巧用case when做SQL優化

阿新 • • 發佈：2019-02-05

下面是一張報表的SQL實現，不用深入業務，你可以看到上面的4條SQL和下面的4條SQL其實大致相同，只是部分條件不一樣，想要優化的第一感覺是如何減少掃描的次數，這個時候case when就派上用場了。

SELECT DECODE(ALL_COUNT, 0, 0, ROUND(ARRIVED_COUNT / ALL_COUNT, 4) * 100) ARRIVED_RATE
  FROM (SELECT E.COUNT_RESULT + F.COUNT_RESULT + G.COUNT_RESULT +
               H.COUNT_RESULT ARRIVED_COUNT
          FROM (SELECT COUNT(1) COUNT_RESULT
                  FROM GG_TASK_WORK_SHEET          PTWS,
                       GG_TASK_PATROL_SHEET_DETAIL PSD,
                       GG_PATROL_WORK_POINT        PPWP,
                       GG_PATROL_POINT             PPP,
                       GG_PATROL_EQUIPMENT         PPE
                 WHERE PTWS.WORK_SHEET_ID = PSD.WORK_SHEET_ID
                   AND TO_CHAR(PTWS.WORK_SHEET_ID) = PPWP.WORK_ID
                   AND PPWP.PATROL_POINT_ID = PPP.PATROL_POINT_ID
                   AND PPP.PATROL_POINT_ID = PPE.PATROL_POINT_ID
                   AND PTWS.TASK_TYPE = 4
                   AND PTWS.SHEET_TYPE = 2
                   AND PSD.PATROL_TYPE <> -1
                   AND PSD.IS_NEED_PDA = '1'
                   AND PPWP.PATROL_FLAG = 1
                   AND PTWS.IS_COMPLETE = '3'
                   AND TO_CHAR(PTWS.REAL_END_DATE, 'YYYY') = '2013'
                   AND PPE.DEVICE_ID = 1003551930) E,
               (SELECT COUNT(1) COUNT_RESULT
                  FROM GG_TASK_WORK_SHEET          PTWS,
                       GG_TASK_PATROL_SHEET_DETAIL PSD,
                       GG_PATROL_WORK_POINT        PPWP,
                       GG_PATROL_POINT             PPP,
                       GG_TASK_PATROL_OBJECT       PTPO
                 WHERE PTWS.WORK_SHEET_ID = PSD.WORK_SHEET_ID
                   AND TO_CHAR(PTWS.WORK_SHEET_ID) = PPWP.WORK_ID
                   AND PPWP.PATROL_POINT_ID = PPP.PATROL_POINT_ID
                   AND PPP.PATROL_POINT_ID = PTPO.PATROL_POINT_ID
                   AND PTWS.TASK_TYPE = 4
                   AND PTWS.SHEET_TYPE = 2
                   AND PSD.PATROL_TYPE <> -1
                   AND PSD.IS_NEED_PDA = '1'
                   AND PPWP.PATROL_FLAG = 1
                   AND PTWS.IS_COMPLETE = '3'
                   AND TO_CHAR(PTWS.REAL_END_DATE, 'YYYY') = '2013'
                   AND PTPO.DEVICE_ID = 1003551930) F,
               (SELECT COUNT(1) COUNT_RESULT
                  FROM GG_TASK_WORK_SHEET          PTWS,
                       GG_TASK_PATROL_SHEET_DETAIL PSD,
                       GG_TASK_PATROL_OBJECT       PTPO,
                       GG_PATROL_EQUIPMENT         PPE
                 WHERE PTWS.WORK_SHEET_ID = PSD.WORK_SHEET_ID
                   AND PTWS.WORK_SHEET_ID = PTPO.WORK_SHEET_ID
