As it currently stands, the iterator argument to
itertools.starmap() must yield tuples, even those any
iterable can be *-expanded in function calls. The
attached patch changes starmap()'s behaviour (as well
as docs and tests) to allow the provided iterator to
return any iterable object.

The patch is against r46582.

Logged In: YES 
user_id=671362

Logged In: YES 
user_id=591932

The patch can be simplified by always calling
PySequence_Tuple() and doing the decref instead of special
casing tuples.  If you pass PySequence_Tuple() a tuple it
will  incref it and hand it back to you.

Logged In: YES 
user_id=1344176

I've updated the patch per jackdied's suggestion, plus
improved the test suite a bit.

The patch is now against r47103.

Logged In: YES 
user_id=80475

Since starmap() is documented to behave like "yield 
function(*iterable.next())", I think this fixup can be 
considered a bugfix.  If Anthony is okay with this, it 
should into Py2.5.

While Jack's suggested modification leads to a 
conceptually simpler patch, I prefer an alternate version 
that only calls PySequence_Tuple() if PyTuple_CheckExact() 
fails.  That will make sure that we don't slow down the 
most common case.

Also, I prefer slightly different updates to the docs and 
test suite.  If Anthony approves for inclusion in Py2.5, I 
will upload my own version of a patch to starmap().

Logged In: YES 
user_id=591932

Raymond's intuition for bare metal is correct.  My suggested
simplifaction slowed down the tuple case by most of a
percent (I benchmarked it but left it out of the comment). 
Preternatural, I tell you.

Logged In: YES 
user_id=29957

Hm. Making the code agree with the docs is a good plan.
Please make sure to include a very clear NEWS item.

This was fixed in r60013.