postgresql - 为什么过滤时使用相同的字段会导致不同的执行时间？（不同的索引用法）

Question

当我运行查询并按它过滤agreement_id时很慢，
但是当我按别名过滤时id它很快。（查看查询末尾）

为什么过滤时使用相同的字段会导致不同的执行时间？

链接到explain analyze：

慢1，慢2，快
1，快2

差异从 #20 开始：使用不同索引的地方：
Index Cond: (o.sys_period @> sys_time())VSIndex Cond: (o.agreement_id = 38)

PS。如果我可以联系此功能的开发人员，那就太好了（我还有一个类似的问题）

UPD 我做了一些实验。当我从查询中删除窗口函数时，它在任何情况下都能快速运行。那么为什么窗口函数在某些情况下会停止使用索引呢？如何逃避/解决这个问题？

带有最小测试用例的dbfiddle
服务器版本为 v13.1
完整查询：

WITH gconf AS
  -- https://www.postgresql.org/docs/current/queries-with.html#QUERIES-WITH-SELECT
  NOT MATERIALIZED -- force it to be merged into the parent query
  -- it gives a net savings because each usage of the WITH query needs only a small part of the WITH query's full output.
( SELECT
  ocd.*, 
  tstzrange( '2021-05-01', '2021-05-01', '[]') AS acc_period,

  (o).agreement_id      AS id,       -- Required to passthrough WINDOW FUNCTION
  (o).id                AS order_id,
  (ic).consumed_period  AS consumed_period,
  dense_rank()  OVER ( PARTITION BY (o).agreement_id, (o).id ORDER BY (ic).consumed_period )  AS nconf,
  row_number()  OVER ( wconf ORDER BY (c).sort_order NULLS LAST                            )  AS nitem,

  (sum( ocd.item_cost )  OVER wconf)::numeric( 10, 2) AS conf_cost,
  max((ocd.ic).consumed) OVER wconf                   AS consumed,
  CASE WHEN true
    THEN (sum( ocd.item_suma )  OVER wconf)::numeric( 10, 2 )
    ELSE (sum( ocd.item_cost )  OVER wconf)::numeric( 10, 2 )
  END AS conf_suma
FROM order_cost_details( tstzrange( '2021-05-01', '2021-05-01', '[]') ) ocd
WHERE true  OR  (ocd.ic).consumed_period @> lower( tstzrange( '2021-05-01', '2021-05-01', '[]') )

WINDOW wconf AS ( PARTITION BY (o).agreement_id, (o).id, (ic).consumed_period )
),
gorder AS (
SELECT *,
  (conf_suma/6)::numeric( 10, 2 ) as conf_nds,
  sum( conf_suma ) FILTER (WHERE nitem = 1) OVER worder AS order_suma
FROM gconf
WINDOW worder AS ( PARTITION BY gconf.id, (o).id )
-- TODO: Ask PG developers: Why changing to (o).agreement_id slows down query?
-- WINDOW worder AS ( PARTITION BY (o).agreement_id, (o).id )
)
SELECT
  u.id, consumed_period, nconf, nitem,
  (c).id                                                as item_id,
  COALESCE( (c).sort_order,   pd.sort_order          )  as item_order,
  COALESCE( st.display, st.name, rt.display, rt.name )  as item_name,
  COALESCE( item_qty,         (c).amount/rt.unit     )  as item_qty,
  COALESCE( (p).label,        rt.label               )  as measure,
  item_price, item_cost, item_suma,
  conf_cost, consumed, conf_suma, conf_nds, order_suma,
  (order_suma/6)::numeric( 10, 2 )                      as order_nds,
  sum( conf_suma                        ) FILTER (WHERE nitem = 1                ) OVER wagreement  AS total_suma,
  sum( (order_suma/6)::numeric( 10, 2 ) ) FILTER (WHERE nitem = 1  AND  nconf = 1) OVER wagreement  AS total_nds,

  pkg.id                            as package_id,
  pkg.link_1c_id                    as package_1c_id,
  COALESCE( pkg.display, pkg.name ) as package,

  acc_period

FROM gorder u
LEFT JOIN resource_type rt ON rt.id  = (c).resource_type_id
LEFT JOIN service_type  st ON st.id  = (c).service_type_id
LEFT JOIN package      pkg ON pkg.id = (o).package_id
LEFT JOIN package_detail     pd  ON pd.package_id = (o).package_id
  AND pd.resource_type_id IS NOT DISTINCT FROM (c).resource_type_id
  AND pd.service_type_id  IS NOT DISTINCT FROM (c).service_type_id

-- WHERE (o).agreement_id = 38   --  slow
WHERE u.id = 38                  --  fast
WINDOW wagreement AS ( PARTITION BY (o).agreement_id )

标签： postgresqlquery-optimizationpostgresql-performancequery-planner

Answer 1

作为问题解决方法，我们还可以为表达式中SELECT使用的列添加别名。PARTITION BY然后 PG 应用优化并使用索引。

问题的答案可能是：如果使用复合类型， PG 不会应用优化。注意它的工作原理：

PARTITION | FILTER | IS USED?
------------------------------
ALIAS     | ORIG   | NO
ALIAS     | ALIAS  | YES
ORIG      | ALIAS  | NO
ORIG      | ORIG   | NO

看到这个 dbfiddle

create table agreement ( ag_id int, name text, cost numeric(10,2) );
create index ag_idx on agreement (ag_id);
insert into agreement (ag_id, name, cost) values ( 1, '333', 22 ),
(1,'333', 33), (1, '333', 7), (2, '555', 18 ), (2, '555', 2), (3, '777', 4);
select * from agreement;

create function initial () 
returns table( agreement_id int, ag agreement ) language sql stable AS $$
select ag_id, t from agreement t;
$$;
select * from initial() t;

explain( analyze, costs, buffers, verbose ) with totals_by_ag as (
  select 
    *,
    sum( (t.ag).cost ) over ( partition by agreement_id ) as total
  from initial() t
)
select * from totals_by_ag t
where (t.ag).ag_id = 1; -- index is NOT USED

explain( analyze, costs, buffers, verbose ) with totals_by_ag as (
  select 
    *,
    sum( (t.ag).cost ) over ( partition by agreement_id ) as total
  from initial() t
)
select * from totals_by_ag t
where agreement_id = 1; -- index is used when alias for column is used

explain( analyze, costs, buffers, verbose ) with totals_by_ag as (
  select 
    *,
    sum( (t.ag).cost ) over ( partition by (t.ag).ag_id ) as total --renamed
  from initial() t
)
select * from totals_by_ag t
where agreement_id = 1; -- index is NOT USED because grouping by original column

explain( analyze, costs, buffers, verbose ) with totals_by_ag as (
  select 
    *,
    sum( (t.ag).cost ) over ( partition by (t.ag).ag_id ) as total --renamed
  from initial() t
)
select * from totals_by_ag t
where (t.ag).ag_id = 1; -- index is NOT USED even if at both cases original column