Hello Stefan,

you wrote:

> Kay Hayen, 21.07.2010 17:57:
>> I can take pure Python 2.5 syntax and compile it to C++ and it
>> does pass most of the CPython tests.
>
> You didn't provide a link to the sources. But why did you start from
> scratch instead of using ShedSkin?

I am not yet ready to release the sources to the world. I want it to be 
actually useful before people note that it's not. And before people see 
relicts of previous attempts and are just confused by it.

Lets say it's not cleaned up well enough, for me to feel comfortable 
with a release right now. Only on a functional level, I have become 
confident.

I think one month is a good time frame. I promise to release it no 
matter what the state will be at the end of August. For private reasons, 
I will have 3 weeks in August, where I can invest significant time, to 
make it a good code drop. Plus, I have not even decided on a name yet.

I evaluated ShedSkin too, and it seemed at the time to have different 
goals and not as much driven by a community either. It seemed to try and 
use C++ standard containers over Python containers to achieve speedups.

Actually I am not at all interested in the C++ part. I am still somewhat 
competent in it as a programming language, but also had to research 
quite a bit on it, to get some things represented. But the use of C++ 
was not at clear to me.

I have also considered to generate Vala or D, or even Ada (which I truly 
love as much as Python for different tasks), but the plan was always 
that this language is not supposed to be important at all, and newer 
languages are evolving themselves too much.

I do not consider that integration of C++ code with Python code is 
useful to me. I want a tool that lets me write in Python and still not 
be "slow". Emphasis on that last part, "to me". I just want something 
faster, that falls back to CPython. I imagine people run the compiler 
over any program, and it's faster, but nothing changed in its behaviour.

For example, I wrote a bit level decoding tool in Python, even in nice 
to read Python, and I want the compiler to make it efficient, based on 
the knowledge the input was a str and what ord(), len(). [] typically do 
to it. Obviously writing it in C would be much more efficient. But also 
much less fun.

In the end C++0x support in recent gcc releases convinced me that C++ is 
a good and useful target language. With some variadic template functions 
e.g., I was able to push a couple of things into the C++ part, where 
previously I would have generated code for it.

I noticed from discussions here that many people do care about C++ and I 
am amazed at some of the things possible with Cython now. It clearly has 
become ever more powerful is this domain. Impressive work you did there 
guys.

Best regards,
Kay Hayen
_______________________________________________
Cython-dev mailing list
[email protected]
http://codespeak.net/mailman/listinfo/cython-dev

Reply via email to