域名預(yù)訂/競價(jià)，好“米”不錯(cuò)過

對于龐大的數(shù)據(jù)，檢索sql的編寫要格外小心，有很多平時(shí)不注意的sql可能就會變成瓶頸。

比如, 我們有個(gè)系統(tǒng), 其中t96_pd_log表，記錄數(shù)8000w多，在開發(fā)階段乃至用了那么多年都沒問題, 最近卻發(fā)生頻繁死鎖的問題, 查數(shù)據(jù)庫后臺發(fā)現(xiàn)問題出在一個(gè)select語句上, 它耗時(shí)高達(dá)2.4-2.7s，這對于一個(gè)需要高并發(fā)的系統(tǒng)來說當(dāng)然是致命的。

數(shù)據(jù)表t96_pd_log有兩條index， 一條的字段組成是f96_mgtbarcd,另一條的字段組成是f96_result_type， 檢索sql是這樣寫的：

select recseq,f96_create_dt,f96_op from t96_pd_log where f96_mgtbarcd='113D1907032385'
and f96_station='AS01-L113' and f96_result_type='TFT'
and f96_qty=1 order by f96_create_dt desc limit 1

意在找出AS01-L113站位最近一條有效的記錄，而這條sql有足夠的index支持，但耗時(shí)高達(dá)2.7s。

F7查一下，居然包含f96_result_type的這支索引也參與運(yùn)算， 這就出事了， 因?yàn)閒96_result_type有相同值的記錄極其多，從圖3也可以看出， 這也是主要耗費(fèi)時(shí)間的環(huán)節(jié)。

通過數(shù)據(jù)分析， f96_mgtbarcd值相同的記錄數(shù)很少， 所以我們可以將順序改一下，先用f96_mgtbarcd作為過濾器生成一個(gè)子集后， 再從這個(gè)子集里面做f96_result_type等過濾， 跑了一下，13ms, 足足快了這樣就快200多倍，如下：

with a as (
select recseq,f96_create_dt,f96_op,f96_station,f96_result_type,f96_qty from t96_pd_log where f96_mgtbarcd='113D1907032385'
)
select recseq,f96_create_dt,f96_op from a where f96_station='AS01-L113' and f96_result_type='TP' and f96_qty=1 order by f96_create_dt desc limit 1

ps：我用的工具是 pgAdmin自帶的，F(xiàn)7, Shift-F7

補(bǔ)充：PostgreSql查詢優(yōu)化之根據(jù)執(zhí)行計(jì)劃優(yōu)化SQL

1、執(zhí)行計(jì)劃路徑選擇

postgresql查詢規(guī)劃過程中，查詢請求的不同執(zhí)行方案是通過建立不同的路徑來表達(dá)的，在生成許多符合條件的路徑之后，要從中選擇出代價(jià)最小的路徑（基于成本運(yùn)算），把它轉(zhuǎn)化為一個(gè)計(jì)劃，傳遞給執(zhí)行器執(zhí)行，規(guī)劃器的核心工作就是生成多條路徑，然后從中找出最優(yōu)的那一條。

1.1代價(jià)評估

評估路徑優(yōu)劣的依據(jù)是用系統(tǒng)表pg_statistic中的統(tǒng)計(jì)信息估算出來的不同路徑的代價(jià)(cost)，PostgreSQL估計(jì)計(jì)劃成本的方式：基于統(tǒng)計(jì)信息估計(jì)計(jì)劃中各個(gè)節(jié)點(diǎn)的成本。PostgreSQL會分析各個(gè)表來獲取一個(gè)統(tǒng)計(jì)信息樣本（這個(gè)操作通常是由autovacuum這個(gè)守護(hù)進(jìn)程周期性的執(zhí)行analyze，來收集這些統(tǒng)計(jì)信息，然后保存到pg_statistic和pg_class里面）。

