pyobjc-dev — CVS commit notices go here

[Pyobjc-dev] more crasher app

[Burris T. E. <bu...@gi...>]

I should mention that the reason each worker has it's own thread is 
because in the real app I'm working on, each worker calls a blocking 
function that doesn't return for hours.  The blocking function takes 
callables that are used as callbacks for display.  Since you can't 
touch appkit objects from any thread but the main thread, the callbacks 
use some mechanism (I've tried performOnMainThread, NSTimer, and DO) to 
pass the display data to the main thread so the appkit objects can be 
updated.

This is the third time I've tried to rewrite this existing ObjC/python 
app (that has over a million downloads, BTW) in pyobjc and it never 
works. I do not believe that pyobjc can be used for multithreaded apps 
at all.  I think my only choice is to run the workers in a subprocess 
and use DO to communicate with the display process, but I'm skeptical 
even that will work.


burris

[Pyobjc-dev] crasher demo app

[Burris T. E. <bu...@gi...>]

Okay, here is a bare bones implementation of what I was describing.

http://www.gigagig.org/BT/mcfoo.tgz

Start this up.  It helps to click on the cells, try to drag them.  
Sometimes it will hang.  Sometimes it will crash like this:

*** malloc[1540]: Deallocation of a pointer not malloced: 0x726df0; 
This could be a double free(), or free() called with the middle of an 
allocated block; Try setting environment variable MallocHelp to see 
tools to help debug
*** malloc_zone_malloc[1540]: argument too large: 4294967292

mcfoo has exited due to signal 10 (SIGBUS).




thanks,
burris

[Pyobjc-dev] =?iso-2022-jp?B?GyRCJWIlUCUkJWslaSVzJS0lcyUwJTUlJCVIPzdAXyROJCpDTiRpGyhC?= =?iso-2022-jp?B?GyRCJDshKiEqISobKEI=?=

[<a0...@bb...>]

=1B$B!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A=1B(B
=1B$B%5%$%H4IM}<TMM$X=1B(B

=1B$B$*K;$7$$=3Dj!"BgJQ<:Ni$$$?$7$^$9!#=1B(B

=1B$B$"!!$-!!$H?=3D$7$^$9!#=1B(B

=1B$B%i%s%-%s%0$N?75,3+@_$KEv$?$j!"=1B(B
=1B$B5.%5%$%HMM$N$h$&$JM%NI%5%$%HMM$K;22C$$$?$@$-$?$/=1B(B
=1B$B%a!<%k$rCW$7$^$7$?!#=1B(B

=1B$B!V%9!<%Q!<%i%s%-%s%0!W=1B(B
http://www.hotclick.tv/24781

=1B$B$*<j;}$A$N%5%$%H$,$"$j$^$7$?$i!"=1B(B
=1B$B$44uK>$N%8%c%s%k$K$I$s$I$sEPO?$7$F$/$@$5$$$M!#=1B(B
=1B$B$b$7=3DEJ#Aw?.Ey8f:B$$$^$7$?$i$*OM$S?=3D$7>e$2$^$9!#=1B(B
=1B$B$^$?!"$3$N%i%s%-%s%0$X$NM6$$$r<u$1<h$j$?$/$J$$$H$*9M$($N4IM}<TMM$O!"=
=1B(B
=1B$B$=3D$N;]$rJV?.$7$F$/$@$5$$!#=1B(B
=1B$B:G8e$^$GFI$s$G$$$?$@$$$FM-Fq$&$4$6$$$^$7$?!#=1B(B
=1B$B!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A!A=1B(B

Re: [Pyobjc-dev] tableview crash

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

On 19-mrt-04, at 7:30, b.bum wrote:

> On Mar 18, 2004, at 9:45 PM, Burris T. Ewell wrote:
>> I think I'm running latest CVS and built in 2.3 on 10.3.2
>
> (0) It sounds like you are tripping over some threading weirdness....
>
> ... but before you go down that route ....
>
> (1) make sure you are leveraging Cocoa Bindings as much as possible.   
> It will eliminate code -- one less line of code is one less potential  
> bug.
>
> http://developer.apple.com/documentation/Cocoa/Conceptual/ 
> CocoaBindings/index.html
>
> (2) have a look at the Twisted+WebServices+ControllerLayer example  
> found in the examples.   It does multithreaded filling of a table view  
> and I haven't encountered a crash....

