select count(*) from A group by f1和select count(ic) from A group by f1哪个快

[learningjsp]

表中记录千万行级的情况下，这两个语句哪个快些？
有没有更好的方法？

[caomm2001]

count(*)快

[sujinlin]

ic上有索引么

[lhb2000]

count(索引键)>count(0)>count(*)

[marco08]

count(*) --包括NULL
count(col1) --忽略NULL

create table T(col1 char(1))
insert T select 'A'
union all select 'A'
union all select 'N'
union all select 'B'
union all select 'B'
union all select NULL
union all select NULL

select col1, count(*) as num from T group by col1
--result
/*
col1 num
---- -----------
NULL 2
A 2
B 2
N 1

(4 row(s) affected)
*/

select col1, count(col1) as num from T group by col1
--result
/*
col1 num
---- -----------
NULL 0
A 2
B 2
N 1

(4 row(s) affected)
*/

[shuqinpeng]

mark

[wanglv]

建议采用count(ic)方式

[caixia615]

建议LZ用查询计划看看..

[youqiong]

当然count(*)快