1.2用于估算代價(jià)的參數(shù)postgresql.conf

# - Planner Cost Constants -
#seq_page_cost = 1.0  # measured on an arbitrary scale 順序磁盤掃描時(shí)單個(gè)頁面的開銷
#random_page_cost = 4.0  # same scale as above 隨機(jī)磁盤訪問時(shí)單頁面的讀取開銷
#cpu_tuple_cost = 0.01  # same scale as above cpu處理每一行的開銷
#cpu_index_tuple_cost = 0.005 # same scale as above cpu處理每個(gè)索引行的開銷
#cpu_operator_cost = 0.0025 # same scale as above cpu處理每個(gè)運(yùn)算符或者函數(shù)調(diào)用的開銷
#parallel_tuple_cost = 0.1 # same scale as above 計(jì)算并行處理的成本，如果成本高于非并行，則不會開啟并行處理。
#parallel_setup_cost = 1000.0 # same scale as above
#min_parallel_relation_size = 8MB
#effective_cache_size = 4GB 再一次索引掃描中可用的文件系統(tǒng)內(nèi)核緩沖區(qū)有效大小

也可以使用 show all的方式查看

1.3 路徑的選擇

--查看表信息
highgo=# \d t_jcxxgl_tjaj
  Table "db_jcxx.t_jcxxgl_tjaj"
 
 Column |  Type  | Modifiers --------------+--------------------------------+-----------
 c_bh | character(32)   | not null
 c_xzdm | character varying(300)  |
 c_jgid | character(32)   |
 c_ajbm | character(22)   |
...
Indexes:
 "t_jcxxgl_tjaj_pkey" PRIMARY KEY, btree (c_bh)
 "idx_ttjaj_cah" btree (c_ah)
 "idx_ttjaj_dslrq" btree (d_slrq)

首先更新統(tǒng)計(jì)信息vacuum analyze t_jcxxgl_tjaj，許多時(shí)候可能因?yàn)榻y(tǒng)計(jì)信息的不準(zhǔn)確導(dǎo)致了不正常的執(zhí)行計(jì)劃--執(zhí)行計(jì)劃。

--執(zhí)行計(jì)劃，全表掃描
highgo=# explain (analyze,verbose,costs,buffers,timing)select c_bh,c_xzdm,c_jgid,c_ajbm from t_jcxxgl_tjaj where d_slrq >='2018-03-18';
       QUERY PLAN      ------------------------------------------------------------------------------------------------------------
 Seq Scan on db_jcxx.t_jcxxgl_tjaj (cost=0.00..9.76 rows=3 width=96) (actual time=1.031..1.055 rows=3 loops
=1)
 Output: c_bh, c_xzdm, c_jgid, c_ajbm
 Filter: (t_jcxxgl_tjaj.d_slrq >= '2018-03-18'::date)
 Rows Removed by Filter: 138
 Buffers: shared hit=8
 Planning time: 6.579 ms
 Execution time: 1.163 ms
(7 rows)

如上，d_slrq是有索引的，但是執(zhí)行計(jì)劃中并沒有走索引，為什么呢？我們繼續(xù)往下看。

--執(zhí)行計(jì)劃，關(guān)閉全表掃描
highgo=# set session enable_seqscan = off;
SET
highgo=# explain (analyze,verbose,costs,buffers,timing)select c_bh,c_xzdm,c_jgid,c_ajbm from t_jcxxgl_tjaj where d_slrq >='2018-03-18';
        QUERY PLAN        ------------------------------------------------------------------------------------------------------------
 
 Index Scan using idx_ttjaj_dslrq on db_jcxx.t_jcxxgl_tjaj (cost=0.14..13.90 rows=3 width=96) (actual time=0.012..0.026 rows=3 loops=1)
 Output: c_bh, c_xzdm, c_jgid, c_ajbm
 Index Cond: (t_jcxxgl_tjaj.d_slrq >= '2018-03-18'::date)
 Buffers: shared hit=4
 Planning time: 0.309 ms
 Execution time: 0.063 ms
(6 rows)

d_slrq上面有btree索引，但是查看執(zhí)行計(jì)劃并沒有走索引，這是為什么呢？

代價(jià)計(jì)算：

一個(gè)路徑的估算由三部分組成：啟動代價(jià)(startup cost)，總代價(jià)(totalcost)，執(zhí)行結(jié)果的排序方式(pathkeys)

