Using C struct in Python

[John Machin]

On 8/06/2006 7:35 AM, Sudheer Gupta wrote:
> Hi,

Hi.

Your later "correction" doesn't clear up the confusion below. Quick 
eye-balling revealed no difference. If you have to correct a minor typo 
in a posting, please consider saying "change X to Y" instead of 
reposting the whole thing.

> 
> I am having trouble using C struct in python. Hope anyone can help me 
> out ...
> 
> Say, I have my C struct as
> 
> typedef struct call
> {
>       struct call *next;
>       // .....
> 
> } call_t;
> 
> I have a global variable, namely call_pool, which is of type call_t *

You really need to explain what sort of glue you have between your 
Python code and your C code.

1. Are you the author of the glue? If not, better ask the author.
2. Are you extending Python with a module written in C, or are you 
embedding Python in a C program?
3. If extending, what's your glue? SWIG? something else? hand-crafted?

> 
> My python program:
> 
> cp = call_pool  # no error doing this, means that call_pool is accessable

Acessible from where? How do you bind the name "call_pool" to an object?

> 
> while cp:
>       print cp
>       print cp.next
> 
> 
> This is giving me error: " There is no member or method name c_next"

Not a Python error message. Must be coming from inside your extension 
module.

> 
> Now, If I just do:
> 
> print cp
> print cp.next
> 
> there is no problem.

Sorry, I don't understand. Above you said it was "giving me error".

> But I am seeing a difference in the way python is 
> looking at the struct:
> 
> print cp       ->  (call_t*) 0xb0...
> print cp.next ->  (struct call *) 0xb0...

If they are actual results from a Python print statement, then I can 
only assume the extension module defines types which have (in effect) 
str() and/or repr() methods which produce such output.

Python does not know that the C type of cp is (call_t*) and that of 
cp.next is (struct call *). In fact it doesn't care, and it shouldn't 
care. The extension module could be written in assembly language or APL 
or even INTERCAL if it obeys the conventions, which don't include 
exposing a C type for each object. Another way of looking at it: methods 
in extension modules are mostly expected to behave like methods written 
in Python.

I get the impression that you are using some "superglue" that parses C 
declarations and writes (parts of) Python extension modules in C. Do you 
think you could possibly tell us what the name of this superglue is?

> Is python not intelligent enough to diagnose the next pointer ??


To the extent to which I can understand what your question means, the 
answer is: It by design makes no attempt to be what you are calling 
intelligent.

Python does only limited inspection of the tables of methods that an 
extension type says it supports. It uses only the very basic 
information: is the pointer to method X NULL?

I hope some of the above helps you amplify your question.

Cheers,
John