postgresql 性能优化 之 模糊匹配大杀器

ArteryBase-模糊匹配大杀器

问题背景

随着pg越来越强大，abase目前已经升级到5.0(postgresql10.4)，目前abase5.0继承了全文检索插件(zhparser)，使用全文检索越来越方便。本文会对abase支持的like模糊匹配，全文检索，创建何种索引，如何使用进行说明。针对于各种模糊匹配均可走索引

前模糊匹配(%xxx),后模糊匹配(xxx%)

使用场景：如果简单的前模糊匹配或者后模糊匹配则可以建一个简单的btree索引。

–1.后模糊匹配(xxx%)
create index i_t_msys_btrre_c_ajmc on db_msys.t_msys using btree(c_ajmc text_pattern_ops);
CREATE INDEX
Time: 4189.886 ms (00:04.190)
db_15fb=# select c_ajmc from db_msys.t_msys where c_ajmc like ‘北京%’;
c_ajmc
————————
北京决定和华宇
北京和华宇信息
北京
北京华宇，北京华宇
、、、
db_15fb=# explain analyze select c_ajmc from db_msys.t_msys where c_ajmc like ‘北京%’;
QUERY PLAN
————————————————————————-
Bitmap Heap Scan on t_msys (cost=43.92..2177.91 rows=4204 width=80) (actual time=0.570..2.667 rows=1570 loops=1)
Filter: ((c_ajmc)::text ~~ ‘北京%’::text)
Heap Blocks: exact=500
-> Bitmap Index Scan on i_t_msys_btrre_c_ajmc (cost=0.00..42.87 rows=632 width=0) (actual time=0.477..0.477 rows=1570 loops=1)
Index Cond: (((c_ajmc)::text ~>=~ ‘北京’::text) AND ((c_ajmc)::text ~<~ ‘北亭’::text))
Planning time: 0.956 ms
Execution time: 2.841 ms
(7 rows)

Time: 4.848 ms
–2.前模糊匹配(%xxx),查询以c_ajmc以信息结尾的记录，使用反转函数reverse
db_15fb=# create index i_t_msys_reverse_c_ajmc on db_msys.t_msys using btree(reverse(c_ajmc) text_pattern_ops);
CREATE INDEX
Time: 4011.131 ms (00:04.011)
–查询以张三结尾的信息
db_15fb=# select c_ajmc from db_msys.t_msys where reverse(c_ajmc) like reverse(‘%张三’);
c_ajmc
———-
华宇张三
北京张三
(2 rows)

Time: 0.910 ms

–前模糊匹配也可走索引
db_15fb=# explain analyze select c_ajmc from db_msys.t_msys where reverse(c_ajmc) like reverse(‘%张三’);
QUERY PLAN

————————————————————————-
Bitmap Heap Scan on t_msys (cost=114.86..5312.88 rows=1787 width=80) (actual time=0.064..0.066 rows=2 loops=1)
Filter: (reverse((c_ajmc)::text) ~~ ‘三张%’::text)
Heap Blocks: exact=1
-> Bitmap Index Scan on i_t_msys_reverse_c_ajmc (cost=0.00..114.42 rows=1787 width=0) (actual time=0.042..0.042 rows=2 loops=1)
Index Cond: ((reverse((c_ajmc)::text) ~>=~ ‘三张’::text) AND (reverse((c_ajmc)::text) ~<~ ‘三弡’::text))
Planning time: 0.236 ms
Execution time: 0.148 ms
(7 rows)

Time: 1.211 ms

–或者使用like ‘三张%’等效于 reverse(‘%张三’)
db_15fb=# explain analyze select c_ajmc from db_msys.t_msys where reverse(c_ajmc) like ‘三张%’;
QUERY PLAN

————————————————————————-
Bitmap Heap Scan on t_msys (cost=114.86..5312.88 rows=1787 width=80) (actual time=0.056..0.058 rows=2 loops=1)
Filter: (reverse((c_ajmc)::text) ~~ ‘三张%’::text)
Heap Blocks: exact=1
-> Bitmap Index Scan on i_t_msys_reverse_c_ajmc (cost=0.00..114.42 rows=1787 width=0) (actual time=0.036..0.036 rows=2 loops=1)
Index Cond: ((reverse((c_ajmc)::text) ~>=~ ‘三张’::text) AND (reverse((c_ajmc)::text) ~<~ ‘三弡’::text))
Planning time: 0.259 ms
Execution time: 0.108 ms
(7 rows)

Time: 1.119 ms