代價(jià)估算公式：

總代價(jià)=啟動代價(jià)+I/O代價(jià)+CPU代價(jià)（cost=S+P+W*T）

P:執(zhí)行時(shí)要訪問的頁面數(shù)，反應(yīng)磁盤的I/O次數(shù)

T:表示在執(zhí)行時(shí)所要訪問的元組數(shù)，反映了cpu開銷

W:表示磁盤I/O代價(jià)和CPU開銷建的權(quán)重因子

統(tǒng)計(jì)信息：

統(tǒng)計(jì)信息的其中一部分是每個(gè)表和索引中項(xiàng)的總數(shù)，以及每個(gè)表和索引占用的磁盤塊數(shù)。這些信息保存在pg_class表的reltuples和relpages列中。我們可以這樣查詢相關(guān)信息：

--查看統(tǒng)計(jì)信息
highgo=# select relpages,reltuples from pg_class where relname ='t_jcxxgl_tjaj';
 relpages | reltuples ----------+-----------
 8 | 141
(1 row)

total_cost = 1(seq_page_cost)*8(磁盤總頁數(shù))+0.01(cpu_tuple_cost)*141(表的總記錄數(shù))+0.0025(cpu_operation_cost)*141(表的總記錄數(shù))=9.7625

可以看到走索引的cost=13.90比全表掃描cost=9.76要大。所以上面沒有關(guān)閉全表掃描的時(shí)候，根據(jù)成本代價(jià)，執(zhí)行計(jì)劃走的全表掃描。在表較小的情況下，全表掃描比索引掃描更有效， index scan 至少要發(fā)生兩次I/O，一次是讀取索引塊，一次是讀取數(shù)據(jù)塊。

2、一個(gè)SQL優(yōu)化實(shí)例

2.1慢SQL：

1select c_ajbh, c_ah, c_cbfy, c_cbrxm, d_larq, d_jarq, n_dbjg, c_yqly from db_zxzhld.t_zhld_db dbxx join db_zxzhld.t_zhld_ajdbxx dbaj on dbxx.c_bh = dbaj.c_dbbh where dbxx.n_valid=1 and dbxx.n_state in (1,2,3) and dbxx.c_dbztbh='1003' and dbaj.c_zblx='1003' and dbaj.c_dbfy='0' and dbaj.c_gy = '2550' and c_ajbh in (select distinct c_ajbh from db_zxzhld.t_zhld_zbajxx where n_dbzt = 1 and c_zblx = '1003' and c_gy = '2550' ) order by d_larq asc, c_ajbh asc limit 15 offset 0;

慢sql耗時(shí)：7s

先過下這個(gè)sql是干什么的、首先dbxx和dbaj的一個(gè)join連接然后dbaj.c_ajbh要包含在zbaj表里面，做了個(gè)排序，取了15條記錄、大概就這樣。

Sql有個(gè)缺點(diǎn)就是我不知道查詢的字段是從那個(gè)表里面取的、建議加上表別名.字段。

查看該sql的表的數(shù)據(jù)量：

t_zhld_db ：1311

t_zhld_ajdbxx ：341296

t_zhld_zbajxx ：1027619

執(zhí)行計(jì)劃：

01 Limit (cost=36328.67..36328.68 rows=1 width=107) (actual time=88957.677..88957.729 rows=15 loops=1)
 
02 -> Sort (cost=36328.67..36328.68 rows=1 width=107) (actual time=88957.653..88957.672 rows=15 loops=1)
 
03  Sort Key: dbaj.d_larq, dbaj.c_ajbh
 
04  Sort Method: top-N heapsort Memory: 27kB
 
05  -> Nested Loop Semi Join (cost=17099.76..36328.66 rows=1 width=107) (actual time=277.794..88932.662 rows=8605 loops=1)
 
06  Join Filter: ((dbaj.c_ajbh)::text = (t_zhld_zbajxx.c_ajbh)::text)
 
07  Rows Removed by Join Filter: 37018710
 
08  -> Nested Loop (cost=0.00..19200.59 rows=1 width=107) (actual time=199.141..601.845 rows=8605 loops=1)
 
