COUNT（*）返回错误的数字

Question

It seems I don't get it something. Please consider this query

SELECT COUNT(*) AS `numrows` 
FROM (`exp_channel_titles` ch) 
JOIN `exp_channel_data` cd ON `cd`.`entry_id`=`ch`.`entry_id` 
LEFT JOIN `exp_matrix_data` md ON `md`.`entry_id`=`ch`.`entry_id` and field_id = 14 
LEFT JOIN `exp_assessment_users` au ON `au`.`entry_id`=`ch`.`entry_id` 
WHERE ch.channel_id = 4 GROUP BY `ch`.`entry_id`

it returns 2

but if I change it to

SELECT * 
FROM (`exp_channel_titles` ch) 
JOIN `exp_channel_data` cd ON `cd`.`entry_id`=`ch`.`entry_id` 
LEFT JOIN `exp_matrix_data` md ON `md`.`entry_id`=`ch`.`entry_id` and field_id = 14 
LEFT JOIN `exp_assessment_users` au ON `au`.`entry_id`=`ch`.`entry_id` 
WHERE ch.channel_id = 4 GROUP BY `ch`.`entry_id`

result is 1 row only. How so?

Answer 1

You're grouping, which means internally matching rows are collapsed into a single entity. eg consider a fake table like this:

field
-----
a
a

Yes, a one field table, with two records, both of which have the value a in them.

SELECT *
FROM table
GROUP BY field

group by will find all fields which have the same value, and collapse them down into a SINGLE record, so your two records of a become one row in the result set, and you end up with

field
-----
a

But doing

SELECT count(*)
FROM table
GROUP BY field

changes things. Now the DB will literally count how many records were collapsed down into the single row of result set. So you still get a SINGLE row in the result set, which contains a count of how many rows were collapsed by the group by :

count(*)
--------
2

One row, with a value of 2, because there were two rows with a .

Now if you had a table with more records:

field
-----
a
a
b
c
c
c

You would get:

SELECT * ... GROUP BY field

field
-----
a
b
c

SELECT count(*), field ... GROUP BY field

count(*)   field
----------------
2          a
1          b
3          c

again, 3 rows of results, but note how the count represents how many of each grouped field there are in the original table.