RE: [Jython-users] More fun with Readline

[brian z. <bz...@zi...>]

I would ask on comp.lang.python about the preferred method of patching.
Conditionals are used in a number of places to determine the correct
behaviour on a particular os.  A quick grep of the Python 2.2 Lib shows
at a minimum 29 occurences of 'if os.name == X' so it appears to be a
viable solution.

brian

> -----Original Message-----
> From: Steve Cohen 
> [mailto:jyt...@li...] On Behalf 
> Of Steve Cohen
> Sent: Sunday, January 06, 2002 11:19 PM
> To: brian zimmer; Finn Bock; jyt...@li...
> Subject: RE: [Jython-users] More fun with Readline
> 
> 
> Hmm, that won't work with my current implementation.  My 
> current implementation is importing java classes.
> 
> In two places.  In one I am importing PySystemState to read 
> the registry.  Is there an all-python way to read the 
> registry or its properties without referencing jython?  If 
> so, I can avoid that import.
> 
> The other is where I'm importing org.gnu.readline.Readline.  
> This seems to be essential.  I can probably put the import 
> into a conditional code block that would never be met by 
> CPython users but would that be "clean" enough?
> 
> What is standard practice around issues of this type?
> 
> Steve
> 
> 
> 
> 
> 
> -----Original Message-----
> From:	brian zimmer
> Sent:	Sun 1/6/2002 10:36 PM
> To:	Steve Cohen; 'Finn Bock'; jyt...@li...
> Cc:	
> Subject:	RE: [Jython-users] More fun with Readline
> Steve,
> 
> The majority of the python modules in Lib are copied from the 
> same version of the CPython distribution.  Only when the 
> module is new and jython-centric (jreload, dbexts) or 
> drastically differs (string, socket) does it get checked into 
> Jython's CVS.
> 
> For both development and production, I point at the current 
> Jython Lib and then the same version of CPython's Lib.  This 
> makes sure I pick any Jython specific modules first before 
> trying to load the standard Python distribution's modules.
> 
> For something like cmd.py, it's probably best to work with 
> the CPython version and submit your patch there along with 
> the explanation for why it is needed.  Some CPython modules 
> check for os.name == 'java'
> (unittest) and do something accordingly.
> 
> Along this note, there was a change to string.py in CPython 
> 2.2 that includes three attributes not found in the Jython 
> version currently in
> CVS:
> 
>  # not set to 'lowercase' or 'uppercase' because they can 
> change  ascii_lowercase = 'abcdefghijklmnopqrstuvwxyz'  
> ascii_uppercase = 'ABCDEFGHIJKLMNOPQRSTUVWXYZ'  ascii_letters 
> = ascii_lowercase + ascii_uppercase
> 
> Without them, the 2.2 version of cmd.py from (CPython) will 
> not work when imported from Jython 2.1+.  I'm going to raise 
> a bug and provide the patch but you can paste these in your 
> local Jython version of string.py in the interim.
> 
> Hope this helps.
> 
> brian
> 
> 
> > -----Original Message-----
> > From: Steve Cohen
> > [mailto:jyt...@li...] On Behalf 
> > Of Steve Cohen
> > Sent: Sunday, January 06, 2002 9:58 PM
> > To: Steve Cohen; Finn Bock; jyt...@li...
> > Subject: RE: [Jython-users] More fun with Readline
> > 
> > 
> > Well, I developed what I thought was a nice integration of
> > readline into cmd.py.  It worked perfectly and it made no 
> > assumptions about the presence or absence of readline.  
> > 
> > I started preparing a patch file.  It involved a few of the
> > java classes and cmd.py.  All was going well and then I 
> > noticed that cmd.py was NOT in CVS.  Where did it come from 
> > then?  It must have come from the installation.  From the 
> > docs I see that it may have come originally from CPython.  
