pyobjc-dev

[Pyobjc-dev] [ pyobjc-Bugs-678818 ] Missing NSCFNumber to int conversions

[SourceForge.net <no...@so...>]

Bugs item #678818, was opened at 2003-02-01 14:16
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=114534&aid=678818&group_id=14534

Category: None
Group: None
Status: Open
Resolution: None
Priority: 5
Submitted By: David Eppstein (eppstein)
Assigned to: Nobody/Anonymous (nobody)
Summary: Missing NSCFNumber to int conversions

Initial Comment:
The attached code shows two problems and one
non-problem with NSCFNumber (a data type that arises
when objc passes an integer value to a python routine
-- in the code, python ints passed to undoManager get
transmogrified into this type when an undo or redo is
requested).

1. if x: should only execute the if clause when x is
nonzero.  But if x is a zero NSCFNumber, the clause
gets executed erroneously (third line of code's output
is "Increment by ..." instead of "Null increment")

2. str(x) should produce a decimal representation of x,
but if x is an NSCFNumber, it instead produces a string
like '<NSCFNumber objective-c instance 0x832ed0>'
(third and fourth lines of code's output).

Given these bobbles, I would have expected to also have
problems with arithmetic expressions involving
NSCFNumber, but that is a non-problem -- the arithmetic
in the attached code behaves correctly.

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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=114534&aid=678818&group_id=14534

Re: [Pyobjc-dev] [ pyobjc-Bugs-678818 ] Missing NSCFNumber to int conversions

[<bb...@ma...>]

On Saturday, Feb 1, 2003, at 17:16 US/Eastern, SourceForge.net wrote:
> 1. if x: should only execute the if clause when x is
> nonzero.  But if x is a zero NSCFNumber, the clause
> gets executed erroneously (third line of code's output
> is "Increment by ..." instead of "Null increment")

I believe this is working correctly on my system?

[bumbox:~/bbum-developer/sourceforge/pyobjc] bbum% python foo.py
Increment by 3 to 3
Null decrement to 3
Null increment to 3
Decrement by <NSCFNumber objective-c instance 0x470330> to 0

> 2. str(x) should produce a decimal representation of x,
> but if x is an NSCFNumber, it instead produces a string
> like '<NSCFNumber objective-c instance 0x832ed0>'
> (third and fourth lines of code's output).

This is definitely broken.  Both str() and repr() do not do what one 
might expect.

Why doesn't str() invoke -description?  ...or repr()?  __repr__ mapping 
to -description was commented out of _convenience.py.

I "fixed" it and your test prog now spews:

[bumbox:~/bbum-developer/sourceforge/pyobjc] bbum% python foo.py
Increment by 3 to 3
Null decrement to 3
Null increment to 3
Decrement by 3 to 0

Is this correct?

Change committed (along with a unit test).

There is still a bug in NSNumber handling, though:

[bumbox:~/bbum-developer/sourceforge/pyobjc] bbum% python 
Lib/Foundation/test/test_nsnumber.py -v
testAsBool (__main__.TestNSNumber) ... FAIL
testNSNumber (__main__.TestNSNumber) ... ok
testStr (__main__.TestNSNumber) ... ok

======================================================================
FAIL: testAsBool (__main__.TestNSNumber)
----------------------------------------------------------------------
Traceback (most recent call last):
   File "Lib/Foundation/test/test_nsnumber.py", line 34, in testAsBool
     raise AssertionError, "%s is not true"%(a.description())
AssertionError: -0.01 is not true

----------------------------------------------------------------------
Ran 3 tests in 0.035s

FAILED (failures=1)

b.bum

Re: [Pyobjc-dev] [ pyobjc-Bugs-678818 ] Missing NSCFNumber to int conversions

[Just v. R. <ju...@le...>]

bb...@ma... wrote:

> On Saturday, Feb 1, 2003, at 17:16 US/Eastern, SourceForge.net wrote:

Bill, not to be a pain in the neck, but it's really best to add comments
to the bug report itself and not post directly to the list...

