Retrieving position and number of lexem occurrences from tsvector

Question

Is there any way to get information about positions of a lexeme in a sentence and number of occurrences from tsvector?

Something like this

SELECT *
FROM get_position(to_tsvector('english', 'The Fat Rats'), to_tsquery('Rats'));

will return 3

and

SELECT *
FROM get_occurrences(to_tsvector('english', 'The Fat Rats'), to_tsquery('Rats'));

will return 1.

Answer 1

Textual representation of a tsvector contains a list of occurrences for particular lexeme:

test=# select to_tsvector ( 'english', 'new bar in New York' );
        to_tsvector
----------------------------
 'bar':2 'new':1,4 'york':5

Below is exemplary function relying on that. It accepts text parameters and converts them to ts_vector internally, but could be easily rewritten to accept ts_vector.

CREATE OR REPLACE FUNCTION lexeme_occurrences (
    IN _document text
,   IN _word text
,   IN _config regconfig
,   OUT lexeme_count int
,   OUT lexeme_positions int[]
) RETURNS RECORD
AS $$
DECLARE
    _lexemes tsvector := to_tsvector ( _config, _document );
    _searched_lexeme tsvector := strip ( to_tsvector ( _config, _word ) );
    _occurences_pattern text := _searched_lexeme::text || ':([0-9,]+)';
    _occurences_list text := substring ( _lexemes::text, _occurences_pattern );
BEGIN
    SELECT
        count ( a )
    ,   array_agg ( a::int )
    FROM regexp_split_to_table ( _occurences_list, ',' ) a
    WHERE _searched_lexeme::text != '' -- preventing false positives
    INTO
        lexeme_count
    ,   lexeme_positions;
    RETURN;
END $$ LANGUAGE plpgsql;

Example usage:

select * from lexeme_occurrences ( 'The Fat Rats', 'rat', 'english' );
 lexeme_count | lexeme_positions
--------------+-----------------
            1 | {3}
(1 row)