[python-win32] Announcement: Project to get some CPython C extensions running underIronPython

[Trent Nelson]

Giles,

Just a thought: if you've got some energy you're willing to invest,
perhaps you could redirect it to getting win32com to support the
generation of type libraries for Python classes that are exposed as COM
objects.  If you've got a .tlb available, you can use 'tlbimp' to
generate .NET assemblies, which you could use directly in IronPython
(http://msdn2.microsoft.com/en-us/library/ms973800.aspx).

Disadvantage of this approach, obviously, is that it relies on the
Python class to expose itself as a COM object, which isn't very common
for your average CPython library.  Perhaps proxying the object in some
way to 'fill in' the relevant COM bits needed in order for the type
library to be generated would work...

CC'd python-win32 for additional input...


	Trent.

________________________________

	From: python-announce-list-bounces at python.org
[mailto:python-announce-list-bounces at python.org] On Behalf Of Giles
Thomas
	Sent: 12 October 2007 18:27
	To: python-announce-list at python.org
	Subject: Announcement: Project to get some CPython C extensions
running underIronPython
	
	
	The great thing about CPython is that it comes with the
batteries included.  The problem with IronPython is that some of these
batteries just don't fit - in particular, most of the the C extensions
don't work.  We'd like to help fix at least some of this problem, to
help people who use IronPython to use their CPython scripts without
having to port everything over to .NET.
	
	Solving the general problem - plugging an arbitrary C extension
into IronPython - is a huge project, and we're not even sure we could
work out *how much work it is* without a lot of investigation.  What we
intend to do is to solve a specific problem, to integrate just one
extension, and to use that project as a testbed to examine the
possibilities for getting other extensions working - and perhaps, in the
long term, solving the general problem.
	
	We think that any solution like this will be valuable not just
to us, but to the Python community as a whole.  And so, we want to make
it Open Source.
	
	Right now, we'd really like to hear from people about the
following:
	

	*	Who wants to get involved?  We're really keen on working
with other people on this.
		
	*	Which module should we go for?  NumPy looks like a good
start, as it gives us a start on getting SciPy working.  But perhaps
there are better choices.
		
	*	Should this be a new project, or should we be talking to
other people about getting it into other projects? 
	*	Which license?  If we're to work on it with a view to
building it into Resolver One, then it will need to be
commercial-software-friendly.  Apart from that - we have no view.
		
	*	What is the best architecture?  We're thinking of this
as being a bit of C# managed code to interface with the C extension, and
a thin Python wrapper on top.  The module's existing C extension and
Python code would "sandwich" this layer.  Let us know if this is a silly
idea :-)
		
	*	Is there anything else we should be thinking about to
get this started?
		

	Any thoughts much appreciated!
	
	
	Regards,
	
	Giles 
	-- 
	Giles Thomas
	MD & CTO, Resolver Systems Ltd.
	giles.thomas at resolversystems.com
	+44 (0) 20 7253 6372
	
	We're hiring! http://www.resolversystems.com/jobs/ 
	
	17a Clerkenwell Road, London EC1M 5RD, UK
	VAT No.: GB 893 5643 79 
	Registered in England and Wales as company number 5467329.
	Registered address: 843 Finchley Road, London NW11 8NA, UK