Patch: pg_trgm: gin index scan performance for similarity search

Поиск
Список
Период
Сортировка
От Fornaroli Christophe
Тема Patch: pg_trgm: gin index scan performance for similarity search
Дата
Msg-id CAEnM=mmtpunGtWQTA20wkeNzMrdGViTWMJ33=xmsyzQJPD5VYQ@mail.gmail.com
обсуждение исходный текст
Ответы Re: Patch: pg_trgm: gin index scan performance for similarity search  (Alexander Korotkov <a.korotkov@postgrespro.ru>)
Re: Patch: pg_trgm: gin index scan performance for similarity search  (Teodor Sigaev <teodor@sigaev.ru>)
Список pgsql-hackers
Hi,

I think that we can improve the gin index scan performance for similarity search defined in the pg_trgm extension. The similarity function is (for the default case where DIVUNION is defined in the code):

        count / (len1 + len2 - count) >= trgm_limit

where
  len1 is the number of unique trigrams for the first string,
  len2 is the same number for the second string,
  count is the number of common trigrams between both strings,
  trgm_limit is a user specfied limit in [0, 1].

The code used to determine if a tuple may match the query string is:

        case SimilarityStrategyNumber:
            /* Count the matches */
            ntrue = 0;
            for (i = 0; i < nkeys; i++)
            {
                if (check[i] != GIN_FALSE)
                    ntrue++;
            }
#ifdef DIVUNION
            res = (nkeys == ntrue) ? GIN_MAYBE : (((((float4) ntrue) / ((float4) (nkeys - ntrue))) >= trgm_limit) ? GIN_MAYBE : GIN_FALSE);
#else
            res = (nkeys == 0) ? GIN_FALSE : (((((float4) ntrue) / ((float4) nkeys)) >= trgm_limit) ? GIN_MAYBE : GIN_FALSE);
#endif

where
  ntrue is the number of common trigrams in both strings,
  nkeys is the number of trigrams in the search string.

This code uses this upper bound for the similarity: ntrue / (nkeys - ntrue). But if there is ntrue trigrams in common, we know that the indexed string is at least ntrue trigrams long. We can then use a more aggressive upper bound: ntrue / (ntrue + nkeys - ntrue) or ntrue / nkeys. Attached is a patch that changes this.

Here are some performance gains with this test case:

create table foo as select
  substring(md5(random()::text) for random() * 5) || '123' as bar
from generate_series(1,1000000);

create index on foo using gin (bar gin_trgm_ops);

patched:

test=# explain analyze select count(*) from foo where bar % 'abc123';
                                                              QUERY PLAN                                                              
---------------------------------------------------------------------------------------------------------------------------------------
 Aggregate  (cost=2511.14..2511.15 rows=1 width=0) (actual time=807.434..807.435 rows=1 loops=1)
   ->  Bitmap Heap Scan on foo  (cost=99.75..2508.64 rows=1000 width=0) (actual time=109.893..787.261 rows=54746 loops=1)
         Recheck Cond: (bar % 'abc123'::text)
         Rows Removed by Index Recheck: 55125
         Heap Blocks: exact=4514
         ->  Bitmap Index Scan on foo_bar_idx  (cost=0.00..99.50 rows=1000 width=0) (actual time=108.456..108.456 rows=109871 loops=1)
               Index Cond: (bar % 'abc123'::text)
 Planning time: 0.353 ms
 Execution time: 807.593 ms
(9 rows)

test=# explain analyze select count(*) from foo where bar % 'abcdef';
                                                           QUERY PLAN                                                          
--------------------------------------------------------------------------------------------------------------------------------
 Aggregate  (cost=2511.14..2511.15 rows=1 width=0) (actual time=4.829..4.830 rows=1 loops=1)
   ->  Bitmap Heap Scan on foo  (cost=99.75..2508.64 rows=1000 width=0) (actual time=3.512..4.794 rows=5 loops=1)
         Recheck Cond: (bar % 'abcdef'::text)
         Rows Removed by Index Recheck: 137
         Heap Blocks: exact=139
         ->  Bitmap Index Scan on foo_bar_idx  (cost=0.00..99.50 rows=1000 width=0) (actual time=3.355..3.355 rows=142 loops=1)
               Index Cond: (bar % 'abcdef'::text)
 Planning time: 0.363 ms
 Execution time: 5.061 ms
(9 rows)


master:

test=# explain analyze select count(*) from foo where bar % 'abc123';
                                                              QUERY PLAN                                                              
---------------------------------------------------------------------------------------------------------------------------------------
 Aggregate  (cost=2511.14..2511.15 rows=1 width=0) (actual time=6416.554..6416.554 rows=1 loops=1)
   ->  Bitmap Heap Scan on foo  (cost=99.75..2508.64 rows=1000 width=0) (actual time=484.359..6389.819 rows=54746 loops=1)
         Recheck Cond: (bar % 'abc123'::text)
         Rows Removed by Index Recheck: 945250
         Heap Blocks: exact=4514
         ->  Bitmap Index Scan on foo_bar_idx  (cost=0.00..99.50 rows=1000 width=0) (actual time=482.677..482.677 rows=999996 loops=1)
               Index Cond: (bar % 'abc123'::text)
 Planning time: 0.359 ms
 Execution time: 6416.945 ms
(9 rows)

test=# explain analyze select count(*) from foo where bar % 'abcdef';
                                                           QUERY PLAN                                                          
---------------------------------------------------------------------------------------------------------------------------------
 Aggregate  (cost=2511.14..2511.15 rows=1 width=0) (actual time=30.678..30.679 rows=1 loops=1)
   ->  Bitmap Heap Scan on foo  (cost=99.75..2508.64 rows=1000 width=0) (actual time=9.020..30.643 rows=5 loops=1)
         Recheck Cond: (bar % 'abcdef'::text)
         Rows Removed by Index Recheck: 2789
         Heap Blocks: exact=2110
         ->  Bitmap Index Scan on foo_bar_idx  (cost=0.00..99.50 rows=1000 width=0) (actual time=7.696..7.696 rows=2794 loops=1)
               Index Cond: (bar % 'abcdef'::text)
 Planning time: 0.254 ms
 Execution time: 30.809 ms
(9 rows)


Cheers,

Christophe
Вложения

В списке pgsql-hackers по дате отправления:

Предыдущее
От: Caleb Meredith
Дата:
Сообщение: Row level security (RLS) for updatable views
Следующее
От: Emre Hasegeli
Дата:
Сообщение: Re: BRIN cost estimate