[ python-Bugs-1119418 ] xrange() builtin accepts keyword arg silently

SourceForge.net noreply at sourceforge.net
Fri Aug 26 08:45:57 CEST 2005


Bugs item #1119418, was opened at 2005-02-09 17:57
Message generated for change (Settings changed) made by birkenfeld
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=105470&aid=1119418&group_id=5470

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: Python Library
Group: None
>Status: Closed
>Resolution: Fixed
Priority: 6
Submitted By: Martin Blais (blais)
Assigned to: Reinhold Birkenfeld (birkenfeld)
Summary: xrange() builtin accepts keyword arg silently

Initial Comment:
Calling ``xrange(10, 100, step=10)`` results in a
xrange(10, 100) iterator silently.  In contrast,
``range(10, 100, step=10)`` raises an exception.  See
test program below.

Two possible fixes:
1. fix xrange() so that it returns a xrange(10, 100,
10) iterator
2. make sure that xrange() raises an exception too.



#!/usr/bin/env python

def foo( min_, max_, step=1 ):
    print min_, max_, step

print '===================='
foo(10, 100, 10)
foo(10, 100, step=10)

print '===================='
print xrange(10, 100, 10)
print xrange(10, 100, step=10)

print '===================='
print range(10, 100, 10)
print range(10, 100, step=10)





elbow:/usr/.../lib/python2.4$ /tmp/a.py
====================
10 100 10
10 100 10
====================
xrange(10, 100, 10)
xrange(10, 100)
====================
[10, 20, 30, 40, 50, 60, 70, 80, 90]
Traceback (most recent call last):
  File "/tmp/a.py", line 16, in ?
    print range(10, 100, step=10)
TypeError: range() takes no keyword arguments

> /tmp/a.py(16)?()
-> print range(10, 100, step=10)
(Pdb) 
elbow:/usr/.../lib/python2.4$ 


----------------------------------------------------------------------

>Comment By: Reinhold Birkenfeld (birkenfeld)
Date: 2005-08-26 08:45

Message:
Logged In: YES 
user_id=1188172

Committed. Must the new API function be somehow documented?

Commit report follows:

Checking in Python/getargs.c;
new revision: 2.106; previous revision: 2.105
Checking in Include/modsupport.h;
new revision: 2.42; previous revision: 2.41
Checking in Objects/rangeobject.c;
new revision: 2.55; previous revision: 2.54
Checking in Objects/setobject.c;
new revision: 1.56; previous revision: 1.55
Checking in Objects/bufferobject.c;
new revision: 2.27; previous revision: 2.26
Checking in Objects/sliceobject.c;
new revision: 2.23; previous revision: 2.22
Checking in Modules/arraymodule.c;
new revision: 2.99; previous revision: 2.98
Checking in Modules/itertoolsmodule.c;
new revision: 1.41; previous revision: 1.40
Checking in Modules/operator.c;
new revision: 2.31; previous revision: 2.30
Checking in Modules/_randommodule.c;
new revision: 1.8; previous revision: 1.7
Checking in Modules/zipimport.c;
new revision: 1.19; previous revision: 1.18
Checking in Modules/collectionsmodule.c;
new revision: 1.39; previous revision: 1.38

Checking in Python/getargs.c;
new revision: 2.102.2.3; previous revision: 2.102.2.2
Checking in Include/modsupport.h;
new revision: 2.41.4.1; previous revision: 2.41
Checking in Objects/rangeobject.c;
new revision: 2.53.2.1; previous revision: 2.53
Checking in Objects/setobject.c;
new revision: 1.31.2.4; previous revision: 1.31.2.3
Checking in Objects/bufferobject.c;
new revision: 2.26.2.1; previous revision: 2.26
Checking in Objects/sliceobject.c;
new revision: 2.22.4.1; previous revision: 2.22
Checking in Modules/arraymodule.c;
new revision: 2.97.2.1; previous revision: 2.97
Checking in Modules/itertoolsmodule.c;
new revision: 1.39.2.1; previous revision: 1.39
Checking in Modules/operator.c;
new revision: 2.29.4.1; previous revision: 2.29
Checking in Modules/_randommodule.c;
new revision: 1.7.4.1; previous revision: 1.7
Checking in Modules/zipimport.c;
new revision: 1.18.2.1; previous revision: 1.18
Checking in Modules/collectionsmodule.c;
new revision: 1.36.2.1; previous revision: 1.36


----------------------------------------------------------------------

Comment By: Reinhold Birkenfeld (birkenfeld)
Date: 2005-08-25 22:21

Message:
Logged In: YES 
user_id=1188172

Attaching a huge patch.

