sql - 改善SQLite查询

Question

我有一个事件表，其中包含约100k行。
以下是有关“事件”表的复杂查询，希望对如何优化此查询有意见...

SELECT id FROM event 
NATURAL JOIN (
    SELECT subj_id, max(timestamp) AS timestamp
    FROM event WHERE (
        timestamp >= 1342052128597 AND timestamp <= 9223372036854775807
        AND NOT subj_interpretation = 46))
    GROUP BY subj_id)
GROUP BY subj_id
ORDER BY  
timestamp DESC

查询在这里执行0.06-0.07秒。任何想法都是查询的良好索引和/或重组。

目前，我正在使用：

CREATE INDEX event_subj_id ON event(subj_id, timestamp, subj_interpretation)

查询计划如下所示：

1 0 0 SCAN TABLE event USING COVERING INDEX event_subj_id (~27777 rows)
0 0 1 SCAN SUBQUERY 1 (~100 rows)
0 1 0 SEARCH TABLE event USING INDEX event_subj_id (subj_id=? AND timestamp=?) (~9 rows)
0 0 0 USE TEMP B-TREE FOR GROUP BY
0 0 0 USE TEMP B-TREE FOR ORDER BY

Answer 1

您可以通过删除DISTINCT优化第二个查询，因为GROUP BY subj_id已经确保了没有重复的记录：

SELECT id,
       subj_id,
       max(timestamp) AS timestamp
FROM event
WHERE timestamp BETWEEN 1342055894621 AND 9223372036854775807
  AND subj_interpretation != 46
GROUP BY subj_id
ORDER BY timestamp

对于GROUP BY / timestamp / subj_interpretation查找，索引本身已经是最佳的。
但是，如果向其中添加id列，SQLite可以将其用作覆盖索引，这样就无需在events表本身中查找任何记录，这实际上可能使性能提高一倍：

0 0 0 SCAN TABLE event USING COVERING INDEX event_subj_id_plus_id
0 0 0 USE TEMP B-TREE FOR ORDER BY