How about this: why use C when you can use C++?

malune - You can write code in C++, taking advantage of the C++ library, and yet expose a pure C interface to python if you want to. Even if you just write procedural C++, the library can be useful.

You can also write C++ code and use boost::python (hey, thedevdan, see, more Boost goodness [grin]) or SWIG and have them take care of the interfacing.

But other than that, you are right, it's generally easier to use Python to do the high-level design and C (or C++) to do the low-level optimizations than it is to use C++ everywhere. The only difficult point left is the actual interfacing code, and even then, it's not overly complicated, just tedious (and be careful to keep your reference counts right).

As for differences in design, spend some time programming in python and you'll quickly realize how different it is from C++. [grin]

If you have more specific questions, feel free to ask.

Quote:Original post by malune
[EDIT] -- Woops i replied at the same time as you Fruny :P. That's quite interesting, does the interfacing code slow down the overall program? Or does this not add much overhead?

Yes, it slows things down somewhat, but it'll still be faster than 'pure python'. Python is a dynamically-typed language, C++ is a statically-typed language. There are checks and conversions to be done when passing objects from python to C++ which, when done properly, do carry an overhead.

I can only reiterate my prior advice: try it out and see if it suits you. It works fine for me because I tend to spend much more time in C code than doing python/C transitions: I manipulate large multidimensional numeric arrays with the numarray library. So, while each element of the expression triggers a python/C transition, I do spend enough time looping over the array to compensate for that.

And even if you stay in 'pure python', python itself is still implemented in C (or Java, or MSIL - depending on the version you pick up) so, at some level, that 'overhead' is a non-issue. By implementing more complex functionality in C++, you incur that overhead only once instead of for each elementary python operation that may need such a check.

So, get started in python, it really is an easy language to learn, then start delving in python/C (or python/C++) interfacing -- I haven't found the online tutorials on this specific topic to be *that* helpful; much less, at any rate than the corresponding section of "Python in a Nutshell" (handy reference, by the way). Even if you decide not to use it in the end, you'll have learned something new and useful.