SQL Group按日期排序

Question

好的，我有一張表，其中包括：名稱，c_id，日期。

該表中的某些條目符合某些標準，我可以使用WHERE有效選擇，所以這不是問題。

我想什麼，然後，是一種分組的，除了名稱和C_ID列，三個日期的組：日期 - 1，日期，日期+ 1。換句話說，我希望輸出的每一行代表具有相同名稱和c_id作爲某個相關條目的所有條目，並且該條目發生在該條目之前的一天之前（包括其自身）。

我該怎麼做呢？

--edit：

（EDIT2：起源表應該是Table 1和表2的INNER JOIN ON Table1.id = Table2.id）

樣本數據：

Table1: 
    id | c_id |  date  | other stuff 
----------------------------------------------------- 
    01 | abc | 2015/12/09 | whatever 
    02 | abc | 2015/12/09 | whatever 
    03 | abc | 2015/12/10 | relevant criterion 
    04 | abc | 2015/12/11 | whatever 
    05 | def | 2015/11/15 | whatever 
    06 | def | 2015/11/16 | relevant criterion 
    07 | abc | 2015/11/17 | whatever 
    08 | mnc | 2016/01/02 | whatever 
    09 | mnc | 2016/01/02 | whatever 
    10 | mnc | 2016/01/03 | whatever 
    11 | mnc | 2016/01/03 | whatever 
    12 | mnc | 2016/01/03 | whatever 
    13 | mnc | 2016/01/04 | relevant criterion 
    14 | mnc | 2016/01/05 | whatever 
    15 | mnc | 2016/01/05 | whatever 
    16 | mnc | 2016/01/06 | whatever 

Table2: 
    id | Name | other stuff 
-------------------------------------- 
    01 | John | whatever 
    02 | John | whatever 
    03 | John | whatever 
    04 | John | whatever 
    05 | Mary | whatever 
    06 | Mary | whatever 
    07 | Mary | whatever 
    08 | Alice | whatever 
    09 | Alice | whatever 
    10 | Alice | whatever 
    11 | Alice | whatever 
    12 | Alice | whatever 
    13 | Alice | whatever 
    14 | Alice | whatever 
    15 | Alice | whatever 
    16 | Alice | whatever 

樣品所需的輸出：

Name | c_id | pivot_date | count 
------------------------------------------ 
    John | abc | 2015/12/10 | 4 
    Mary | def | 2015/11/16 | 2 
    Alice | mnc | 2016/01/04 | 6 

（該pivot_date部分沒有特別的必要的一個與相關的標準，所涉及的日期中的任何一個都好）

Answer 1

更新了新的樣本數據：

SELECT t.name, t.c_id, t.date pivot_date, COUNT(*) count 
    FROM record t 
    JOIN record t2 
     ON t2.name = t.name 
    AND t2.c_id = t.c_id 
    AND t2.date >= t.date - INTERVAL 1 DAY   
    AND t2.date <= t.date + INTERVAL 1 DAY 
    WHERE t.other_stuff = 'relevant criterion' 
GROUP BY t.name, t.c_id, t.date 

See SQLFiddle

-

更新了新的樣本數據：

SELECT t2.name, t1.c_id, t1.date pivot_date, COUNT(*) count 
    FROM table1 t1 
    JOIN table1 to1 
     ON to1.c_id = t1.c_id 
    AND to1.date >= t1.date - INTERVAL 1 DAY   
    AND to1.date <= t1.date + INTERVAL 1 DAY 
    JOIN table2 
     ON t2.id = t1.id 
    WHERE t1.other_stuff = 'relevant criterion' 
GROUP BY t2.name, t1.c_id, t1.date 

Answer 2

這是一種方式：

@dt := '2015-12-10' 

SELECT Name , c_id, MAX(`date`) maxdate, COUNT(*) countof 
FROM table1 
WHERE date >= @dt - INTERVAL 1 DAY AND `date` < @dt + INTERVAL 2 DAY 
GROUP BY Name , c_id 

注日期列可以是MIN（日期）作爲替代

Answer 3

我不得不使用分析LAG/LEAD函數Oracle中建立這個。我不知道你是否能在MYSQL模擬這些功能（看here仿真）

table1是你的數據

with 
relevant as 
(
select * 
    from (
    select 
     nm, 
     c_id, 
     stuff, 
     dt, 
     lag(dt,1) over (order by nm, dt, c_id) prev, 
     lead(dt,1) over (order by nm, dt, c_id) nxt 
    from table1) 
where stuff = 'relevant criterion' 
) 
select 
    d.nm, 
    d.c_id, 
    to_char(d.dt,'DD-MON-YYY') dt, 
    d.stuff, 
    count(*) over (partition by d.nm, d.c_id) cnt 
    from table1 d 
where (d.dt in (select rp.prev from relevant rp where d.nm = rp.nm and d.c_id = rp.c_id) 
     OR 
     d.dt in (select rn.nxt from relevant rn where d.nm = rn.nm and d.c_id = rn.c_id ) 
     OR 
     d.dt in (select rn.dt from relevant rn where d.nm = rn.nm and d.c_id = rn.c_id ) 
    ) 
; 

結果：

NM C_ID DT STUFF CNT 
Alice mnc 03-JAN-016 whatever 6 
Alice mnc 03-JAN-016 whatever 6 
Alice mnc 03-JAN-016 whatever 6 
Alice mnc 05-JAN-016 whatever 6 
Alice mnc 04-JAN-016 relevant criterion 6 
Alice mnc 05-JAN-016 whatever 6 
John abc 11-DEC-015 whatever 4 
John abc 10-DEC-015 relevant criterion 4 
John abc 09-DEC-015 whatever 4 
John abc 09-DEC-015 whatever 4 
Mary def 15-NOV-015 whatever 2 
Mary def 16-NOV-015 relevant criterion 2 

您可以通過創建過濾上述結果此查詢爲視圖並添加where stuff = 'relevant criterion'篩選條件

NM C_ID DT STUFF CNT 
Alice mnc 04-JAN-016 relevant criterion 6 
John abc 10-DEC-015 relevant criterion 4 
Mary def 16-NOV-015 relevant criterion 2 

Answer 4

這是一種天真的做法。這類型的作品：

select t2.Name, t1.c_id, min(t1.Date) as pivot_date, count(*) as cnt 
from Table1 t1 inner join Table2 t2 on t2.id = t1.id 
where <criteria...> 
group by t2.Name, t1.c_id, datediff(t1.Date, '2015-01-01') div 3 

這種方式任意劃分曆法轉換成基於一個出發點的三階會阻塞。

將有可能有連續的日期，形成第一個塊的結尾和下一個的開始。我懷疑你是否想要這樣，我會看看我能否找到能夠提供更好平衡的東西。

您遺漏的過濾標準有多複雜？