Re: [Pywebsvcs-talk] Type objects

[Rich S. <rs...@zo...>]

> I don't understand that bit. Under what circumstance would a typecode be
> "heavy"? If I understand typecodes correctly they are basically
> bidirectional functions converting XML to Python and Python to XML. I
> would expect that the bytecodes would overwhelm the associated data in
> almost every circumstance. You are "stuck with" the bytecodes whether
> you have any associated data or nor.


You are probably right.  ZSI was my first large-scale Python module. But 
I was concerned about situations where the typecode wasn't directly or 
easily available, but was only as an attribute within other big objects. 
And also in places where the typecode was generated by a wsdl compiler.


 
> Now I absolutely believe that we need the static type checking stuff as
> infrastructure and for the circumstances that really need it. I just
> don't think it is the use interface that most Python programmers expect.


Sure, and for positional parameters I think things are intuitive.  If 
you wanna place with the ZSI run-time type system, maybe it's 
idiosyncratic (less harsh than un-pythonic :), but maybe not.

> You may be right. But you may not be. Consider the most popular "data
> binding" library in the Perl world is called XML::Simple and it works
> exactly in this way. Building dictionaries is easy for Python
> programmers.



I don't disagree.  In fact, I agree with most of the rest of what you 
wrote, so I deleted it.  That's why I added TC.Any looking for a 
per-object typecode. :)

Now we just need to wrangle the name and calling sequence of cast/wrap 
to the ground.
	/r$
-- 
Zolera Systems, http://www.zolera.com
Information Integrity, XML Security