Idiosyncratic python

jmp jeanmichel at sequans.com
Thu Sep 24 20:04:17 CEST 2015


On 09/24/2015 04:26 PM, Ian Kelly wrote:
> On Thu, Sep 24, 2015 at 8:07 AM, jmp <jeanmichel at sequans.com> wrote:
>> result = getResult()
>>
>> For the later, the original weird form come from a C habit to allocate
>> returned structures within the caller and provide a pointer to it so the
>> function can fill the data in, otherwise the structure is lost as the stack
>> is popped out and the structure content is garbage. None of this make any
>> sense in python.
>
> Only if the structure is allocated on the stack and returned by
> pointer. If it's returned by value, then the content remains intact,
> but at the expense of copying it. The other option of course would be
> to allocate it on the heap and return the pointer, but this needlessly
> incurs malloc overhead and creates an opportunity for a memory leak if
> the variable is naturally stack-scoped. Python effectively takes this
> option, as everything is allocated on the heap. Leaving it up to the
> caller to provide a pointer also gives the caller the option of
> allocating on the stack or the heap as best fits the context.
>

I'm not an expert but I think this "return by value thing" is only for C++.
In vintage C, you can only return something that fits within a register.

Anyway, there's a lot of legit C code in which functions are plagued by 
'out parameters' and somehow it has transpired in some python code for 
no reason :o)

jm



More information about the Python-list mailing list