09   Join Filter: (dbxx.c_bh = dbaj.c_dbbh)
 
10   Rows Removed by Join Filter: 111865
 
11   -> Seq Scan on t_zhld_ajdbxx dbaj (cost=0.00..19117.70 rows=219 width=140) (actual time=198.871..266.182 rows=8605 loops=1)
 
12    Filter: ((n_valid = 1) AND ((c_zblx)::text = '1003'::text) AND ((c_dbfy)::text = '0'::text) AND ((c_gy)::text = '2550'::text))
 
13    Rows Removed by Filter: 332691
 
14   -> Materialize (cost=0.00..66.48 rows=5 width=33) (actual time=0.001..0.017 rows=14 loops=8605)
 
15    -> Seq Scan on t_zhld_db dbxx (cost=0.00..66.45 rows=5 width=33) (actual time=0.044..0.722 rows=14 loops=1)
 
16     Filter: ((n_valid = 1) AND ((c_dbztbh)::text = '1003'::text) AND (n_state = ANY ('{1,2,3}'::integer[])))
 
17     Rows Removed by Filter: 1297
 
18  -> Materialize (cost=17099.76..17117.46 rows=708 width=32) (actual time=0.006..4.890 rows=4303 loops=8605)
 
19   -> HashAggregate (cost=17099.76..17106.84 rows=708 width=32) (actual time=44.011..54.924 rows=8605 loops=1)
 
20    Group Key: t_zhld_zbajxx.c_ajbh
 
21    -> Bitmap Heap Scan on t_zhld_zbajxx (cost=163.36..17097.99 rows=708 width=32) (actual time=5.218..30.278 rows=8605 loops=1)
 
22     Recheck Cond: ((n_dbzt = 1) AND ((c_zblx)::text = '1003'::text))
 
23     Filter: ((c_gy)::text = '2550'::text)
 
24     Rows Removed by Filter: 21849
 
25     Heap Blocks: exact=960
 
26     -> Bitmap Index Scan on i_tzhldzbajxx_zblx_dbzt (cost=0.00..163.19 rows=5876 width=0) (actual time=5.011..5.011 rows=30458 loops=1)
 
27     Index Cond: ((n_dbzt = 1) AND ((c_zblx)::text = '1003'::text))
 
28 Planning time: 1.258 ms
29 Execution time: 88958.029 ms

執(zhí)行計(jì)劃解讀：

1：第27->21行，通過索引i_tzhldzbajxx_zblx_dbzt過濾表t_zhld_zbajxx的數(shù)據(jù)，然后根據(jù)過濾條件(c_gy)::text = '2550'::text過濾最終返回8605條數(shù)據(jù)

2：第17->15行，根據(jù)條件過濾t_zhld_db表的數(shù)據(jù)，最終返回了14條數(shù)據(jù)

3：第20->19行，對表t_zhld_zbajxx做group by的操作

4：第13->11行，全表掃描t_zhld_ajdbxx 最終返回了8605條數(shù)據(jù)

5：第08行，根據(jù)t_zhld_ajdbxx返回的8605條結(jié)果集作為驅(qū)動表和t_zhld_db的結(jié)果集(14條)做嵌套循環(huán)，t_zhld_db的結(jié)果集被循環(huán)了8605次。然后過濾掉了其中的111865條記錄，那么最終將得到（8605*14-111865） = 8605

6：第07->05行，根據(jù)第08和18行返回的結(jié)果集最終做了Nested Loop Semi Join，第18行的4303條結(jié)果集被循環(huán)了8605次,(4303*8605-37018710)=8605

7: 第04->02行，對最終的8605條記錄進(jìn)行排序

8：第01行，limit最終獲取15條記錄

整個(gè)執(zhí)行計(jì)劃中耗時(shí)最長的地方在05行Nested Loop Semi Join，actual time=277.794..88932.662， 表db_zxzhld.t_zhld_db dbxx和db_zxzhld.t_zhld_ajdbxx均是全表掃描