> > OK, so what does this mean?  Does it mean that my changes 
> > also have to work in CPython?
> > 
> > I'm confused.
> > 
> > 
> > -----Original Message-----
> > From:	Steve Cohen on behalf of Steve Cohen
> > Sent:	Sat 1/5/2002 6:28 PM
> > To:	Finn Bock; jyt...@li...
> > Cc:	
> > Subject:	RE: [Jython-users] More fun with Readline
> > OK.  On one level getting cmd.Cmd to use the readline is
> > easy.  Just import org.gnu.readline.Readline and I'm on my 
> > way.  It works exactly as I want it to work.
> > 
> > But of course, that won't do, since org.gnu.readline is
> > optional.  What I'd like to do is somehow grab the "interp" 
> > InteractiveConsole that is created in jython.main() and call 
> > IT'S raw_input method.  But I don't see a way to do this.  
> > interp is a local variable to the main method.  Is there some 
> > behind the scenes way I can hook this, to take advantage of 
> > its readline capabilities?  That seems like the "right" way 
> > to do this, but is there a way to do it short of modifying 
> > jython.java?  Seems kind of an extreme step for my first 
> > attempt to fix something in jython.
> > 
> > 
> > -----Original Message-----
> > From:	Finn Bock
> > Sent:	Sat 1/5/2002 4:11 PM
> > To:	jyt...@li...
> > Cc:	Steve Cohen
> > Subject:	Re: [Jython-users] More fun with Readline
> > [Steve Cohen]
> > 
> > >There is a stack trace and it's the same whether or not
> > >showJavaExceptions is true:
> > 
> > That is OK. Thanks.
> > 
> > >[scohen@sleepingbear scohen]$ jython
> > >-Dpython.options.showJavaExceptions=3Dtrue
> > >Jython 2.1 on java1.3.0 (JIT: null)
> > >Type "copyright", "credits" or "license" for more information.
> > >>>> import dbexts, isql
> > >>>> d=3Ddbexts.dbexts("prod_sport")
> > >>>> D=3Disql.IsqlCmd(d)
> > >>>> D.use_rawinput=3D0
> > >>>> D.cmdloop()
> > >Traceback (innermost last):
> > >  File "<console>", line 1, in ?
> > >  File "/usr/local/jython/jython-2.1/Lib/cmd.py", line 79, 
> in cmdloop
> > >TypeError: write(): 1st arg can't be coerced to String
> > 
> > Not quite the error situation that I expected, but that just
> > goes to show how important the context is. I have added a bug 
> > report about the situation.
> > 
> > >The 1st arg is None, which is arguably a bug in isql.IsqlCmd.
> > 
> > Are you absolutely sure that self.prompt is None. AFAICT
> > D.prompt should be an instance of the Prompt class and that 
> > will cause the stacktrace above. A None value will cause a 
> > NPE which is a different bug. Please recheck the value of D.prompt.
> > 
> > >The code
> > >in cmdloop can't handle it (can't coerce it to String).  I can work
> > >around the bug by explicitly setting D.prompt (which is the first 
> > >argument passed in) before executing D.cmdloop() but then 
> > you are quite
> > >right, there still isn't support for any form of readline,
> > either the
> > >java_readline that is in the interactive mode or even the Ctrl-N,
> > >Ctrl-P stuff that the documentation speaks of.
> > >
> > >I suppose I could take a look at fixing this, although I'm not
> > >extremely familiar with it.  If you could point me to where the 
> > >java_readline stuff is integrated into the interactive 
> mode, I could 
> > >have a go at it.
> > 
> > That code is located in org\python\util\ReadlineConsole.java.
> > 
> > >Or, if you'd rather have someone more familiar with the 
> internals do
> > >it, I certainly understand.
> > 
> > Well, I prefer that it is implemented by someone who really
> > want it to work. If you decide to try, feel free to ask 
> > questions and seek advice on jython-dev.
> > 
> > regards,
> > finn
> > 
> > 
> > 
> > 
> 
> 
> 
>