns strings and can't yield a number, meaning that BIGINT(DATE_FORMAT(...)) will fail.

BIGINT()

This function promotes an integer argument to a 64-bit type, leaving floating-point arguments untouched. It's designed to ensure the evaluation of specific expressions (such as a*b) in 64-bit mode, even if all arguments are 32-bit.

DOUBLE()

The DOUBLE() function promotes its argument to a floating-point type. This is designed to help enforce the evaluation of numeric JSON fields.

INTEGER()

The INTEGER() function promotes its argument to a 64-bit signed type. This is designed to enforce the evaluation of numeric JSON fields.

TO_STRING()

This function forcefully converts its argument to a string type.

UINT()

The UINT() function promotes its argument to a 32-bit unsigned integer type.

UINT64()

The UINT64() function promotes its argument to a 64-bit unsigned integer type.

SINT()

The SINT() function forcefully reinterprets its 32-bit unsigned integer argument as signed and extends it to a 64-bit type (since the 32-bit type is unsigned). For instance, 1-2 ordinarily evaluates to 4294967295, but SINT(1-2) evaluates to -1.

Arrays and conditions functions

ALL()

ALL(cond FOR var IN json.array) applies to JSON arrays and returns 1 if the condition is true for all elements in the array and 0 otherwise. cond is a general expression that can also use var as the current value of an array element within itself.

‹›
  • ALL() with json
  • ALL() with json ex. 2
📋
select *, ALL(x>0 AND x<4 FOR x IN j.ar) from tbl
‹›
Response
+------+--------------+--------------------------------+
| id   | j            | all(x>0 and x<4 for x in j.ar) |
+------+--------------+--------------------------------+
|    1 | {"ar":[1,3]} |                              1 |
|    2 | {"ar":[3,7]} |                              0 |
+------+--------------+--------------------------------+
2 rows in set (0.00 sec)

ALL(mva) is a special constructor for multi-value attributes. When used with comparison operators (including comparison with IN()), it returns 1 if all values from the MVA attribute are found among the compared values.

‹›
  • ALL() with MVA
  • ALL() with MVA and IN()
📋
select * from tbl where all(m) >= 1
‹›
Response
+------+------+
| id   | m    |
+------+------+
|    1 | 1,3  |
|    2 | 3,7  |
+------+------+
2 rows in set (0.00 sec)

To compare an MVA attribute with an array, avoid using <mva> NOT ALL(); use ALL(<mva>) NOT IN() instead.

‹›
  • ALL() with MVA and NOT IN()
ALL() with MVA and NOT IN()
📋
select * from tbl where all(m) not in (2, 4)
‹›
Response
+------+------+
| id   | m    |
+------+------+
|    1 | 1,3  |
|    2 | 3,7  |
+------+------+
2 rows in set (0.00 sec)

ALL(string list) is a special operation for filtering string tags.

If all of the words enumerated as arguments of ALL() are present in the attribute, the filter matches. The optional NOT inverts the logic.

This filter internally uses doc-by-doc matching, so in the case of a full scan query, it might be slower than expected. It is intended for attributes that are not indexed, like calculated expressions or tags in PQ tables. If you need such filtering, consider the solution of putting the string attribute as a full-text field, and then use the full-text operator match(), which will invoke a full-text search.

‹›
  • ALL() with strings
  • ALL() with strings and NOT
📋
select * from tbl where tags all('bug', 'release')
‹›
Response
+------+---------------------------+
| id   | tags                      |
+------+---------------------------+
|    1 | bug priority_high release |
|    2 | bug priority_low release  |
+------+---------------------------+
2 rows in set (0.00 sec)

ANY()

ANY(cond FOR var IN json.array) applies to JSON arrays and returns 1 if the condition is true for any element in the array and 0 otherwise. cond is a general expression that can also use var as the current value of an array element within itself.

‹›
  • ANY() with json
  • ANY() with json ex. 2
📋
select *, ANY(x>5 AND x<10 FOR x IN j.ar) from tbl
‹›
Response
+------+--------------+---------------------------------+
| id   | j            | any(x>5 and x<10 for x in j.ar) |
+------+--------------+---------------------------------+
|    1 | {"ar":[1,3]} |                               0 |
|    2 | {"ar":[3,7]} |                               1 |
+------+--------------+---------------------------------+
2 rows in set (0.00 sec)

ANY(mva) is a special constructor for multi-value attributes. When used with comparison operators (including comparison with IN()), it returns 1 if any of the MVA values is found among the compared values.

When comparing an array using IN(), ANY() is assumed by default if not otherwise specified, but a warning will be issued regarding the missing constructor.