2.2具體優(yōu)化步驟

查看索引頁并沒有索引，創(chuàng)建c_ajbh，c_dbbh等邏輯外鍵的索引

drop index if exists I_T_ZHLD_AJDBXX_AJBH;
create index I_T_ZHLD_AJDBXX_AJBH on T_ZHLD_AJDBXX (c_ajbh);
commit;
drop index if exists I_T_ZHLD_AJDBXX_DBBH;
create index I_T_ZHLD_AJDBXX_DBBH on T_ZHLD_AJDBXX (c_dbbh);
commit;

創(chuàng)建d_larq,c_ajbh的排序索引：

drop index if exists I_T_ZHLD_AJDBXX_m6;create index I_T_ZHLD_AJDBXX_m6 on T_ZHLD_AJDBXX (c_zblx,c_dbfy,c_gy,d_larq asc,c_ajbh asc);
commit;
drop index if exists I_T_ZHLD_ZBAJXX_h3 ;
create index I_T_ZHLD_ZBAJXX_h3 on db_zxzhld.t_zhld_zbajxx (n_dbzt,c_zblx,c_gy,c_gy);
commit;

創(chuàng)建索引后執(zhí)行計(jì)劃有了改變，原來的dbaj表和dbxx表先做nestedloop變成了zbaj和dbaj表先做了nestedloop join,總的cost也從36328.68降到了12802.87，

執(zhí)行計(jì)劃

Limit (cost=12802.87..12802.87 rows=1 width=107) (actual time=4263.598..4263.648 rows=15 loops=1)
 -> Sort (cost=12802.87..12802.87 rows=1 width=107) (actual time=4263.592..4263.609 rows=15 loops=1)
 Sort Key: dbaj.d_larq, dbaj.c_ajbh
 Sort Method: top-N heapsort Memory: 27kB
 -> Nested Loop (cost=2516.05..12802.86 rows=1 width=107) (actual time=74.240..4239.723 rows=8605 loops=1)
  Join Filter: (dbaj.c_dbbh = dbxx.c_bh)
  Rows Removed by Join Filter: 111865
  -> Nested Loop (cost=2516.05..12736.34 rows=1 width=140) (actual time=74.083..327.974 rows=8605 loops=1)
   -> HashAggregate (cost=2515.62..2522.76 rows=714 width=32) (actual time=74.025..90.185 rows=8605 loops=1)
    Group Key: ("ANY_subquery".c_ajbh)::text
    -> Subquery Scan on "ANY_subquery" (cost=2499.56..2513.84 rows=714 width=32) (actual time=28.782..59.823 rows=8605 loops=1)
    -> HashAggregate (cost=2499.56..2506.70 rows=714 width=32) (actual time=28.778..39.968 rows=8605 loops=1)
     Group Key: zbaj.c_ajbh
     -> Index Scan using i_t_zhld_zbajxx_h3 on t_zhld_zbajxx zbaj (cost=0.42..2497.77 rows=715 width=32) (actual time=0.062..15.104 rows=8605 loops=1)
      Index Cond: ((n_dbzt = 1) AND ((c_zblx)::text = '1003'::text) AND ((c_gy)::text = '2550'::text))
   -> Index Scan using i_t_zhld_ajdbxx_ajbh on t_zhld_ajdbxx dbaj (cost=0.42..14.29 rows=1 width=140) (actual time=0.015..0.021 rows=1 loops=8605)
    Index Cond: ((c_ajbh)::text = ("ANY_subquery".c_ajbh)::text)
    Filter: (((c_zblx)::text = '1003'::text) AND ((c_dbfy)::text = '0'::text) AND ((c_gy)::text = '2550'::text))
    Rows Removed by Filter: 1
  -> Seq Scan on t_zhld_db dbxx (cost=0.00..66.45 rows=5 width=33) (actual time=0.015..0.430 rows=14 loops=8605)
   Filter: ((n_valid = 1) AND ((c_dbztbh)::text = '1003'::text) AND (n_state = ANY ('{1,2,3}'::integer[])))
   Rows Removed by Filter: 1298
Planning time: 1.075 ms
Execution time: 4263.803 ms

