Re: [Pyobjc-dev] shouldSelectRow not working as expected

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

On Friday, Feb 28, 2003, at 12:17 Europe/Amsterdam, Just van Rossum 
wrote:

> Ronald Oussoren wrote:
>
>> Right. Does the patch I just checked in solve the problem?
>
> Partly, it seems. However, you _must_ return an int/bool for methods
> that are declared to return a BOOL; when I return None instead I get an
> exception. I assume that's because ObjC can't see the difference 
> between
> a BOOL and a char at the time the conversion happens.

Right. As far as the runtime, and therefore the bridge, is concerned 
BOOL is an alias for 'unsigned char'.

>
> In my PythonBrowser playground I have an inconsitency, though. Two
> methods are supposed to return BOOLs, yet one is part of the
> NSOutlineViewDataSource protocol, the other is an NSOutlineView 
> delegate
> method (is there even a NSOutlineViewDelegate protocol?).

Not yet :-)

>
> class PythonBrowserModel(AutoBaseClass, NSOutlineViewDataSource):
>
>     [..snippo..]
>
>     def outlineView_isItemExpandable_(self, view, item):
>         # part of the NSOutlineViewDataSource protocol,
>         # must return int/bool
>         if item is None:
>             item = self.root
>         return item.isExpandable()
>
>     def outlineView_shouldEditTableColumn_item_(self, view, col, item):
>         # this is a delegate method of NSOutlineView, I _must_ return
>         # None as "NO", any other object as "YES", hence the "or None"
>         # hack.
>         return item.isEditable() or None
>
> I suppose this is not solvable without adding an NSOutlineViewDelegate
> informal protocol?

Yup. Feel free to add this to Lib/AppKit/__init__.py ;-). It should 
contain all the methods a NSOutlineView delegate can implement.

We should probably add informal_protocol definitions for all delegates 
in Cocoa.

>
> Here's a thought: what if the bridge would convert the False bool
> instance to nil? Hm, this would only work with Python 2.3 unless we add
> proper bool support to 2.2 ourselves (btw. plistlib.py contains code
> that does just that, but only for that module).

That would work only by accident. If the bit-pattern of the pointer to 
the (proxy-)object for a true value happened to contain 8 0-bits at the 
right location it would also be interpreted as NO by the objective-C 
side of thinks.

Not that adding a private bool type on Python 2.2 would be bad, it 
would allow for the creation of the right kind of NSNumber for boolean 
values, NSNumbers created using numberWithBool: have a different kind 
of representation in plist files. BTW. Later 2.2 versions introduce 
builtins for True and False, does that also introduce a bool type or 
are these just names for 1 and 0?

Ronald