SELECT *
FROM trips,
dates
WHERE places_number_on_flag > 0
AND places_number > 0
AND ( places_number - ( (SELECT Count(id)
FROM resrv_customers
WHERE trip = trips.id
AND ow > 0)
+ (SELECT Count(id)
FROM resrv_customers
WHERE trip = trips.id
AND extra_seat = 1
AND ow > 0) ) < 20 )
AND dates.id = trips.trip_date
AND dates.from_date > 2458553;
它在大约一分钟和50秒后返回5行。
我希望它更快。
一个优化是仅运行子查询一次并使用条件聚合来计算席位:
SELECT *
FROM trips
JOIN dates on dates.id = trips.trip_date
WHERE places_number_on_flag > 0
AND places_number > 0
AND ( places_number - ( (SELECT Count(id) + count(id) filter (where extra_seat = 1)
FROM resrv_customers
WHERE trip = trips.id
AND ow > 0)) < 20 )
AND dates.from_date > 2458553;
请注意,我使用显式JOIN
运算符替换了WHERE子句中的古老和脆弱的隐式连接。它对性能没有任何影响,它只是更好的编码风格。
我们可以尝试使用resrv_customers
表的连接重写您的查询,而不是使用昂贵的相关子查询:
WITH cte AS (
SELECT
trip,
COUNT(CASE WHEN ow > 0 THEN 1 END) AS cnt1,
COUNT(CASE WHEN extra_seat = 1 AND ow > 0 THEN 1 END) AS cnt2
FROM resrv_customers
GROUP BY trip
)
SELECT *
FROM trips t
INNER JOIN dates d
ON t.trip_date = d.id
LEFT JOIN cte r
ON r.trip = t.id
WHERE
places_number > 0 AND
places_number - (r.cnt1 + r.cnt2) < 20 AND
d.from_date > 2458553;
在这里索引可能很棘手,因为你正在做一个SELECT *
。要让Postgres使用任何索引,您可能需要进行大量的列覆盖,即创建大型索引。
试试这个:
with rc_count as
(
select trip, count(*) as ow_count, sum( (extra_seat = 1)::int ) as ow_count_with_extra_seat
from resrv_customers
where ow > 0
group by trip
)
SELECT *
FROM trips
join dates on dates.id = trips.trip_date
join rc_count rc on trips.id = trip
WHERE places_number_on_flag > 0
AND places_number > 0
AND dates.from_date > 2458553
and places_number - (ow_count + ow_count_with_extra_seat) < 20