‹›
  • ANY() with MVA
  • ANY() with MVA and IN()
📋
mysql> select * from tbl

+------+------+
| id   | m    |
+------+------+
|    1 | 1,3  |
|    2 | 3,7  |
+------+------+
2 rows in set (0.01 sec)

mysql> select * from tbl where any(m) > 5

+------+------+
| id   | m    |
+------+------+
|    2 | 3,7  |
+------+------+
1 row in set (0.00 sec)
‹›
Response
+------+------+
| id   | m    |
+------+------+
|    1 | 1,3  |
|    2 | 3,7  |
+------+------+
2 rows in set (0.00 sec)

To compare an MVA attribute with an array, avoid using <mva> NOT ANY(); use <mva> NOT IN() instead or ANY(<mva>) NOT IN().

‹›
  • ANY() with MVA and NOT IN()
ANY() with MVA and NOT IN()
📋
mysql> select * from tbl

+------+------+
| id   | m    |
+------+------+
|    1 | 1,3  |
|    2 | 3,7  |
+------+------+
2 rows in set (0.00 sec)

mysql> select * from tbl where any(m) not in (1, 3, 5)

+------+------+
| id   | m    |
+------+------+
|    2 | 3,7  |
+------+------+
1 row in set (0.00 sec)

ANY(string list) is a special operation for filtering string tags.

If any of the words enumerated as arguments of ANY() is present in the attribute, the filter matches. The optional NOT inverts the logic.

This filter internally uses doc-by-doc matching, so in the case of a full scan query, it might be slower than expected. It is intended for attributes that are not indexed, like calculated expressions or tags in PQ tables. If you need such filtering, consider the solution of putting the string attribute as a full-text field, and then use the full-text operator match(), which will invoke a full-text search.

‹›
  • ANY() with strings
  • ANY() with strings and NOT
📋
select * from tbl where tags any('bug', 'feature')
‹›
Response
+------+---------------------------+
| id   | tags                      |
+------+---------------------------+
|    1 | bug priority_high release |
|    2 | bug priority_low release  |
+------+---------------------------+
2 rows in set (0.00 sec)

CONTAINS()

CONTAINS(polygon, x, y) checks whether the (x,y) point is within the given polygon, and returns 1 if true, or 0 if false. The polygon has to be specified using either the POLY2D() function. The former function is intended for "small" polygons, meaning less than 500 km (300 miles) a side, and it doesn't take into account the Earth's curvature for speed. For larger distances, you should use GEOPOLY2D, which tessellates the given polygon in smaller parts, accounting for the Earth's curvature.

IF()

The behavior of IF() is slightly different from its MySQL counterpart. It takes 3 arguments, checks 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) due to a threshold, and IF() will always return ** as a result. In the second case, the same sqrt(3)*sqrt(3)-3 expression will be compared with zero without a threshold by the IF() function itself. However, its value will be slightly different from zero due to limited floating-point calculation precision. Because of this, the comparison with 0.0 done by IF() will not pass, and the second variant will return 'a' as a result.

IN()

IN(expr,val1,val2,...) takes 2 or more arguments and returns 1 if the 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. The constants are pre-sorted, and binary search is used, so IN() even against a large arbitrary list of constants will be very quick. The first argument can also be an MVA attribute. In that case, IN() will return 1 if any of the MVA values are equal to any of the other arguments. IN() also supports IN(expr,@uservar) syntax to check whether the value belongs to the list in the given global user variable. The first argument can be a JSON attribute.

INDEXOF()

INDEXOF(cond FOR var IN json.array) function iterates through all elements in the array and returns the index of the first element for which 'cond' is true, and -1 if 'cond' is false for every element in the array.

INTERVAL()

INTERVAL(expr,point1,point2,point3,...) 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.

LENGTH()

LENGTH(attr_mva) function returns the number of elements in an MVA set. It works with both 32-bit and 64-bit MVA attributes. LENGTH(attr_json) returns the length of a field in JSON. The return value depends on the type of field. For example, LENGTH(json_attr.some_int) always returns 1, and LENGTH(json_attr.some_array) returns the number of elements in the array. LENGTH(string_expr) function returns the length of the string resulting from an expression. TO_STRING() must enclose the expression, regardless of whether the expression returns a non-string or it's simply a string attribute.

REMAP()

REMAP(condition, expression, (cond1, cond2, ...), (expr1, expr2, ...)) function allows you to make some exceptions to expression values depending on condition values. The condition expression should always result in an integer, while the expression can result in an integer or float.

Example:

SELECT id, size, REMAP(size, 15, (5,6,7,8), (1,1,2,2)) s
FROM products
ORDER BY s ASC;

This will put documents with sizes 5 and 6 first, followed by sizes 7 and 8. In case there's an original value not listed in the array (e.g. size 10), it will default to 15, and in this case, will be placed at the end.

More examples:

SELECT REMAP(userid, karmapoints, (1, 67), (999, 0)) FROM users;
SELECT REMAP(id%10, salary, (0), (0.0)) FROM employes;