                   AND PTPO.PATROL_LINE_ID = PPE.PATROL_LINE_ID
                   AND PTPO.FUNCTION_LOCATION_ID = PPE.FUNCTION_LOCATION_ID
                   AND PTWS.TASK_TYPE = 4
                   AND PTWS.SHEET_TYPE = 1
                   AND PTPO.PATROL_TYPE = 1
                   AND PSD.PATROL_TYPE = -1
                   AND PSD.IS_NEED_PDA = '1'
                   AND PTPO.ARRIVE_TYPE = 1
                   AND PTWS.IS_COMPLETE = '3'
                   AND TO_CHAR(PTWS.REAL_END_DATE, 'YYYY') = '2013'
                   AND PPE.DEVICE_ID = 1003551930) G,
               (SELECT COUNT(1) COUNT_RESULT
                  FROM GG_TASK_WORK_SHEET          PTWS,
                       GG_TASK_PATROL_SHEET_DETAIL PSD,
                       GG_TASK_PATROL_OBJECT       PTPO,
                       GG_PATROL_EQUIPMENT         PPE
                 WHERE PTWS.WORK_SHEET_ID = PSD.WORK_SHEET_ID
                   AND PTWS.WORK_SHEET_ID = PTPO.WORK_SHEET_ID
                   AND PTPO.PATROL_LINE_ID = PPE.PATROL_LINE_ID
                   AND PTPO.FUNCTION_LOCATION_ID = PPE.FUNCTION_LOCATION_ID
                   AND PTWS.TASK_TYPE = 4
                   AND PTWS.SHEET_TYPE = 1
                   AND PTPO.PATROL_TYPE = 1
                   AND PSD.PATROL_TYPE <> -1
                   AND PSD.IS_NEED_PDA = '1'
                   AND PTPO.ARRIVE_TYPE = 1
                   AND PTWS.IS_COMPLETE = '3'
                   AND TO_CHAR(PTWS.REAL_END_DATE, 'YYYY') = '2013'
                   AND PPE.DEVICE_ID = 1003551930) H) ARRIVED_DEVICE,
       (SELECT A.COUNT_RESULT + B.COUNT_RESULT + C.COUNT_RESULT +
               D.COUNT_RESULT ALL_COUNT
          FROM (SELECT COUNT(1) COUNT_RESULT
                  FROM GG_TASK_WORK_SHEET          PTWS,
                       GG_TASK_PATROL_SHEET_DETAIL PSD,
                       GG_PATROL_WORK_POINT        PPWP,
                       GG_PATROL_POINT             PPP,
                       GG_PATROL_EQUIPMENT         PPE
                 WHERE PTWS.WORK_SHEET_ID = PSD.WORK_SHEET_ID
                   AND TO_CHAR(PTWS.WORK_SHEET_ID) = PPWP.WORK_ID
                   AND PPWP.PATROL_POINT_ID = PPP.PATROL_POINT_ID
                   AND PPP.PATROL_POINT_ID = PPE.PATROL_POINT_ID
                   AND PTWS.TASK_TYPE = 4
                   AND PTWS.SHEET_TYPE = 2
                   AND PSD.PATROL_TYPE = -1
                   AND PSD.IS_NEED_PDA = '1'
                   AND PTWS.IS_COMPLETE = '3'
                   AND TO_CHAR(PTWS.REAL_END_DATE, 'YYYY') = '2013'
                   AND PPE.DEVICE_ID = 1003551930) A,
               (SELECT COUNT(1) COUNT_RESULT
                  FROM GG_TASK_WORK_SHEET          PTWS,
                       GG_TASK_PATROL_SHEET_DETAIL PSD,
                       GG_PATROL_WORK_POINT        PPWP,
                       GG_PATROL_POINT             PPP,
                       GG_TASK_PATROL_OBJECT       PTPO
                 WHERE PTWS.WORK_SHEET_ID = PSD.WORK_SHEET_ID
                   AND TO_CHAR(PTWS.WORK_SHEET_ID) = PPWP.WORK_ID
                   AND PPWP.PATROL_POINT_ID = PPP.PATROL_POINT_ID
                   AND PPP.PATROL_POINT_ID = PTPO.PATROL_POINT_ID
                   AND PTWS.TASK_TYPE = 4
                   AND PTWS.SHEET_TYPE = 2
                   AND PSD.PATROL_TYPE <> -1
                   AND PSD.IS_NEED_PDA = '1'
                   AND PTWS.IS_COMPLETE = '3'
                   AND TO_CHAR(PTWS.REAL_END_DATE, 'YYYY') = '2013'
                   AND PTPO.DEVICE_ID = 1003551930) B,
               (SELECT COUNT(1) COUNT_RESULT
                  FROM GG_TASK_WORK_SHEET          PTWS,
                       GG_TASK_PATROL_SHEET_DETAIL PSD,
                       GG_TASK_PATROL_OBJECT       PTPO,
                       GG_PATROL_EQUIPMENT         PPE
                 WHERE PTWS.WORK_SHEET_ID = PSD.WORK_SHEET_ID
                   AND PTWS.WORK_SHEET_ID = PTPO.WORK_SHEET_ID
                   AND PTPO.PATROL_LINE_ID = PPE.PATROL_LINE_ID
                   AND PTPO.FUNCTION_LOCATION_ID = PPE.FUNCTION_LOCATION_ID
                   AND PTWS.TASK_TYPE = 4
                   AND PTWS.SHEET_TYPE = 1
                   AND PSD.PATROL_TYPE = -1
                   AND PSD.IS_NEED_PDA = '1'
                   AND PTPO.PATROL_TYPE = 1
                   AND PTWS.IS_COMPLETE = '3'
                   AND TO_CHAR(PTWS.REAL_END_DATE, 'YYYY') = '2013'
                   AND PPE.DEVICE_ID = 10035519304) C,
               (SELECT COUNT(1) COUNT_RESULT
                  FROM GG_TASK_WORK_SHEET          PTWS,
                       GG_TASK_PATROL_SHEET_DETAIL PSD,
                       GG_TASK_PATROL_OBJECT       PTPO,
                       GG_PATROL_EQUIPMENT         PPE
                 WHERE PTWS.WORK_SHEET_ID = PSD.WORK_SHEET_ID
                   AND PTWS.WORK_SHEET_ID = PTPO.WORK_SHEET_ID
                   AND PTPO.PATROL_LINE_ID = PPE.PATROL_LINE_ID
                   AND PTPO.FUNCTION_LOCATION_ID = PPE.FUNCTION_LOCATION_ID
                   AND PTWS.TASK_TYPE = 4
                   AND PTWS.SHEET_TYPE = 1
                   AND PTPO.PATROL_TYPE = 1
                   AND PSD.PATROL_TYPE <> -1
                   AND PSD.IS_NEED_PDA = '1'
                   AND PTWS.IS_COMPLETE = '3'
                   AND TO_CHAR(PTWS.REAL_END_DATE, 'YYYY') = '2013'
                   AND PPE.DEVICE_ID = 1003551930) D) ALL_DEVICE

使用case when改造後的SQL如下，不用執行就知道效果明顯。

SELECT DECODE(ALL_COUNT, 0, 0, ROUND(ARRIVED_COUNT / ALL_COUNT, 4) * 100) ARRIVED_RATE
  FROM (SELECT (E.COUNT_RESULT1 + F.COUNT_RESULT1 + G.COUNT_RESULT1 +
               H.COUNT_RESULT1) ARRIVED_COUNT,
               (E.COUNT_RESULT2 + F.COUNT_RESULT2 + G.COUNT_RESULT2 +
               H.COUNT_RESULT2) ALL_COUNT
          FROM (SELECT COUNT(case
                               when (PSD.PATROL_TYPE <> -1 AND PPWP.PATROL_FLAG = 1) then
                                1
                               else
                                null
                             end) COUNT_RESULT1,
                       COUNT(case
                               when PSD.PATROL_TYPE = -1 then
                                1
                               else
                                null
                             end) COUNT_RESULT2
                  FROM GG_TASK_WORK_SHEET          PTWS,
                       GG_TASK_PATROL_SHEET_DETAIL PSD,
                       GG_PATROL_WORK_POINT        PPWP,
                       GG_PATROL_POINT             PPP,
                       GG_PATROL_EQUIPMENT         PPE
                 WHERE PTWS.WORK_SHEET_ID = PSD.WORK_SHEET_ID
                   AND TO_CHAR(PTWS.WORK_SHEET_ID) = PPWP.WORK_ID
                   AND PPWP.PATROL_POINT_ID = PPP.PATROL_POINT_ID
                   AND PPP.PATROL_POINT_ID = PPE.PATROL_POINT_ID
                   AND PTWS.TASK_TYPE = 4
                   AND PTWS.SHEET_TYPE = 2
                   AND PSD.IS_NEED_PDA = '1'
                   AND PTWS.IS_COMPLETE = '3'
                   AND TO_CHAR(PTWS.REAL_END_DATE, 'YYYY') = '2013'
                   AND PPE.DEVICE_ID = 1003551930) E,
               (SELECT COUNT(case
                               when PPWP.PATROL_FLAG = 1 then
                                1
                               else
                                null
                             end) COUNT_RESULT1,
                       COUNT(1) COUNT_RESULT2
                  FROM GG_TASK_WORK_SHEET          PTWS,
                       GG_TASK_PATROL_SHEET_DETAIL PSD,
                       GG_PATROL_WORK_POINT        PPWP,
                       GG_PATROL_POINT             PPP,
                       GG_TASK_PATROL_OBJECT       PTPO
                 WHERE PTWS.WORK_SHEET_ID = PSD.WORK_SHEET_ID
                   AND TO_CHAR(PTWS.WORK_SHEET_ID) = PPWP.WORK_ID
                   AND PPWP.PATROL_POINT_ID = PPP.PATROL_POINT_ID
                   AND PPP.PATROL_POINT_ID = PTPO.PATROL_POINT_ID
                   AND PTWS.TASK_TYPE = 4
                   AND PTWS.SHEET_TYPE = 2
                   AND PSD.PATROL_TYPE <> -1
                   AND PSD.IS_NEED_PDA = '1'
                   AND PPWP.PATROL_FLAG = 1
                   AND PTWS.IS_COMPLETE = '3'
                   AND TO_CHAR(PTWS.REAL_END_DATE, 'YYYY') = '2013'
                   AND PTPO.DEVICE_ID = 1003551930) F,
               (SELECT COUNT(case
                               when PTPO.ARRIVE_TYPE = 1 then
                                1
                               else
                                null
                             end) COUNT_RESULT1,
                       COUNT(case
                               when PSD.PATROL_TYPE = -1 then
                                1
                               else
                                null
                             end) COUNT_RESULT2
                  FROM GG_TASK_WORK_SHEET          PTWS,
                       GG_TASK_PATROL_SHEET_DETAIL PSD,
                       GG_TASK_PATROL_OBJECT       PTPO,
                       GG_PATROL_EQUIPMENT         PPE
                 WHERE PTWS.WORK_SHEET_ID = PSD.WORK_SHEET_ID
                   AND PTWS.WORK_SHEET_ID = PTPO.WORK_SHEET_ID
                   AND PTPO.PATROL_LINE_ID = PPE.PATROL_LINE_ID
                   AND PTPO.FUNCTION_LOCATION_ID = PPE.FUNCTION_LOCATION_ID
                   AND PTWS.TASK_TYPE = 4
                   AND PTWS.SHEET_TYPE = 1
                   AND PTPO.PATROL_TYPE = 1
                   AND PSD.PATROL_TYPE = -1
                   AND PSD.IS_NEED_PDA = '1'
                   AND PTWS.IS_COMPLETE = '3'
                   AND TO_CHAR(PTWS.REAL_END_DATE, 'YYYY') = '2013'
                   AND PPE.DEVICE_ID = 1003551930) G,
               (SELECT COUNT(case
                               when PTPO.ARRIVE_TYPE = 1 then
                                1
                               else
                                null
                             end) COUNT_RESULT1,
                       COUNT(1) COUNT_RESULT2
                  FROM GG_TASK_WORK_SHEET          PTWS,
                       GG_TASK_PATROL_SHEET_DETAIL PSD,
                       GG_TASK_PATROL_OBJECT       PTPO,
                       GG_PATROL_EQUIPMENT         PPE
                 WHERE PTWS.WORK_SHEET_ID = PSD.WORK_SHEET_ID
                   AND PTWS.WORK_SHEET_ID = PTPO.WORK_SHEET_ID
                   AND PTPO.PATROL_LINE_ID = PPE.PATROL_LINE_ID
                   AND PTPO.FUNCTION_LOCATION_ID = PPE.FUNCTION_LOCATION_ID
                   AND PTWS.TASK_TYPE = 4
                   AND PTWS.SHEET_TYPE = 1
                   AND PTPO.PATROL_TYPE = 1
                   AND PSD.PATROL_TYPE <> -1
                   AND PSD.IS_NEED_PDA = '1'
                   AND PTWS.IS_COMPLETE = '3'
                   AND TO_CHAR(PTWS.REAL_END_DATE, 'YYYY') = '2013'
                   AND PPE.DEVICE_ID = 1003551930) H) ARRIVED_DEVICE

