To remove the distinction between int and long
int(1e200) should return a long object instead of
raising an OverflowError.

This is related to bug #629989:
int("123123123123123123") doesn't work.

Fixing this could be done by changing
Objects/floatobject.c/float_int (see attached patch),
but this changes the meaning of the nb_int slot.

Logged In: YES 
user_id=6380

PyInt_AsLong() insists that nb_int returns a PyInt_Object.

I think it's okay that nb_int returns a non-int -- classes
don't require __int__ to do this either. So PyInt_AsLong()
should accept a PyLong_Object as well, I guess.

Logged In: YES 
user_id=89016

This second patch (diff2.txt) fixes PyInt_AsLong(), so it is
able to cope with longs. In addition to that int(long) has
been changed to return a long, if the int overflows:

>>> int(1L<<100) 
1267650600228229401496703205376L

For subinstances of long the patch ensures that a proper
long instance will be created. (See the additional test in
test_long.py).

If I understood the sourcecode correctly nb_int, nb_long and
nb_float might return objects, that are not instances of
int, long or float, but instances of subclasses of int, long
or float. Is this a bug or an (undocumented?) feature?

Logged In: YES 
user_id=6380

Looks good (I haven't run this).

I was shocked to see that int(2**1000) will now return a
long rather than failing, but that's actually correct! :-)

Logged In: YES 
user_id=89016

So should I add a note to Misc/NEWS and check it in?

Are there any changes to the documentation that should be done?

Logged In: YES 
user_id=6380

Please do check it in and add a note to NEWS.

I dunno about docs, please look and ask around. int() should
be easy to find.

Logged In: YES 
user_id=89016

Doc/api/abstract.tex 1.20
Checked in as:
Doc/lib/libfuncs.tex 1.120
Lib/test/test_b1.py 1.58
Lib/test/test_long.py  1.21
Lib/test/test_types.py 1.40
Misc/NEWS 1.529
Objects/floatobject.c 2.117
Objects/intobject.c 2.95
Objects/longobject.c 1.145