h2py.py bug?

[Gabriel Rossetti]

Gabriel Genellina wrote:
> En Wed, 11 Jun 2008 04:21:03 -0300, Gabriel Rossetti 
> <gabriel.rossetti at arimaz.com> escribió:
>
>> Gabriel Genellina wrote:
>>> En Tue, 10 Jun 2008 09:44:13 -0300, Gabriel Rossetti 
>>> <gabriel.rossetti at arimaz.com> escribió:
>>>
>>>> I wanted to use the h2py.py script (Tools/scripts/h2py.py) and it 
>>>> didn't like char litterals :
>>>>
>>>> Skipping: PC_ERROR = ord()
>>>>
>>>> where my *.h file contained :
>>>>
>>>> #define PC_ERROR '0'
>>>>
>>>> I searched the web and found a post with the same error :
>>>>
>>>> http://mail.python.org/pipermail/python-list/2005-September/340608.html 
>>>>
>>>>
>>>> but it got no replies, I tried the fix and it works. I have the 
>>>> following questions:
>>>>
>>>> 1) Why did it not get any attention, is something wrong with it?
>>>> 2) If nothing is wrong, did the fix not get applied because a bug 
>>>> report wasn't filed?
>>>
>>> Very probably - bug reports outside the tracker are likely to go 
>>> unnoticed or forgotten.
>>>
>> Ok, I'll file one then.
>>>> 3) Isn't turning a char literal into the ordinal value not contrary 
>>>> to what a C programmer had in mind when he/she defined it? I mean 
>>>> if you define a char literal then in python you would have used a 
>>>> string value :
>>>>
>>>> #define PC_ERROR '0'
>>>>
>>>> would become :
>>>>
>>>> PC_ERROR = '0'
>>>>
>>>> in python, and if you intended to use the char type for an 8 bit 
>>>> numerical value you would have done :
>>>>
>>>> #define PC_ERROR 0x30
>>>>
>>>> where 0x30 is the '0' ascii hex value, so shouldn'it the line in 
>>>> the diff (see the post) be :
>>>>
>>>> body = p_char.sub("'\\1'", body)
>>>>
>>>> instead of :
>>>>
>>>> body = p_char.sub("ord('\\1')", body)
>>>
>>> It's not so clear what's the intended usage - chars are also 
>>> integers in C. (I prefer the current behavior, but certainly it may 
>>> be wrong in several places).
>>>
>> Yes, true, but if you intend to use it as an integer, wouldn't you 
>> use a numeric value instead of a character literal?
>
> Not always. Using characters as ordinal numbers is very common in that 
> language. A small example:
>
> #define FIRST 'A'
> #define LAST  'Z'
> #define index(letter) ((letter)-FIRST)
>
> int accum[LAST-FIRST+1];
> ...
> char x = some_word[0];
> accum[index(x)]++;
>
> accum is an array of 26 integers, element 0 corresponding to letter 
> 'A'. If one has to reproduce the exact same structure, in Python it 
> would be:
>
> FIRST = ord('A')
> LAST = ord('Z')
> def index(letter):
>   return ord(letter)-FIRST
>
> accum = [0 for i in range(LAST-FIRST+1)]
> x = some_word[0]
> accum[index(x)] += 1
>
> Of course, in other cases, character constants are intended to be used 
> as character data, so ord(...) is not the appropiate thing to do when 
> converting to Python. But I don't think my example above is very 
> contrived or uncommon - chars *are* integers from the C point of view, 
> and some people isn't worried too much about the distinction, 
> specially those who use C as their main or only language.
>
Yes, you have a point there, maybe there could be a parameter to switch 
between methods, like keep it as it is by default (for compatibility 
reasons) and use the switch to choose the method I suggested.