(3) try to reproduce your crash in a minimal program

I'd like to know what causes this crash, even when it turns out to be a  
bug in your script. If it is a bug in PyObjC it should be fixed, and  
that is not likely to happen unless you either send a patch or a  
program that reproduces your problem.

Ronald
--
X|support bv            http://www.xsupport.nl/
T:  +31 610271479       F:  +31 204416173

Re: [Pyobjc-dev] tableview crash

[Bob I. <bo...@re...>]

On Mar 19, 2004, at 1:30 AM, b.bum wrote:

> On Mar 18, 2004, at 9:45 PM, Burris T. Ewell wrote:
>> I think I'm running latest CVS and built in 2.3 on 10.3.2
>
> (0) It sounds like you are tripping over some threading weirdness....
>
> ... but before you go down that route ....
>
> (1) make sure you are leveraging Cocoa Bindings as much as possible.   
> It will eliminate code -- one less line of code is one less potential  
> bug.
>
> http://developer.apple.com/documentation/Cocoa/Conceptual/ 
> CocoaBindings/index.html
>
> (2) have a look at the Twisted+WebServices+ControllerLayer example  
> found in the examples.   It does multithreaded filling of a table view  
> and I haven't encountered a crash....

It's only technically multithreaded because CoreFoundation uses an  
additional thread behind the scenes to do a select loop.  Python only  
knows about the event loop thread, so I would consider it single  
threaded.

-bob

Re: [Pyobjc-dev] tableview crash

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

On Mar 18, 2004, at 9:45 PM, Burris T. Ewell wrote:
> I think I'm running latest CVS and built in 2.3 on 10.3.2

(0) It sounds like you are tripping over some threading weirdness....

... but before you go down that route ....

(1) make sure you are leveraging Cocoa Bindings as much as possible.   
It will eliminate code -- one less line of code is one less potential  
bug.

http://developer.apple.com/documentation/Cocoa/Conceptual/ 
CocoaBindings/index.html

(2) have a look at the Twisted+WebServices+ControllerLayer example  
found in the examples.   It does multithreaded filling of a table view  
and I haven't encountered a crash....

b.bum

[Pyobjc-dev] tableview crash

[Burris T. E. <bu...@gi...>]

Forgive me, my Cocoa is pretty rusty...  I have a single column  
NSTableView using an NSCell subclass.  The cell gets a reference to a  
view (a box of regular controls) in setObjectValue_  Each row has a  
worker thread, started with start_new_thread, that posts a method and  
arguments onto a queue.  I have an NSTimer going off in the main thread  
that pulls the methods off the queue, which update the views that are  
displayed in the cell.

When there is more than one row in the TV I get a strange crash shortly  
after selecting rows in quick succession or in dragging on a row (the  
TV is set to single selection and no rearranging of rows).  Usually in  
the cell's drawInteriorWithFrame_inView_ the attribute containing the  
content view will disappear mid-method.  Or an extra argument will show  
up in the NSTimer fired event loop.  Now I'm getting "Fatal Python  
error: GC object already tracked"

I think I'm running latest CVS and built in 2.3 on 10.3.2

Any help would be appreciated!


interesting bits from cell:

     def setObjectValue_(self, controller):
         self.controller = controller

     def drawInteriorWithFrame_inView_(self, frame, view):
         if self.controller.cview.superview() != view:
             view.addSubview_(self.controller.cview)
         if self.controller.resized:
             self.controller.resized = 0
             self.controller.cview.setFrame_(frame)
         self.controller.cview.lockFocus()
         self.controller.cview.drawRect_(frame)
         self.controller.cview.unlockFocus()

interesting bits from the controller:

     def handleEvents_(self, timer):
         while self.events:
             self.event_lock.acquire()
             f, args, kwargs = self.events.pop()
             self.event_lock.release()
             try:
                 f(*args,**kwargs)
             except:
                 print_exc()
         if not self.flag.isSet():
                  
NSTimer.scheduledTimerWithTimeInterval_target_selector_userInfo_repeats_ 
(0.2, self, self.handleEvents_, self, 0)


burris

Re: [Pyobjc-dev] GetMovieTime() from quicktime movie

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

On 17-mrt-04, at 18:58, Birch Browning wrote:

> Please forgive me if I've gone about this the wrong way. I'm still 
> quite
> a newbie to pyobjc. I think I know just enough to be dangerous.

