Question

我在sqlite3 db中有表位于内存映射分区。我有这个查询计划的select语句：

sqlite3> EXPLAIN QUERY PLAN SELECT sst.prefix, 0 AS pb, sst.rate, 
sst.rate_n, sst.interval_1, sst.interval_n
FROM sch AS sst
WHERE
sst.i_workbook_id = 989 AND 
sst.prefix IN ('', '1', '19', '191', '1919', '19198') AND 
sst.activation_date <= DATETIME('now') AND 
(sst.expiration_date > DATETIME('now') OR sst.expiration_date IS NULL) AND
sst.start_time <= TIME('now') AND 
sst.end_time >= TIME('now');


0|0|0|SCAN TABLE sch AS sst (~185 rows)
0|0|0|EXECUTE LIST SUBQUERY 1

现在，如果我按顺序添加订单，那么他的查询计划会两次点击主表，并且比之前的表格长4倍，而表格有130万行，但过滤的行只有两行。

以下是新的查询计划：

sqlite3> EXPLAIN QUERY PLAN SELECT sst.prefix, 0 AS pb, sst.rate, 
sst.rate_n, sst.interval_1, sst.interval_n
FROM sch AS sst
WHERE
sst.i_workbook_id = 989 AND 
sst.prefix IN ('', '1', '19', '191', '1919', '19198') AND 
sst.activation_date <= DATETIME('now') AND 
(sst.expiration_date > DATETIME('now') OR sst.expiration_date IS NULL) AND
sst.start_time <= TIME('now') AND 
sst.end_time >= TIME('now') order by sst.prefix;     

0|0|0|EXECUTE LIST SUBQUERY 1
0|0|0|SEARCH TABLE sch AS sst USING AUTOMATIC COVERING INDEX (i_workbook_id=?) (~7 rows)
0|0|0|EXECUTE LIST SUBQUERY 1
0|0|0|SEARCH TABLE sch AS sst USING AUTOMATIC COVERING INDEX (i_workbook_id=?) (~7 rows)
0|0|0|EXECUTE LIST SUBQUERY 1
0|0|0|USE TEMP B-TREE FOR ORDER BY

有什么想法吗？我做错了什么？

Answer 1

SQLite尝试通过在i_workbook_id列上创建临时索引来优化您的第二个查询。该查找已多次执行，因为OR表达式无法以更好的方式进行优化。（请参阅query optimizer和query planner文档。）

你可以尝试做的是

在i_workbook_id列上创建索引;或
在i_workbook_id和prefix列上创建索引;或
在这两个上创建覆盖索引，然后在SELECT和WHERE子句中提及所有其他列。

sqlite3订单需要时间

1 个答案: