oracle-按年、月、周、日、時、分 分組查詢統計數據，無數據補零（connect by）

目的：統計一段時間內每年、每月、每周、每日、每時、每分數據，無數據時自動補零

思路：1. 生成給定時間段對應日期

           2. 將原表中該時間段內的不為0的數據量統計出來

           3. 用left join連接起來，無數據的自動補零

難點主要在於步驟一中生成該時間段對應的日期，話不多說，直接貼代碼：

- - 獲取某時間段內的每年

SELECT TO_CHAR(ADD_MONTHS(TO_DATE(SUBSTR('2019-12-12 08:01:00', 1, 4), 'yyyy'), (ROWNUM - 1) * 12), 'yyyy')  AS DATES  FROM DUAL
CONNECT BY ROWNUM <= CEIL(MONTHS_BETWEEN(TO_DATE(SUBSTR('2020-01-12 08:01:00', 1, 4), 'yyyy'), TO_DATE(SUBSTR('2019-12-12 08:01:00', 1, 4),
'yyyy')) / 12 +1)

- - 獲取某時間段內的每月

SELECT TO_CHAR(ADD_MONTHS(TO_DATE(SUBSTR('2019-12-12 08:01:00', 1, 7), 'YYYY-MM'), ROWNUM - 1), 'YYYY-MM') AS DATES FROM DUAL
CONNECT BY ROWNUM <= CEIL(MONTHS_BETWEEN(TO_DATE(SUBSTR('2020-01-12 08:01:00', 1, 7), 'YYYY-MM'), TO_DATE(SUBSTR('2019-12-12 08:01:00', 1, 7),
'YYYY-MM'))+1)

- - 獲取某時間段內的每周

SELECT TO_CHAR(TO_DATE(SUBSTR('2019-12-12 08:01:00', 1, 10), 'YYYY-MM-DD') + (ROWNUM - 1) * 7， 'IW') AS DATE FROM DUAL
CONNECT BY ROWNUM <=TRUNC(TO_DATE(SUBSTR('2020-01-12 08:01:00', 1, 10), 'YYYY-MM-DD') - TO_DATE(SUBSTR('2019-12-12 08:01:00', 1, 10),
 'YYYY-MM-DD')) / 7 + 1

- - 獲取某時間段內的每日

SELECT TO_CHAR(TO_DATE(SUBSTR('2019-12-12 08:01:00', 1, 10), 'yyyy-mm-dd') + ROWNUM - 1, 'yyyy-mm-dd') AS DATAS FROM DUAL
CONNECT BY ROWNUM <= TRUNC(TO_DATE(SUBSTR('2020-01-12 08:01:00', 1, 10), 'yyyy-mm-dd') - TO_DATE(SUBSTR('2019-12-12 08:01:00', 1, 10), 'yyyy-mm-dd')) + 1

- - 獲取某時間段內的每時

SELECT to_char(TO_DATE(SUBSTR('2019-12-12 08:01:00', 1, 13), 'YYYY-MM-DD HH24') + (ROWNUM - 1) / 24,'yyyy-mm-dd HH24') AS DATES FROM DUAL
CONNECT BY ROWNUM <= floor(to_number(TO_DATE(SUBSTR('2020-01-12 08:01:00', 1, 13), 'yyyy-mm-dd HH24')-to_date(SUBSTR('2019-12-12 08:01:00', 1, 13),
'yyyy-mm-dd hh24'))*24 +1)

- - 獲取某時間段內的每分

SELECT to_char(TO_DATE(SUBSTR('2019-12-12 08:01:00', 1, 16), 'YYYY-MM-DD HH24:MI') + (ROWNUM-1 ) / (24 * 60) ,'yyyy-mm-dd HH24:MI') AS DATES FROM DUAL
CONNECT BY ROWNUM <= floor(to_number(TO_DATE(SUBSTR('2020-01-12 08:01:00', 1, 16), 'yyyy-mm-dd HH24:MI')-to_date(SUBSTR('2019-12-12 08:01:00', 1, 16),
'yyyy-mm-dd hh24:MI'))*(24 * 60) +1)

 

完整版（以周為例）

SELECT NVL(D.DATA1, 0) POWER,NVL(D.UNIT, 'Week') UNIT, C.WEEK TIME1
       FROM(SELECT TO_CHAR(TO_DATE(SUBSTR('2019-12-12 08:01:00', 1, 10), 'YYYY-MM-DD') + (ROWNUM - 1) * 7， 'IW') AS WEEK
                   FROM DUAL CONNECT BY ROWNUM <= 
                TRUNC(TO_DATE(SUBSTR('2020-01-12 08:01:00', 1, 10), 'YYYY-MM-DD') - TO_DATE(SUBSTR('2019-12-12 08:01:00', 1, 10), 'YYYY-MM-DD')) / 7 + 1)C
       LEFT JOIN (SELECT SUM(DATA1) DATA1,TIME1,'Week' Unit 
                   FROM (SELECT SUM(TO_NUMBER(DATA1)) DATA1,to_char(INSERT_TIME,'iw') Time1 FROM TABLE1
                                where to_char(INSERT_TIME,'yyyy-mm-dd')>=SUBSTR('2019-12-12 08:01:00', 1, 10) and 
                                      to_char(INSERT_TIME,'yyyy-mm-dd')<=SUBSTR('2020-01-12 08:01:00', 1, 10)
                                      GROUP BY to_char(INSERT_TIME,'iw'))
                               GROUP BY TIME1)D
                   ON  C.WEEK = D.TIME1
       ORDER  BY WEEK

Oracle connect by 基本語法：

select * from table [start with condition1]
    connect by [prior] id=parentid

一般用來查找存在父子關系的數據，也就是樹形結構的數據；其返還的數據也能夠明確的區分出每一層的數據。

   start with condition1 是用來限制第一層的數據，或者叫根節點數據；以這部分數據為基礎來查找第二層數據，然后以第二層數據查找第三層數據以此類推。

   connect by [prior] id=parentid 這部分是用來指明oracle在查找數據時以怎樣的一種關系去查找；比如說查找第二層的數據時用第一層數據的id去跟表里面記錄的parentid字段進行匹配，如果這個條件成立那么查找出來的數據就是第二層數據，同理查找第三層第四層…等等都是按這樣去匹配。