Constraints on Index Use

The following is an example of using an index. Run the following statements in a database that uses the UTF-8 or GBK encoding:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12

create table table1 (c_int int,c_bigint bigint,c_varchar varchar,c_text text) with(orientation=row);

create text search configuration ts_conf_1(parser=POUND);
create text search configuration ts_conf_2(parser=POUND) with(split_flag='%');

set default_text_search_config='ts_conf_1';
create index idx1 on table1 using gin(to_tsvector(c_text));

set default_text_search_config='ts_conf_2';
create index idx2 on table1 using gin(to_tsvector(c_text));

select c_varchar,to_tsvector(c_varchar) from table1 where to_tsvector(c_text) @@ plainto_tsquery('¥#@...&**')   and to_tsvector(c_text) @@ plainto_tsquery('Company ')   and c_varchar is not null order by 1 desc limit 3;

In this example, table1 has two GIN indexes created on the same column c_text, idx1 and idx2, but these two indexes are created under different settings of default_text_search_config. Differences between this example and the scenario where one table has common indexes created on the same column are as follows:

• GIN indexes use different parsers (that is, different delimiters). In this case, the index data of idx1 is different from that of idx2.
• In the specified scenario, the index data of multiple common indexes created on the same column is the same.

As a result, using idx1 and idx2 for the same query returns different results.

• Multiple GIN indexes are created on the same column of the same table.
• The GIN indexes use different parsers (that is, different delimiters).
• The column is used in a query, and an index scan is used in the execution plan.

  To avoid different query results caused by different GIN indexes, ensure that only one GIN index is available on a column of the physical table.