It introduces the function you described, and calls it in
the constructors of xrange, set, frozenset, buffer and slice.

Moreover, I found it to be necessary in objects in the
following modules:
- array (array)
- itertools (cycle, dropwhile, takewhile, islice, imap,
chain, ifilter, count, izip, repeat)
- operator (attrgetter, itemgetter)
- _random (Random)
- zipimport (zipimporter)
- collections (deque)


----------------------------------------------------------------------

Comment By: Raymond Hettinger (rhettinger)
Date: 2005-08-25 16:35

Message:
Logged In: YES 
user_id=80475

xrange() needs to be kept closely parallel with range() . 
Accordingly, it should not sprout keyword arguments.

For all of these, I think the solution is to raise a
TypeError when keyword arguments are supplied to
functions/types that don't handle them.

Encapsulate the logic in a new internal API function:

int
 _PyArg_NoKeywords(char *funcname, PyObject *kwds)
{
    if (kwds == NULL) return 1;
    if (!PyDict_CheckExact(kwds)){
         bad internal call
         return 0;
    }
    if (PyDict_Size(kwds) == 0)
         return 1;
    set_exc_typeerror(funcname does not take kw args)
    return 0;
}


Then go through the code finding all the type constructors
(grep for PyTypeObject) and add the error check if the kwds
arg is being ignored).  For example:

range_new(PyTypeObject *type, PyObject *args, PyObject *kw)
{
   if (!_PyArg_NoKeywords("xrange", kw))
         return NULL;
    . . .


Be sure to add test cases for every constructor that gets
changed.  Also, go ahead an backport to Py2.4.2.

----------------------------------------------------------------------

Comment By: Reinhold Birkenfeld (birkenfeld)
Date: 2005-08-25 15:20

Message:
Logged In: YES 
user_id=1188172

Please review. Should I work on a patch to buffer, set and
slice, too?

----------------------------------------------------------------------

Comment By: Reinhold Birkenfeld (birkenfeld)
Date: 2005-06-09 23:04

Message:
Logged In: YES 
user_id=1188172

I delved deeper into this, and it seems that the difference
is caused by range being a method (of bltinmodule, defined
as METH_VARARGS), while xrange is a constructor for a
rangeobject.

Such constructor functions get three arguments (the type
object, the args and the kw args), and when the kw args are
not checked like in str(), they can pass freely and are ignored.

I have attached a patch which changes the range object
constructor (xrange) to accept keyword arguments.

Other builtin types that need such a correction include
buffer, set, slice.

----------------------------------------------------------------------

Comment By: Terry J. Reedy (tjreedy)
Date: 2005-02-16 23:21

Message:
Logged In: YES 
user_id=593130

Functions coded in C generally do not take keyword 
arguments.  (Special coding is required to achieve 
otherwise.)  In 2.2, range and xrange both followed this rule:

>>> xrange(1,20,step=2)
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
TypeError: xrange() takes no keyword arguments

>>> range(1,20,step=2)
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
TypeError: range() takes no keyword arguments

So, removal of the error message by 2.4 seem to be a bug.

Surprise:
>>> str(object=1)
'1'
>>> str(i=2)
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
TypeError: 'i' is an invalid keyword argument for this function

There is nothing in the doc(Lib Ref) or doc string of str vs. 
range and xrange that would lead me to expect this.

I looked around CVS a bit to see if the past or possible future 
change was something simple, but I could not find source of 
error message in bltinmodule.c, ceval.c, getargs.c,  
rangeobject.c, or typeobject.c, so I will leave this to 
someone else.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=105470&aid=1119418&group_id=5470


More information about the Python-bugs-list mailing list