You're lucky ;-), I thought you ran into a PyObjC bug and I wanted to 
fix that.


The attached archive contains two python files: main.py is the 
translation of the Objective-C source-code into python, buildapp.py can 
be used to build the application.

Copy both files into the Objective-C example tree and then run 'python 
buildapp.py build'.

This should create 'build/NSMovie_Example.app', which seems fully 
functional.

[Pyobjc-dev] GetMovieTime() from quicktime movie

[Birch B. <b.b...@cs...>]

Please forgive me if I've gone about this the wrong way. I'm still quite
a newbie to pyobjc. I think I know just enough to be dangerous.

I'm developing a research application for evaluating pre-service
teachers (I have to get tenure!). The application is all set with the
exception of extracting time data from the quicktime file. I've seen it
done in cocoa/objc (see nsmovie_example at
www.macdevcenter.com/pub/a/mac/2003/08/15/quicktime_cocoa.html) but I
can't quite figure out how to translate this to pyobjc.

If I do:
	m = self.movie.QTMovie()
	s = str(m)
	self.results.insertText_(s)
I can see that m is a QT.Movie object. Doing this, however, also
promptly crashes the application.

I've tried adding Movies.h to the nib file, but the NibClassBuilder
doesn't seem to recognize it. Consequently, 'import Movies', 'import
QuickTime' and 'from Quicktime import Movies' all fail.

Any suggestions, including help translating the nsmovie_example
referenced above, would be greatly appreciated.

Birch

--
Birch P. Browning, Ph.D.
Assistant Professor of Music Education
Cleveland State University

[Pyobjc-dev] The NEW ISSUE STOCK I was telling you about, Mon, 15 Mar 2004 00:41:28 -0500

[Sherman W. <alb...@zo...>]

Mon, 15 Mar 2004 00:41:28 -0500

Here is that new issue stock.  Looks like it has now started to make its move.  http://otcbb.com/asp/quotes.asp?Quotes=pwrm&Bullet.x=85&Bullet.y=5

Sherman Washington

Re: [Pyobjc-dev] Calling Python code from Cocoa bundle

[<li...@po...>]

On 13.03.2004, at 18:44, Bob Ippolito wrote:
> On Mar 13, 2004, at 6:21 PM, Ralph P=F6llath wrote:
>
>> I'm a python newbie and I'm trying to call a python method from=20
>> within a Obj-c Cocoa bundle.
>>
>> I've worked around the "undefined symbol" problem
>> ( http://sourceforge.net/mailarchive/message.php?msg_id=3D7358512 ),
>> but python doesn't find my abstract Obj-C class and therefore won't=20=

>> let me subclass it:
>
> Python's module namespaces are mutable, but "stupid" and "fixed". =20
> Basically, it won't search for additional things if your class is not=20=

> already in that namespace.  Additionally, every module gets its own=20
> namespace, very much unlike Obj-C where namespaces only exist by=20
> "NSConvention".  This is one way to add your Abstract class to the=20
> current namespace, and it should work so long as your Abstract class=20=

> is already loaded somehow by the Obj-C runtime:
>
> import objc
> Abstract =3D objc.lookUpClass('Abstract')

Works like a charm. Thanks, Bob!

Cheers,
-Ralph.

