I've checked current code. Reporter is right.
I need to check when this has been changed and I think row should be required parameter.  I'll keep this behavior if and only if this was the implemented from the beginning :)
(Current code internally advance row in result structure)

So DO NOT take advantage of this.

Additional comment:
All pg_fetch_* works as reporter mentioned, but
DO NOT take advantage of this.
I may change behavior after I check php3 and php4 CVS log :)

Works wonderfully without the row. That's certainly a win convienence wise, especially for folks coming over from mysql where the loop syntax is trivial 

(while $row = fetch())
{}

I'd prefer to see the docs change then the functionality. I'll take a look at the code itself but it is hard to see how this is a performance loss, most users end up keeping a counter anyways.

Why do you prefer making the row required? It seems like a silly added restriction, especially when I bet 80% of the cases where the function is called don't need it. 

Implemented 6 months ago by jon (1.112) pgsql.c.
Looks reasonable.

Thanks for checking the change :)
I'm the new module maintainer, so I'm asking previous module maintainers about it. Since MySQL seems it does the same, I'll keep this behavior probably, but it is not a promise yet. ;)

Sounds good. I think it would be silly to drop this change.

- It does not appear to impact existing code (patch switches based on argument count) 

- The reason I discovered it because I was annoyed with what I thought the existing behavior was.

Good luck with the new maintainership. 

Status -> assigned

Note added to doc.