巧用case when做SQL優化

下面是一張報表的SQL實現，不用深入業務，你可以看到上面的4條SQL和下面的4條SQL其實大致相同，只是部分條件不一樣，想要優化的第一感覺是如何減少掃描的次數，這個時候case when就派上用場了。 SELECT DECODE(ALL_COUNT, 0, 0,

用sum case when做分類統計sql優化

背景：本來一個頁面涉及到多個類別的統計，最初的程式碼是每一個類別做一個sql來count出數目，可想而知，最後的載入速度和體驗都是不好，能明顯感覺至少5、6秒甚至更多的響應時間之前的程式碼： //136集團客戶故障分析統計 public List<Analysis

CASE WHEN進行SQL改寫優化 – 運維派

導讀今天給大家分享一個通過SQL改寫而獨闢蹊徑的SQL優化案例待優化場景發現SLOW QUERY LOG中有下面這樣一條記錄： ... # Query_time: 59.503827 Lock_time: 0.000198 Rows_sent: 641227 Rows_examined

Sql Server裡巧用Case將多行顯示的資料合併為一行顯示

轉載地址：http://www.cnblogs.com/kingthy/archive/2008/04/29/1175697.html 昨晚在CSDN論壇上看到有某個人問了類似這樣的一個問題，現有三個資料表,分別是學生表,課程表,成績表。它們的結構與樣例資料如下：學生表

優化案例 | CASE WHEN進行SQL改寫優化

導讀今天給大家分享一個通過SQL改寫而獨闢蹊徑的SQL優化案例待優化場景發現SLOW QUERY LOG中有下面這樣一條記錄： ... # Query_time: 59.503827 Lock_time: 0.000198 Rows_sent: 6

mysql中使用case when 做where條件篩選表記錄

平時我們專案中經常用到 where 欄位名=值來篩選查詢結果，但實際也會遇到這樣的情況，如果表中某個欄位code的值是“_1”結尾的，那麼就查全部，否則，根據輸入的引數查詢。這時 case when 就派上用途了： select * from ta

讓天下沒有難用的資料庫 » mysql sql優化之straight_join

在oracle中可以指定的表連線的hint有很多：ordered hint 指示oracle按照from關鍵字後的表順序來進行連線；leading hint 指示查詢優化器使用指定的表作為連線的首表，即驅動表；use_nl hint指示查詢優化器使用nested loops方式連線指定表和其他行源，

讓天下沒有難用的資料庫 » MySql sql優化之order by desc/asc limit M

Order by desc/asc limit M是我在mysql sql優化中經常遇到的一種場景，其優化原理也非常的簡單，就是利用索引的有序性，優化器沿著索引的順序掃描，在掃描到符合條件的M行資料後，停止掃描；看起來非常的簡單，但是我經常看到很多效能較差的sql沒有利用這個優化規律，下面將結合一些

