Reprinted: https://blog.csdn.net/u014717572/article/details/80687042
EDITORIAL words: group by using a long time, woke up this morning and suddenly felt a strange group by good, there is always a gluten do not, however, why not be able to select * from Table group by id, why must not be a *, and aggregate function is one column or a column, group by more than one field can be a good understanding of how to do? But finally turn up, simple to write about it, Daniel had a direct skip it.
Text start =========== =========
First look at Table 1, Table named test:
Table 1
Execute the following SQL statement:
|
1
2
|
SELECT
name
FROM
test
GROUP
BY
name
|
You should be very easy to know the results of running, yes, that is Table 2 below:
Table 2
However, in order to better understand the "group by" multiple columns "and" aggregate function ", I suggest in the process of thinking in the process of Table 1 to Table 2, the addition of a fictional middle of the table: a virtual table 3. Here to talk about how to think about the implementation of the above SQL statement:
1.FROM test: After the execution of the sentence, and should result in Table 1, as is the original table.
2.FROM test Group BY name: After the execution of the sentence, we can imagine generating the virtual table 3, as shown in the FIG., Such generation processes are: group by name, then the name to find the row with the same name the row values, It merged into one line, as described for aa name value, then <1 aa 2> and <2 aa 3> two lines merged into one line, and all the values of id number value is written to a cell surface Gerry.
3. The next step is for the virtual table 3 to perform a Select statement:
(1) If you do select *, then the results should be returned virtual table 3, but the number of content-id and some units Gerry face multiple values, and relational database is based on the relationship of the cell is not It allows multiple values, so you see, perform a select * statement on the error.
（2）我们再看name列，每个单元格只有一个数据，所以我们select name的话，就没有问题了。为什么name列每个单元格只有一个值呢，因为我们就是用name列来group by的。
（3）那么对于id和number里面的单元格有多个数据的情况怎么办呢？答案就是用聚合函数，聚合函数就用来输入多个数据，输出一个数据的。如cout(id)，sum(number)，而每个聚合函数的输入就是每一个多数据的单元格。
（4）例如我们执行select name,sum(number) from test group by name，那么sum就对虚拟表3的number列的每个单元格进行sum操作，例如对name为aa的那一行的number列执行sum操作，即2+3，返回5，最后执行结果如下：
（5）group by 多个字段该怎么理解呢：如group by name,number，我们可以把name和number 看成一个整体字段，以他们整体来进行分组的。如下图
（6）接下来就可以配合select和聚合函数进行操作了。如执行select name,sum(id) from test group by name,number，结果如下图：
至此，我已经对我自己对如此简单的问题有如此天马行空的想法所折服，洗洗睡觉。