Re: [Pyobjc-dev] Calling Python code from Cocoa bundle

[Bob I. <bo...@re...>]

On Mar 13, 2004, at 6:21 PM, Ralph P=F6llath wrote:

> I'm a python newbie and I'm trying to call a python method from within=20=

> a Obj-c Cocoa bundle.
>
> I've worked around the "undefined symbol" problem
> ( http://sourceforge.net/mailarchive/message.php?msg_id=3D7358512 ),
> but python doesn't find my abstract Obj-C class and therefore won't=20
> let me subclass it:

Python's module namespaces are mutable, but "stupid" and "fixed". =20
Basically, it won't search for additional things if your class is not=20
already in that namespace.  Additionally, every module gets its own=20
namespace, very much unlike Obj-C where namespaces only exist by=20
"NSConvention".  This is one way to add your Abstract class to the=20
current namespace, and it should work so long as your Abstract class is=20=

already loaded somehow by the Obj-C runtime:

import objc
Abstract =3D objc.lookUpClass('Abstract')

-bob

[Pyobjc-dev] Calling Python code from Cocoa bundle

[<li...@po...>]

Hi,

I'm a python newbie and I'm trying to call a python method from within 
a Obj-c Cocoa bundle.

I've worked around the "undefined symbol" problem
( http://sourceforge.net/mailarchive/message.php?msg_id=7358512 ),
but python doesn't find my abstract Obj-C class and therefore won't let 
me subclass it:

> Traceback (most recent call last):
>   File "MyBundle.bundle/Contents/Resources/PythonGlue.py", line 36, in 
> ?
>     main()
>   File "MyBundle.bundle/Contents/Resources/PythonGlue.py", line 30, in 
> main
>     __import__(filename[:-3])
>   File "MyBundle.bundle/Contents/Resources/Test.py", line 7, in ?
>     class Test (Abstract):
> NameError: name 'Abstract' is not defined

Here's what I have:

PythonGlue.h, PythonGlue.m, PythonGlue.py from 
pyobjc-1.0/Doc/tutorial_embed/src/

### file: Abstract.h
@interface Abstract : NSObject
+ instantiateSubclassNamed: (NSString *) aClassName;
@end

@interface Abstract (Test)
- (void) hello;
@end

### file: Abstract.m
@implementation Abstract
  + instantiateSubclassNamed:(NSString*) aClassName
  {
     Class aClass = NSClassFromString(aClassName);
     return [[aClass alloc] init];
  }
  @end

### file: Test.py
from objc import YES, NO
from Foundation import *
from AppKit import *

DEBUG = 1

class Test (Abstract):
     def init(self):
	if DEBUG:
	    print "Test.init() called"
	return self
	
     def hello_(self, obj):
	if DEBUG:
	    print "Test.hello() called"

And in my bundle's main class:

     [[PythonGlue alloc] init];
     id test = [Abstract instantiateSubclassNamed:@"Test"];
     [test hello];


Thanks in advance,
-Ralph.

[Pyobjc-dev] Try this 6 Mega Pixel DigiCam for Free. nGWFDf[

[e-ticket-ads.com <no...@e-...>]

No text version was provided

[Pyobjc-dev] =?GB2312?B?16Ky4c/juNu5q8u+t9HTw9a70Og1NjAw1KrIy8PxsdI=?=

[<hkr...@to...>]

以下是商业推广邮件，如你暂时未感兴趣，请保存或转交有需要的朋友。
如你不想接收有关商业推广邮件，请发电邮至rem...@to...，标题为退订。


注册成立香港有限公司全套费用只需港币5200元 / 人民币5600元
（包括注册地址、公司秘书、电话、传真、代收及代转邮件等配套服务）

注册成立香港出版社 / 杂志社 / 协会 / 研究中心 / 学院 
全套费用只需港币4600元 / 人民币4900元
（包括地址、秘书、电话、传真、代收及代转邮件等配套服务）

注册香港商标全套费用只需港币3900元 / 人民币4200元
（包括商标查询、政府申请费用、公告费用、商标证书费用及服务费用）

申领国际标准书号（ISBN）费用：每家出版社：港币1500元 / 人民币1600元
（每家出版社可领的国际书号无配额限制，每次可获分配十个书号，用完免费再领十个书号）

申领国际标准期刊号（ISSN）费用：每本期刊：港币1500元 / 人民币1600元


您好!

