Oracle在创建索引时,不会存储NULL值,而PostgreSQL在创建索引时则会存储NULL值.在查询时,如使用Column is null这样的条件查询,Oracle不会使用索引而PostgreSQL则会使用索引. **Oracle** 插入数据,200w多行的数据,然后插入一行值为null的数据. ``` TEST-orcl@DESKTOP-V430TU3>create table tbl1(id int); Table created. TEST-orcl@DESKTOP-V430TU3>create global temporary table tmp(id int); Table created. TEST-orcl@DESKTOP-V430TU3>insert into tmp select rownum from dba_objects; 133456 rows created. TEST-orcl@DESKTOP-V430TU3>insert into tmp select * from tmp; 133455 rows created. TEST-orcl@DESKTOP-V430TU3>/ 266910 rows created. TEST-orcl@DESKTOP-V430TU3>/ 533820 rows created. TEST-orcl@DESKTOP-V430TU3>/ 1067640 rows created. TEST-orcl@DESKTOP-V430TU3>insert into tbl1 select * from tmp; 2135296 rows created. TEST-orcl@DESKTOP-V430TU3>commit; Commit complete. TEST-orcl@DESKTOP-V430TU3>exec dbms_stats.gather_table_stats('TEST','TBL1',cascade=>true); PL/SQL procedure successfully completed. TEST-orcl@DESKTOP-V430TU3>select index_name,index_type,blevel,leaf_blocks,num_rows,status,distinct_keys from user_indexes where table_name='TBL1'; INDEX_NAME INDEX_TYPE BLEVEL ------------------------------ --------------------------- ---------- LEAF_BLOCKS NUM_ROWS STATUS DISTINCT_KEYS ----------- ---------- -------- ------------- IDX_TBL1_ID NORMAL 2 4662 2103843 VALID 134688 TEST-orcl@DESKTOP-V430TU3>insert into tbl1 values(null); 1 row created. TEST-orcl@DESKTOP-V430TU3>commit; Commit complete. TEST-orcl@DESKTOP-V430TU3> TEST-orcl@DESKTOP-V430TU3>exec dbms_stats.gather_table_stats('TEST','TBL1',cascade=>true); PL/SQL procedure successfully completed. TEST-orcl@DESKTOP-V430TU3>select index_name,index_type,blevel,leaf_blocks,num_rows,status,distinct_keys from user_indexes where table_name='TBL1'; INDEX_NAME INDEX_TYPE BLEVEL ------------------------------ --------------------------- ---------- LEAF_BLOCKS NUM_ROWS STATUS DISTINCT_KEYS ----------- ---------- -------- ------------- IDX_TBL1_ID NORMAL 2 4771 2152683 VALID 134688 ``` 执行查询 ``` TEST-orcl@DESKTOP-V430TU3>set autotrace on explain TEST-orcl@DESKTOP-V430TU3>select * from tbl1 where id is null; ID ---------- Execution Plan ---------------------------------------------------------- Plan hash value: 312383637 -------------------------------------------------------------------------- | Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | -------------------------------------------------------------------------- | 0 | SELECT STATEMENT | | 1 | 5 | 898 (2)| 00:00:11 | |* 1 | TABLE ACCESS FULL| TBL1 | 1 | 5 | 898 (2)| 00:00:11 | -------------------------------------------------------------------------- Predicate Information (identified by operation id): --------------------------------------------------- 1 - filter("ID" IS NULL) TEST-orcl@DESKTOP-V430TU3> ``` **PostgreSQL** 数据表tbl1结构与Oracle一致. ``` testdb=# insert into tbl1 select generate_series(1,100000); INSERT 0 100000 testdb=# explain (analyze,verbose) select * from tbl1 where id is null; QUERY PLAN -------------------------------------------------------------------------------------------------------- Seq Scan on public.tbl1 (cost=0.00..1569.33 rows=11 width=4) (actual time=26.052..130.752 rows=1 loops=1) Output: id Filter: (tbl1.id IS NULL) Rows Removed by Filter: 110000 Planning Time: 1.403 ms Execution Time: 130.814 ms (6 rows) testdb=# create index idx_tb1_id on tbl1(id); CREATE INDEX testdb=# explain (analyze,verbose) select * from tbl1 where id is null; QUERY PLAN -------------------------------------------------------------------------------------------------------- Index Only Scan using idx_tb1_id on public.tbl1 (cost=0.42..8.56 rows=4 width=4) (actual time=0.133..0.136 rows=1 loops=1) Output: id Index Cond: (tbl1.id IS NULL) Heap Fetches: 1 Planning Time: 1.512 ms Execution Time: 0.199 ms (6 rows) ``` 使用id is null进行查询,使用的是Index Only Scan.