Question

我有一个函数，它使用传递给它的变量执行动态查询。

我定义了表达式索引，但它们没有被使用。平均查询大约需要7秒钟。如果我这样做：

 SET enable_seqscan=f;

同样的查询只需1秒钟。如何让Postgres使用索引？

这是我的功能：

--Usage: select * from scored_search('hinton', 0.5, 1.0, 10000);

CREATE OR REPLACE FUNCTION scored_search(
vn character varying(500), tn real, wn real, lim integer 
) RETURNS SETOF contact_index_with_score AS 
$func$
DECLARE
  nlim integer := lim/4;
BEGIN
RETURN QUERY EXECUTE
  'SELECT *,
      (CASE gn > $2 WHEN true THEN gn ELSE 0 END) * $3
    AS score
  FROM (
    SELECT *,
    /* Word sim parameters ordered so name token can match multi-token name */
      greatest(multi_word_similarity(coalesce($1,''''), coalesce(contact_name1,'''')),
               multi_word_similarity(coalesce($1,''''), coalesce(contact_name2,'''')),
               multi_word_similarity(coalesce($1,''''), coalesce(contact_name3,'''')),
               multi_word_similarity(coalesce($1,''''), coalesce(contact_name4,'''')),
               multi_word_similarity(coalesce($1,''''), coalesce(contact_name5,''''))) AS gn
    FROM (
      (SELECT * FROM contact_search_index WHERE $1 <%
        normalize_string(coalesce(contact_name1,'''') || '' '' ||
         coalesce(contact_name2,'''') || '' '' ||
         coalesce(contact_name3,'''') || '' '' ||
         coalesce(contact_name4,'''') || '' '' ||
         coalesce(contact_name5,'''')) COLLATE "default" LIMIT $13)
    ) AS filtered_contact_index
  ) AS max_similarities
  ORDER BY score DESC'
USING 
  vn, tn, wn, nlim;
END
$func$
LANGUAGE plpgsql;

这是我的索引定义：

CREATE INDEX contact_search_name_trgm_idx on contact_search_index 
USING GIN((
  normalize_string(coalesce(contact_name1, '') || ' ' ||
  coalesce(contact_name2, '') || ' ' ||
  coalesce(contact_name3, '') || ' ' ||
  coalesce(contact_name4, '') || ' ' ||
  coalesce(contact_name5, '')) 
) COLLATE "default" gin_trgm_ops);

Postgres trgm GIN索引不用于查询内部函数

0 个答案: