SQL Best Practice: count(1) or count(*) [duplicate]

Question

Possible Duplicate:
Count(*) vs Count(1)

I remember anecdotally being told:

never use count(*) when count(1) will do

Recently I passed this advice on to another developer, and was challenged to prove this was true. My argument was what I was told along with when I was given the advice: that the database would only return the first column, which would then be counted. The counterargument was that the database wouldn't evaluate anything in the brackets.

From some (unscientific) testing on small tables, there certainly seems to be no difference. I don't currently have access to any large tables to experiment on.

I was given this advice when I was using Sybase, and tables had hundreds of millions of rows. I'm now working with Oracle and considerably less data.

So I guess in summary, my two questions are:

1. Which is faster, count(1) or count(*)?
2. Would this vary in different database vendors?

Answer 1

According to another similar question (Count(*) vs Count(1)), they are the same.

In Oracle, according to Ask Tom, count(*) is the correct way to count the number of rows because the optimizer changes count(1) to count(*). count(1) actually means to count rows with non-null 1's (all of them are non-null so optimizer will change it for you).