香港公司注册网
专业办理各地客户于香港注册成立有限公司、出版社、杂志社、协会、研究中心、学院等，
注册香港商标、申领国际标准书号（ISBN）、申领国际标准期刊号（ISSN）等，
并提供地址、秘书、电话、传真、带收及代转邮件等各种配套服务，
使各地客户无需亲临香港亦可用遥控方式控制及运作一间在香港注册登记的公司，
助您拓展业务、免交香港企业所得税、将外汇合法留存。

注册成立香港公司的好处：

经营范围不受限制
注册资金无需验资
公司名称自由选择
于香港开立银行户口、接受L/C（信用证）、接受外币电汇、境外支票、
将外汇合法存放
通过电话银行、网上银行服务遥控香港银行帐户
于国内外资银行开立离岸帐号，办理L/C（信用证）、外币电汇、并可提取现金、转帐等
通过电话银行、传真银行、网上银行等服务遥控银行帐户
税率低，只需一年申报企业所得税一次，无需缴纳营业税、增值税等
香港以外业务免交香港企业所得税
货物可自由进出口、存仓，无需申领许可证
以香港公司名义注册商标、专利、域名、出版社、国际书号、国际期刊号，不受限制

由专业律师办理
手续简单、快捷
办理时间只需12个工作天
国内申请人只需提供姓名及国内身份证号码

详情请访问：http://www.hkrnet.net　香港公司注册网   http://www.hkrnet.org

联系我们 :
香港公司注册网
（由新年代信息咨询有限公司主办）
中国办事处：
地址:  深圳市福田区深南中路2号新闻大厦3604室
电话:  0755 - 82092416 　 或   0755 - 82092420　
传真:  0755 - 82092412				
手机:　13510362845	
联系人: 吴先生

电邮:  ma...@hk...     mailto:ma...@hk...

电邮:  ma...@hk...     mailto:ma...@hk...

　　　　致
礼!
  　　　　　　　　　　　　
香港公司注册网

详情请访问：http://www.hkrnet.org  香港公司注册网  http://www.hkrnet.net

电邮请回复：ma...@hk...      mailto:ma...@hk...

电邮请回复：ma...@hk...      mailto:ma...@hk...

[Pyobjc-dev] make_distrib.py

[Pierce T.W. I. <pi...@tw...>]

This code:

  Lines 149-162 in make_distrib.py

if osvers != '10.2':
     # Move the python packages into the right location, 'site-packages'  
is a
     # symlink on Panther and we don't want to clobber that.
     makeDir(BUILDDIR, 'package', 'Library', 'Python', str(PYTHONVER))
     os.rename(
         '%s/package%s/lib/python%s/site-packages/PyObjC'%(BUILDDIR,  
basedir, PYTHONVER),
         '%s/package/Library/Python/%s/PyObjC'%(BUILDDIR, PYTHONVER))
     os.rename(
         '%s/package%s/lib/python%s/site-packages/PyObjC.pth'%(BUILDDIR,  
basedir, PYTHONVER),
         '%s/package/Library/Python/%s/PyObjC.pth'%(BUILDDIR, PYTHONVER))

     # Trash /System/Library/Frameworks/Python.framework in the  
to-be-installed
     # tree, we don't want to replace those directories!
     shutil.rmtree('%s/package/System'%(BUILDDIR,))

  Wasn't working for me. I ended up just commenting it out because it  
seems to be trying to move things from the old places to the new  
places?

  That is, the first rename was trying to do:

os.rename("pyobjc1.1b2/package/System/Library/Frameworks/ 
Python.framework/Versions/2.3/lib/python2.3/site-packages/PyObjC",
"pyobjc1.1b2/package/Library/Python/2.3/PyObjC")

and failing because package/System/etc. didn't exist.

  Perhaps it should be:

if osvers != '10.3'?

Re: [Pyobjc-dev] NSDecimalNumber to the next level

[Pierce T.W. I. <pi...@tw...>]

On Mar 10, 2004, at 12:16 PM, b.bum wrote:

> On Mar 10, 2004, at 11:08 AM, Pierce T.Wetter III wrote:
>>  How so? NSNumbers are still an object, and I think that NSDecimal 
>> actually only uses the minimum amount of memory necessary.
>>
>>  Remember this would only be necessary when passing a python number 
>> back to objective c.
>
> NSNumber & NSValue are designed as a minimal encapsulation of a C 
> scalar type.  NSDecimalNumber can represent any number that can be 
> expressed as mantissa x 10 exponent where mantissa is a decimal 
> integer up to 38 digits long, and exponent is an integer between -128 
> and 127.