巧用這19條MySQL優化，效率至少提高3倍

本文我們來談談專案中常用的MySQL優化方法，共19條，具體如下： 1、EXPLAIN 做MySQL優化，我們要善用EXPLAIN檢視SQL執行計劃。下面來個簡單的示例，標註（1、2、3、4、5）我們要重點關注的資料： type列，連線型別。一個好的SQL語句至少要達到range級別。杜絕出現a

巧用case進行分類求和（case & group by & sum）

table1 code(商品程式碼) price（價格） num（數量） type（買賣） 1 10 5 B 1 10 10 S 2 5 20 B 2 5 40 S 問題：求商品(code)在某個價格（price）時買進（type=B）和賣出（type=S）的數量（nu

mysql 分組統計(直播內容狀態,帶上內容id),count case when group_concat sql

SELECT group_concat(case live_status when 0 then content_id end) as closed_ids, group_concat(case live_status when 1 then content_id when

關於case when複雜sql語句查詢

問題描述：需要查詢出學生在不同流程狀態對應的活動個數，流程狀態分為五個狀態audit、havePass、notPass、reject、giveup；活動有社會活動(xg_credit_comm_a

MySQL巧用sum,case...when...優化統計查詢

最近在做專案，涉及到開發統計報表相關的任務，由於資料量相對較多，之前寫的查詢語句查詢五十萬條資料大概需要十秒左右的樣子，後來經過老大的指點利用sum,case...when...重寫SQL效能一下子提高到一秒鐘就解決了。這裡為了簡潔明瞭的闡述問題和解決的方法，我簡化一下需求模型。現在資料庫有一

SQL 邏輯優化 case when 轉為 union all

mysq ins 完成 cli 操作 nbsp 過程 date lec 通常數據庫的優化從硬件層面去考慮可分為4個方面： CPU：即降低計算復雜度，如減少sql各類聚合函數，窗口函數，case when等。 IO ：(較少查詢結果集過程中對數據的訪問量。數據優化很大程度從這

sql語句中多個union all的case when優化

今天有點空，打算把報表專案中前人留下的一段sql優化下。原sql大致如下(刪除了設計公司資料的敏感資訊)： select '現場司機' 環節,'勞斯萊斯' 品牌, 'CIQ 直接喂料' 採集專案, count(1) 統計數量 from test_table where

Mysql-巧用join來優化sql

0. 準備相關表來進行接下來的測試相關建表語句請看：https://github.com/YangBaohust/my_sql user1表，取經組 +----+-----------+-----------------+---------------------------------+ |

【sql技巧】mysql修改時，動態指定要修改的字段 update `table` set (case when ....) = 1 where id = xx

幫助 ase stack pan 希望需求記錄開始 lec 如果你點進了這篇帖子，那麽你一定遇到了跟我一樣的問題。別看題目的set case when...，我一開始也是第一反應是用case when但是發現並不好使。問題呢，說得高大上一點：動態指定要修改的字段。

spark-sql case when 問題

spark 大數據 hadoop spark-sqlSELECT CASE (pmod(datediff(f0.`4168388__c_0`,‘1970-01-04‘),7)+1) WHEN ‘1‘ THEN ‘星期日‘ WHEN ‘2‘ THEN ‘星期一‘ WHEN ‘3‘ THEN ‘星期二‘ WHE

SQL優化-子查詢&case&limit

子查詢優化load 導數據.notesdxtdb 數據庫 total_time 475.60秒。監控服務：倉頡select t_.*, a.name acquirer_name,m.merchant_name, am.merchant_name acq_merchant_name,

SQL面試題-行列互換-if、【case when】

alt 技術我們 rim case when ade 建表行數 prim http://www.cda.cn/view/21469.html tb_lemon_grade中，表中字段id，student_name，course，score分別表示成績id，學生姓名，課

巧用case when做SQL優化

相關推薦