執(zhí)行的時(shí)間還是要4s左右仍然不滿足需求，并且沒有使用上I_T_ZHLD_AJDBXX_m6這個(gè)索引。

2.3等價(jià)改寫SQL(1)

等價(jià)改寫：將排序條件加入db_zxzhld.t_zhld_ajdbxx讓其先排序，再和t_zhld_db表連接。

修改后sql：

Select dbaj.c_ajbh, dbaj.c_ah, dbaj.c_cbfy, dbaj.c_cbrxm, dbaj.d_larq, dbaj.d_jarq, dbaj.n_dbjg, dbaj.c_yqly from (select * from db_zxzhld.t_zhld_db where n_valid=1 and n_state in (1,2,3) and c_dbztbh='1003' )dbxx
 join (select * from db_zxzhld.t_zhld_ajdbxx where n_valid=1 and c_zblx='1003'
 and c_dbfy='0' and c_gy = '2550' and
c_ajbh in (select distinct c_ajbh from db_zxzhld.t_zhld_zbajxx where n_dbzt = 1 and c_zblx = '1003' and c_gy = '2550' ) order by d_larq asc, c_ajbh asc)dbajon dbxx.c_bh = dbaj.c_dbbh
 limit 15 offset 0

再次查看執(zhí)行計(jì)劃：

Limit (cost=3223.92..3231.97 rows=1 width=107) (actual time=127.291..127.536 rows=15 loops=1)
 
 -> Nested Loop (cost=3223.92..3231.97 rows=1 width=107) (actual time=127.285..127.496 rows=15 loops=1)
 
 -> Sort (cost=3223.64..3223.65 rows=1 width=140) (actual time=127.210..127.225 rows=15 loops=1)
 
  Sort Key: t_zhld_ajdbxx.d_larq, t_zhld_ajdbxx.c_ajbh
 
  Sort Method: quicksort Memory: 2618kB
 
  -> Hash Semi Join (cost=2523.19..3223.63 rows=1 width=140) (actual time=55.913..107.265 rows=8605 loops=1)
 
   Hash Cond: ((t_zhld_ajdbxx.c_ajbh)::text = (t_zhld_zbajxx.c_ajbh)::text)
 
   -> Index Scan using i_t_zhld_ajdbxx_m6 on t_zhld_ajdbxx (cost=0.42..700.28 rows=219 width=140) (actual time=0.065..22.005 rows=8605 loops=1)
 
    Index Cond: (((c_zblx)::text = '1003'::text) AND ((c_dbfy)::text = '0'::text) AND ((c_gy)::text = '2550'::text))
 
   -> Hash (cost=2513.84..2513.84 rows=714 width=32) (actual time=55.802..55.802 rows=8605 loops=1)
 
    Buckets: 16384 (originally 1024) Batches: 1 (originally 1) Memory Usage: 675kB
 
    -> HashAggregate (cost=2499.56..2506.70 rows=714 width=32) (actual time=30.530..43.275 rows=8605 loops=1)
 
    Group Key: t_zhld_zbajxx.c_ajbh
 
    -> Index Scan using i_t_zhld_zbajxx_h3 on t_zhld_zbajxx (cost=0.42..2497.77 rows=715 width=32) (actual time=0.043..15.552 rows=8605 loops=1)
 
     Index Cond: ((n_dbzt = 1) AND ((c_zblx)::text = '1003'::text) AND ((c_gy)::text = '2550'::text))
 
 -> Index Scan using t_zhld_db_pkey on t_zhld_db (cost=0.28..8.30 rows=1 width=33) (actual time=0.009..0.011 rows=1 loops=15)
 
  Index Cond: (c_bh = t_zhld_ajdbxx.c_dbbh)
  Filter: (((c_dbztbh)::text = '1003'::text) AND (n_state = ANY ('{1,2,3}'::integer[])))
Planning time: 1.154 ms
Execution time: 127.734 ms