   Yes, that's true, but not relevant, because we're going the right 
direction in that we're converting a python type to representation that 
is larger then what it cares about.

  There are two issues to my mind:

Memory:

NSNumber(int) is basically

  struct {
	objc_object *_isa;
      int value;
  }

  While NSDecimalNumber is

  struct {
      objc_object *_isa;
      NSDecimal dcvalue;
  }

  Except NSDecimal is a variable size structure, so while its larger 
then sizeof(int), the size of any number representable by by an int 
turns out to be smaller then then the maximum NSDecimal size.

Performance:

   [NSNumber(double) doubleValue]
   [NSNumber(int) doubleValue]

    is slower then

   [NSDecimalNumber doubleValue]

But since you're already doing a method call and going into/out of 
python, I don't think its that big a deal...


>
> Though, given that this will only come into play when passing a python 
> number back to Obj-C, the performance penalty will be a lot 
> smaller....

Here are the changes.

objc_support: lines 1342:1356

WAS:

		} else if (PyObjCBool_Check(argument)) {
			*(id *) datum = [NSNumber
				numberWithBool:PyInt_AS_LONG (argument)];
		} else if (PyInt_Check (argument)) {
			*(id *) datum = [NSNumber
				numberWithLong:PyInt_AS_LONG (argument)];
		} else if (PyFloat_Check (argument)) {
			*(id *) datum = [NSNumber
				numberWithDouble:PyFloat_AS_DOUBLE (argument)];
		} else if (PyLong_Check(argument)) {
			/* XXX: What if the value doesn't fit into a
			 * 'long long'
			 */
			*(id *) datum = [NSNumber
				numberWithLongLong:PyLong_AsLongLong(argument)];

NOW:

		} else if (PyObjCBool_Check(argument)) {
			*(id *) datum = [NSDecimalNumber
				numberWithBool:PyInt_AS_LONG (argument)];
		} else if (PyInt_Check (argument)) {
			*(id *) datum = [NSDecimalNumber
				numberWithLong:PyInt_AS_LONG (argument)];
		} else if (PyFloat_Check (argument)) {
			*(id *) datum = [NSDecimalNumber
				numberWithDouble:PyFloat_AS_DOUBLE (argument)];
		} else if (PyLong_Check(argument)) {
			/* XXX: What if the value doesn't fit into a
			 * 'long long'
			 */
			*(id *) datum = [NSDecimalNumber
				numberWithLongLong:PyLong_AsLongLong(argument)];

Re: [Pyobjc-dev] NSDecimalNumber to the next level

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

On Mar 10, 2004, at 11:08 AM, Pierce T.Wetter III wrote:
>  How so? NSNumbers are still an object, and I think that NSDecimal 
> actually only uses the minimum amount of memory necessary.
>
>  Remember this would only be necessary when passing a python number 
> back to objective c.

NSNumber & NSValue are designed as a minimal encapsulation of a C 
scalar type.  NSDecimalNumber can represent any number that can be 
expressed as mantissa x 10 exponent where mantissa is a decimal integer 
up to 38 digits long, and exponent is an integer between -128 and 127.

Though, given that this will only come into play when passing a python 
number back to Obj-C, the performance penalty will be a lot smaller....

Re: [Pyobjc-dev] NSDecimalNumber to the next level

[Pierce T.W. I. <pi...@tw...>]

>>  Hmmmm.... Anyone mind if PyObjC just build NSDNs instead of 
>> NSNumbers?
>
> They are a lot heavier weight.... a lot.

  How so? NSNumbers are still an object, and I think that NSDecimal 
actually only uses the minimum amount of memory necessary.

  Remember this would only be necessary when passing a python number 
back to objective c.

  (BTW, this works for me so far.)

  Pierce

Re: [Pyobjc-dev] Documentation generation, Xcode templates, locating NibClassBuilder

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

On Mar 10, 2004, at 8:36 AM, Donovan Preston wrote:
> 2) I would like to update the Xcode templates using Bob's new PyMacApp 
> main.m (I don't know if he announced it here, but he announced it on 
> the MacPython mailing list). How many varieties of templates should I 
> produce? The project builder templates had a bewildering variety of 
> project types you could create... Also, should I look into updating 
> the Project Builder templates, or is that too much of a can of worms 
> to try to open. I don't have 10.2 installed but I suppose I could set 
> up a 10.2 testing environment.

