[Python] Best way of extending Python com or lnk or dll?

"Owen F. Ransen" <ransen_spam_me_not at nemo.it> wrote
i wouldn't do it with static recompiling. most people aren't
too hip on rebuilding their python core just for a new module
(especially on Win and Mac (unix is a little better))

look into the distutils. they have been shipping with python
since 2.0, and now in 2.1 they have support for macos and
metrowerks.

first dipping your toes into distutils can take a bit of learning,
but look for some examples and the docs, and you won't be in any
trouble.
http://www.python.org/doc/current/dist/dist.html

this will build your extension as a shared library, and get
it all installed properly. from there you are all set to go.

Unless you're wrapping existing code, do new work in Python first.
Then if you really, really must, and have no other choice.
Take a look at a 'C' extension.
Often it's not worth the head ach to code it in 'C'.

Someone won't like this, but I'd avoid COM.
Unless you share your components with VB developers or the like.
For reuse in a Python environment .py files are great.

Several times now I've converted 'C' and Java apps to Python.
Once you get rid of the visual noise those languages come with.
A simplified solution may fall out.

Oh, don't I whish C++ required you use this->xyz
--Darrell


"Owen F. Ransen" <ransen_spam_me_not at nemo.it> wrote in message
news:3ae64b92.1979659 at news.newsguy.com...

An old quote, who's origin I've forgotten: "premature
optimization is the root of all evil." Which is to say,
profile your code first, then consider external C functions
for the things which actually consume CPU time.

BTW, to the original poster: the more work your code
can do in C once it gets there, the better. Python carries
significant overhead in dispatching to your external code.
If you external function is meant to do very small amounts
of work over very frequent invocations, external functions
aren't as obvious a choice and the benefits aren't nearly
as dramatic as you predict.

Conversely, things which do very large amounts of work
can push Python programs to at-or-near C speeds when
done right. Calculations in Numeric Python (and friends)
come to mind.

C//

Just had an example of this in this thread "Help me use re better"
Where it was faster to stay in the regular expression engine than to loop
though it.

--Darrell

"Courageous" <jkraska1 at san.rr.com> wrote in message >