Postgresql Query执行缓慢

Question

我有PostgreSql查询如下：

SELECT DISTINCT ON (reference) reference, reference_url 
FROM vehicles v 
WHERE NOT EXISTS 
    (select reference 
     from daily_run_vehicle rv 
     WHERE ((
           handled = False 
           AND retries >= 5 ) 
           OR rv.timestamp::timestamp::date = now()::date)  
     AND v.reference=reference);

其中vehicles表有大约400k记录，daily_run_vehicle表有大约5000万条记录。

因此，我需要所有车辆今天没有添加到daily_run_vehicle的车辆或处理列是False并重试column is >= 5。

但问题是查询执行时间太长。

有没有办法更好地编写它以便更快地执行？

Answer 1

我有一个理论，它可能与调用now（）函数数百万次有关。您可以通过运行此查询来验证

SELECT DISTINCT ON (reference) reference, reference_url 
FROM vehicles v 
WHERE NOT EXISTS 
    (select reference 
     from daily_run_vehicle rv 
     WHERE ((
           handled = False 
           AND retries >= 5 ) 
           OR rv.timestamp::timestamp::date = '2019-03-06')  
     AND v.reference=reference);

它的性能得到改善，你必须将今天的日期设置为一个变量并在查询中使用变量，这样现在只需要调用一次。如果你使用EXISTS，那么传统就是选择SELECT ... FROM ...你不关心这些值是否至少有一个或没有。

Answer 2

嗯。我在想：

SELECT DISTINCT ON (v.reference) v.reference, v.reference_url 
FROM vehicles v 
WHERE NOT EXISTS (select 1 
                  from daily_run_vehicle rv 
                  where rv.reference = v.reference and
                        rv.handled = False and
                        rv.retries >= 5
                 ) and
      NOT EXISTS (select 1 
                  from daily_run_vehicle rv 
                  where rv.reference = v.reference and
                        rv.timestamp >= current_date::timestamp and
                        rv.timestamp >= (current_date + interval '1 day'::timestamp 
                 )
ORDER BY v.reference;

对于此查询，您需要索引：

• daily_run_vehicle(reference, handled, retries)
• daily_run_vehicle(reference, timestamp)
• reference_url(reference, reference_url)