We need two templates:

Cocoa-Python Application
Cocoa-Python Document-Based Application

There is already a [mostly] working Cocoa-Python Application template 
in CVS.   It needs some refinement.   In particular, integrating Bob's 
new PymacApp main.m and general cleanup.   The Document-Based template 
is not really working yet.

In both cases, comparing to the "latest and greatest" templates as 
shipped by Apple would be useful to ensure that various settings, etc, 
are as close to the "pure ObjC Cocoa" templates as possible.   There is 
a python script included in the template tree that can filter an Xcode 
template into a normal project and back again;  that is, an Xcode 
template is just a regular project with markers throughout indicating 
where various programmatic substitutions should be made.

The one key difference between the default templates and the PyObjC 
Xcode templates is that I pushed keyword substitution into the NIB 
files, too [using the 10.2 and later text format].  The big win is NO 
MORE MyDocument.{nib,h,m}!!

The Cocoa-Python-ObjC template was simply a workaround for the lack of 
a linkable python runtime on 10.2.   It allowed the developer to push 
the ObjC code into a framework that was dyna-loaded at runtime.... no 
longer necessary.

b.bum

Re: [Pyobjc-dev] NSDecimalNumber to the next level

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

On Mar 10, 2004, at 9:05 AM, Pierce T.Wetter III wrote:
>  Hmmmm.... Anyone mind if PyObjC just build NSDNs instead of NSNumbers?

They are a lot heavier weight.... a lot.

Re: [Pyobjc-dev] Documentation generation, Xcode templates, locating NibClassBuilder

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

On 10-mrt-04, at 17:36, Donovan Preston wrote:

> A couple of questions about documentation and environment setup for 
> PyObjC 1.1:
>
> 1) I started reading the pyobjc documentation in earnest again last 
> night, and found myself editing it for spelling. I plan on reading 
> every last word and touching the documentation up, but I need to know 
> if there is a script I should run to generate the HTML before checking 
> in?

Thank you for proof reading. You don't have to generate the HTML, I'll 
do that before building the release files. If you really want to you 
can use Scripts/make_distrib.py, but that needs docutils and DocArticle 
(from the docutils sandbox).

>
> 2) I would like to update the Xcode templates using Bob's new PyMacApp 
> main.m (I don't know if he announced it here, but he announced it on 
> the MacPython mailing list). How many varieties of templates should I 
> produce? The project builder templates had a bewildering variety of 
> project types you could create... Also, should I look into updating 
> the Project Builder templates, or is that too much of a can of worms 
> to try to open. I don't have 10.2 installed but I suppose I could set 
> up a 10.2 testing environment.

I've seen his announcement, but didn't look at it yet. We need at most 
two Xcode templates, one of simple applications and one for 
document-based applications. The 'Cocoa-Python-ObjC' variants are not 
relevant for Bob's PyMacApp.

>
> 3) Finally, the tutorial has some very ugly instructions about how to 
> locate NibClassBuilder in a variety of situations. Is there any way 
> this situation could be simplified? Perhaps the initial tutorial 
> should instruct users on how to build an application using an Xcode 
> template, and then only later introduce the shell commands for doing 
> the same. I know a shell script which invokes NibClassBuilder is 
> installed in the python bin directory, but this is not on the $PATH by 
> default. Should the installer make a symlink to this script into 
> somewhere that is on the $PATH?

nibclassbuilder is installed in /usr/local/bin when you use the binary 
installer. I'm going to create a PackMan database that will do the 
same. IMHO PackMan should automaticly use /usr/local/bin when 
installing source packages (this is easy enough when you use 
distutils).

>
> In general I would like to see the introductory documentation and 
> environment setup situation improved in time for the final 1.1, and I 
> am willing to spend the time to make this happen.

Thank you!

Ronald
--
X|support bv            http://www.xsupport.nl/
T:  +31 610271479       F:  +31 204416173

Re: [Pyobjc-dev] Using python classes in objc...

