这是我的Postgres查询：
select
date_trunc('month', member.member_since),
count(member_change.member_id)
from
member_change
left join
member on (member_change.member_id = member.member_id)
left join
import_stats on member_change.import_history_id = import_stats.import_id
where
member.account_id = 1050007 and
change_source = 'i' and
change_type = 'a' and
member_status_id = 'a' and
member_since >= '2015-08-01' and
member_change.import_history_id in (select import_id from import_stats where account_id = 1050007)
group by 1;
以下是对查询的解释分析
"HashAggregate (cost=43.50..43.55 rows=4 width=16) (actual time=2.727..2.728 rows=2 loops=1)"
" -> Nested Loop Semi Join (cost=18.00..43.48 rows=4 width=16) (actual time=2.089..2.717 rows=33 loops=1)"
" Join Filter: (public.member_change.import_history_id = public.import_stats.import_id)"
" Rows Removed by Join Filter: 79"
" -> Hash Right Join (cost=18.00..33.68 rows=4 width=24) (actual time=0.121..0.715 rows=33 loops=1)"
" Hash Cond: (public.import_stats.import_id = public.member_change.import_history_id)"
" -> Append (cost=0.00..14.10 rows=411 width=8) (actual time=0.004..0.584 rows=7 loops=1)"
" -> Seq Scan on import_stats (cost=0.00..0.00 rows=1 width=8) (actual time=0.000..0.000 rows=0 loops=1)"
" -> Seq Scan on import_stats (cost=0.00..14.10 rows=410 width=8) (actual time=0.004..0.583 rows=7 loops=1)"
" -> Hash (cost=17.97..17.97 rows=2 width=24) (actual time=0.108..0.108 rows=33 loops=1)"
" Buckets: 1024 Batches: 1 Memory Usage: 2kB"
" -> Hash Join (cost=15.72..17.97 rows=2 width=24) (actual time=0.077..0.098 rows=33 loops=1)"
" Hash Cond: (public.member.member_id = public.member_change.member_id)"
" -> Append (cost=0.00..2.08 rows=38 width=16) (actual time=0.024..0.038 rows=34 loops=1)"
" -> Seq Scan on member (cost=0.00..0.00 rows=1 width=16) (actual time=0.005..0.005 rows=0 loops=1)"
" Filter: ((member_since >= '2015-08-01 00:00:00-05'::timestamp with time zone) AND (account_id = 1050007) AND ((member_status_id)::text = 'a'::text))"
" -> Seq Scan on member (cost=0.00..2.08 rows=37 width=16) (actual time=0.019..0.030 rows=34 loops=1)"
" Filter: ((member_since >= '2015-08-01 00:00:00-05'::timestamp with time zone) AND (account_id = 1050007) AND ((member_status_id)::text = 'a'::text))"
" Rows Removed by Filter: 28"
" -> Hash (cost=15.70..15.70 rows=2 width=16) (actual time=0.048..0.048 rows=61 loops=1)"
" Buckets: 1024 Batches: 1 Memory Usage: 3kB"
" -> Append (cost=0.00..15.70 rows=2 width=16) (actual time=0.007..0.026 rows=61 loops=1)"
" -> Seq Scan on member_change (cost=0.00..0.00 rows=1 width=16) (actual time=0.000..0.000 rows=0 loops=1)"
" Filter: ((change_source = 'i'::text) AND (change_type = 'a'::text))"
" -> Seq Scan on member_change (cost=0.00..15.70 rows=1 width=16) (actual time=0.006..0.022 rows=61 loops=1)"
" Filter: ((change_source = 'i'::text) AND (change_type = 'a'::text))"
" Rows Removed by Filter: 27"
" -> Materialize (cost=0.00..9.62 rows=3 width=8) (actual time=0.059..0.060 rows=3 loops=33)"
" -> Append (cost=0.00..9.61 rows=3 width=8) (actual time=0.030..0.030 rows=4 loops=1)"
" -> Seq Scan on import_stats (cost=0.00..0.00 rows=1 width=8) (actual time=0.001..0.001 rows=0 loops=1)"
" Filter: (account_id = 1050007)"
" -> Bitmap Heap Scan on import_stats (cost=4.27..9.61 rows=2 width=8) (actual time=0.027..0.027 rows=4 loops=1)"
" Recheck Cond: (account_id = 1050007)"
" -> Bitmap Index Scan on ix_import_stats_account_id (cost=0.00..4.27 rows=2 width=0) (actual time=0.019..0.019 rows=4 loops=1)"
" Index Cond: (account_id = 1050007)"
"Total runtime: 3.447 ms"
基本上我只是想在change_source='i'处添加成员的汇总数量。 member_change表非常庞大（超过一百万条记录并计算在内）但我无法让查询得到1秒以下