這一次可以看出，ajdbxx和zbajxx表做了hash semi join 消除了nestedloop,cost降到了3231.97。并且使用上了i_t_zhld_ajdbxx_m6子查詢中in的結(jié)果集有一萬多條數(shù)據(jù)。

繼續(xù)嘗試使用exists等價(jià)改寫in，看能否有更好的結(jié)果

2.4等價(jià)改寫SQL(2)

等價(jià)改寫：將in替換為exists：

select c_ajbh, c_ah, c_cbfy, c_cbrxm, d_larq, d_jarq, n_dbjg, c_yqlyfrom (select c_bh from db_zxzhld.t_zhld_db where n_state in (1,2,3) and c_dbztbh='1003' )dbxx
 join (select c_ajbh, c_ah, c_cbfy, c_cbrxm, d_larq, d_jarq, n_dbjg, c_yqly,c_dbbh from db_zxzhld.t_zhld_ajdbxx ajdbxxwhere c_zblx='1003'
 and c_dbfy='0' and c_gy = '2550' and
exists (select distinct c_ajbh from db_zxzhld.t_zhld_zbajxx zbajxx where ajdbxx.c_ajbh = zbajxx.c_ajbh and n_dbzt = 1 and c_zblx = '1003' and c_gy = '2550' ) order by d_larq asc, c_ajbh asc)dbajon dbxx.c_bh = dbaj.c_dbbh
 limit 15 offset 0

再次查看執(zhí)行計(jì)劃：

Limit (cost=1.12..2547.17 rows=1 width=107) (actual time=0.140..0.727 rows=15 loops=1)
 -> Nested Loop (cost=1.12..2547.17 rows=1 width=107) (actual time=0.136..0.689 rows=15 loops=1)
 -> Nested Loop Semi Join (cost=0.85..2538.84 rows=1 width=140) (actual time=0.115..0.493 rows=15 loops=1)
  -> Index Scan using i_t_zhld_ajdbxx_m6 on t_zhld_ajdbxx t2 (cost=0.42..700.28 rows=219 width=140) (actual time=0.076..0.127 rows=15 loops=1)
   Index Cond: (((c_zblx)::text = '1003'::text) AND ((c_dbfy)::text = '0'::text) AND ((c_gy)::text = '2550'::text))
  -> Index Scan using i_t_zhld_zbajxx_c_ajbh on t_zhld_zbajxx t3 (cost=0.42..8.40 rows=1 width=32) (actual time=0.019..0.019 rows=1 loops=15)
   Index Cond: ((c_ajbh)::text = (t2.c_ajbh)::text)
   Filter: (((c_zblx)::text = '1003'::text) AND ((c_gy)::text = '2550'::text) AND (n_dbzt = 1))
 -> Index Scan using t_zhld_db_pkey on t_zhld_db (cost=0.28..8.30 rows=1 width=33) (actual time=0.007..0.008 rows=1 loops=15)
  Index Cond: (c_bh = t2.c_dbbh)
  Filter: (((c_dbztbh)::text = '1003'::text) AND (n_state = ANY ('{1,2,3}'::integer[])))
Planning time: 1.268 ms
Execution time: 0.859 ms

可以看出使用exist效果更好，最終cost 2547.17

(1).少了t_zhld_zbajxx表的group by操作：Sort Key: t_zhld_ajdbxx.d_larq, t_zhld_ajdbxx.c_ajbh。(這一步是因?yàn)槭褂昧怂饕械呐判?

(2).少了分組的操作：Group Key: t_zhld_zbajxx.c_ajbh。

第(2)為什么這個(gè)查詢消除了t_zhld_zbajxx表的group by操作呢？

原因是exists替換了distinct的功能，一旦滿足條件則立刻返回。所以使用exists的時(shí)候子查詢可以直接去掉distinct。

文章來源：腳本之家

來源地址：https://www.jb51.net/article/204776.htm

申請創(chuàng)業(yè)報(bào)道，分享創(chuàng)業(yè)好點(diǎn)子。點(diǎn)擊此處，共同探討創(chuàng)業(yè)新機(jī)遇！