[Jiva D. <ji...@de...>]

That is exactly what I was looking for!

Thank you so much!

On Mar 10, 2004, at 10:19 AM, b.bum wrote:

> On Mar 10, 2004, at 9:01 AM, Jiva DeVoe wrote:
>> How do I get around this problem?  BBum has mentioned using a factory 
>> method in an abstract superclass.  Ok, that's fine, but even then, 
>> the factory method still has to instantiate the python class.  How 
>> does it do so without linking the python code?
>
> Do something like:
>
> @implementation MyFactoryAbstractObjCClassyThing
> + createAConcreteInstance
> {
> 	Class aClass = 
> NSClassFromString(@"ConcretePythonSubclassOfThisClass"); // only works 
> if you had previously imported the appropriate code in python
>
> 	return [[aClass alloc] init];
> }
>
> To avoid compiler warnings when messaging the class, you can:
>
> - declare the methods on a category of the abstract super.   
> Categories do not have to have an associated implementation.  
> Alternatively, you can provide abstract implementations of the methods 
> that do something like:
>
> - (void) methodThatIsReallyImplementedInPython
> {
> 	@throw [NSException exceptionWithName: 
> NSInternalInconsistencyException reason: @"Subclass must implement 
> this." userInfo: nil];
> }
>
> - use -performSelector:, -performSelector:with: and friends to 
> indirectly invoke the method
>
> - make all your functionality Key-Value Coding and use Bindings and/or 
> -setValue:forKey: and -valueForKey: (not really practical for anything 
> but data passing)
>
> - use NSInvocation to build up the method invocation
>
> I generally go with the first option:
>
> @interface Abstract:NSObject
> + instatiateSubclassNamed: (NSString *) aClassName;
> @end
>
> @interface Abstract(MethodsThatMustBeInPythonSubclass)
> ...
> @end
>
> Then, in python:
>
> class Foo(Abstract)
> 	def ....:
>
>
--
Jiva DeVoe
jiva at devoesquared.com
http://www.devoesquared.com

Re: [Pyobjc-dev] Using python classes in objc...

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

On Mar 10, 2004, at 9:01 AM, Jiva DeVoe wrote:
> How do I get around this problem?  BBum has mentioned using a factory 
> method in an abstract superclass.  Ok, that's fine, but even then, the 
> factory method still has to instantiate the python class.  How does it 
> do so without linking the python code?

Do something like:

@implementation MyFactoryAbstractObjCClassyThing
+ createAConcreteInstance
{
	Class aClass = 
NSClassFromString(@"ConcretePythonSubclassOfThisClass"); // only works 
if you had previously imported the appropriate code in python

	return [[aClass alloc] init];
}

To avoid compiler warnings when messaging the class, you can:

- declare the methods on a category of the abstract super.   Categories 
do not have to have an associated implementation.  Alternatively, you 
can provide abstract implementations of the methods that do something 
like:

- (void) methodThatIsReallyImplementedInPython
{
	@throw [NSException exceptionWithName: 
NSInternalInconsistencyException reason: @"Subclass must implement 
this." userInfo: nil];
}

- use -performSelector:, -performSelector:with: and friends to 
indirectly invoke the method

- make all your functionality Key-Value Coding and use Bindings and/or 
-setValue:forKey: and -valueForKey: (not really practical for anything 
but data passing)

- use NSInvocation to build up the method invocation

I generally go with the first option:

@interface Abstract:NSObject
+ instatiateSubclassNamed: (NSString *) aClassName;
@end

@interface Abstract(MethodsThatMustBeInPythonSubclass)
...
@end

Then, in python:

class Foo(Abstract)
	def ....:

Re: [Pyobjc-dev] NSDecimalNumber to the next level

[Pierce T.W. I. <pi...@tw...>]

> You're using the wrong version of PyObjC:
>

  Testing my use cases then, only one fails:

print Foundation.NSDecimalNumber.zero().decimalNumberByAdding_(4.4)

  Which I'll go dig into, since that's a float -> NSDN issue.

  Hmmmm.... Anyone mind if PyObjC just build NSDNs instead of NSNumbers?

  Pierce