Just

Re: [Pyobjc-dev] [ pyobjc-Bugs-678818 ] Missing NSCFNumber to int conversions

[<bb...@ma...>]

On Saturday, Feb 1, 2003, at 18:24 US/Eastern, Just van Rossum wrote:
> bb...@ma... wrote:
>
>> On Saturday, Feb 1, 2003, at 17:16 US/Eastern, SourceForge.net wrote:
>
> Bill, not to be a pain in the neck, but it's really best to add 
> comments
> to the bug report itself and not post directly to the list...

Yes -- I'm actually going to update the bug, but I wanted to open some 
of the discussion here because it involves adding a feature back to the 
bridge that was clearly removed at some point for some reason...

b.bum

Re: [Pyobjc-dev] [ pyobjc-Bugs-678818 ] Missing NSCFNumber to int conversions

[David E. <epp...@ic...>]

On 2/1/03 6:15 PM -0500 bb...@ma... wrote:
> I "fixed" it and your test prog now spews:
>
> [bumbox:~/bbum-developer/sourceforge/pyobjc] bbum% python foo.py
> Increment by 3 to 3
> Null decrement to 3
> Null increment to 3
> Decrement by 3 to 0
>
> Is this correct?

Yes.  Was there a recent change that would have caused the "if x:" to work 
on your system and not on mine?  My most recent cvs update was Jan. 23.

Also, I'm not sufficiently familiar w/the sf bugtracker -- is there an 
address to include in email so that comments like this can get 
auto-attached to the bug?  Or do all comments have to be done through the 
web page?
--
David Eppstein       UC Irvine Dept. of Information & Computer Science
epp...@ic... http://www.ics.uci.edu/~eppstein/

Re: [Pyobjc-dev] [ pyobjc-Bugs-678818 ] Missing NSCFNumber to int conversions

[Ronald O. <ous...@ci...>]

On Sunday, Feb 2, 2003, at 00:15 Europe/Amsterdam, bb...@ma... wrote:

> On Saturday, Feb 1, 2003, at 17:16 US/Eastern, SourceForge.net wrote:
>> 1. if x: should only execute the if clause when x is
>> nonzero.  But if x is a zero NSCFNumber, the clause
>> gets executed erroneously (third line of code's output
>> is "Increment by ..." instead of "Null increment")
>
> I believe this is working correctly on my system?
Probably because I checked in a fix for this just before the weekend 
:-) NSCFNumber evaluates to true if the value returned by 'boolValue' 
does (as do all other objects that have a boolValue method)

>
> [bumbox:~/bbum-developer/sourceforge/pyobjc] bbum% python foo.py
> Increment by 3 to 3
> Null decrement to 3
> Null increment to 3
> Decrement by <NSCFNumber objective-c instance 0x470330> to 0
>
>> 2. str(x) should produce a decimal representation of x,
>> but if x is an NSCFNumber, it instead produces a string
>> like '<NSCFNumber objective-c instance 0x832ed0>'
>> (third and fourth lines of code's output).
>
> This is definitely broken.  Both str() and repr() do not do what one 
> might expect.
>
> Why doesn't str() invoke -description?  ...or repr()?  __repr__ 
> mapping to -description was commented out of _convenience.py.

I have no idea why this was removed. ... Oh, I do know: If you invoke 
-description from __str__ and __repr__ you may cause a coredump if you 
do something like this in an interactive interpreter:
	>>> NSSomeClass.alloc()
	<NSSomeClass instance at 0xabcdabcd>
	>>> _.init()
	<NSSomeClass instance at 0xabcdabcd>

For some values of 'NSSomeClass' the interpreter will crash when trying 
to print the repr of the return value of alloc. The interpreter seems 
to call __repr__, so we should be safe if we only map __str__ to 
description.

Ronald

Re: [Pyobjc-dev] [ pyobjc-Bugs-678818 ] Missing NSCFNumber to int conversions

[<bb...@ma...>]

On Sunday, Feb 2, 2003, at 12:21 US/Eastern, Ronald Oussoren wrote:
> For some values of 'NSSomeClass' the interpreter will crash when 
> trying to print the repr of the return value of alloc. The interpreter 
> seems to call __repr__, so we should be safe if we only map __str__ to 
> description.

Should probably undo that change that I made, then... :-)

b.bum