order by 1 與 order by 列名
經常見sql語句中order by 1或者order by 2...order by N,有時候很莫名其妙.其實1表示第一個欄位,2表示第二欄位; 依此類推,當表中只有2個欄位時,oder by 3就會出錯,這個跟order by 列名沒有什麼區別,不過在特殊情況下還是很有用的.
例如table1(p_code int,issue_date datetime,issue_num int)
p_code issue_date issue_num
101 2016-12-01 00:00:00.000 45
102 2016-12-01 00:00:00.000 89
102 2016-12-03 00:00:00.000 44
103 2016-12-03 00:00:00.000 44
101 2016-12-02 00:00:00.000 44
101 2016-12-03 00:00:00.000 44
101 2016-12-03 00:00:00.000 45
101 2016-12-03 00:00:00.000 44
101 2016-12-03 00:00:00.000 44
102 2016-12-03 00:00:00.000 47
101 2016-12-03 00:00:00.000 48
104 2016-12-03 00:00:00.000 86
101 2016-12-03 00:00:00.000 56
101 2016-12-03 00:00:00.000 29
101 2016-12-11 00:00:00.000 11
我想獲取某個p_code過去(不含當天)10次的issue_num平均值,
可能我們會這麼寫
select top 10 p_code, avg(issue_num) as avgissue from table1
where issue_date<>(select convert(varchar(10),getdate(),120))
and p_code='101'
group by p_code
order by issue_date desc但是以上寫法是錯誤,的,會提示 Column "table1.issue_date" is invalid in the ORDER BY clause because it is not contained in either an aggregate function or the GROUP BY clause.必須要求issue_date在group by中.
所以我們改成
select top 10 p_code, avg(issue_num) as avgissue from table1
where issue_date<>(select convert(varchar(10),getdate(),120))
and p_code='101'
group by p_code,issue_date
order by issue_date desc但是得到的結果卻是(今天是2016.12.11)
p_codeavgissue
10144
10144
10145
顯然以上不是我們想要的結果.
那麼該如何結果這種問題呢,當然你可以再次進行avg()聚合運算,但是這裡我們既然說oder by N,那麼就用這個知識來解決.
語句如下.
select top 10 p_code, avg(issue_num) as avgissue from table1
where issue_date<>(select convert(varchar(10),getdate(),120))
and p_code='101'
group by p_code
order by 2 descp_codeavgissue
10144
是不是很神奇呢!
當然作為碼農,我們儘量用列名來解決,以便增加程式碼的可讀性和易維護性.