Re: sort - C, python, C++, glibc Was: Baby X is bor nagain

On 19/06/2024 12:17, Michael S wrote:

On Wed, 19 Jun 2024 07:44:44 +0200
David Brown <david.brown@hesbynett.no> wrote:
 

On 18/06/2024 18:39, Malcolm McLean wrote:

>
 

My main experience of Python was that we had some resource files
which were icons, in matching light and dark themes. The light
theme had suffix _L followed by extension, and the dark themes had
_D. And they needed to be sorted alphabetically, except that _L
should be placed before _D.
And it didn't take long to get Python to sort the list
alphabetically, but there seemed no way in to the sort comparision
function itself. And I had to give up.
  

>
Python "sort" is a bit like C "qsort" (desperately trying to relate
this to the group topicality) in that you can define your own
comparison function, and use that for "sort".  For simple comparison
functions, people often use lambdas, for more complicated ones it's
clearer to define a function with a name.
>
>

 Off topic:
Indeed, Python sort has option for specifying comparison function, but
I would not call it very similar to comparison function of C qsort
since in Python comparison is not applied directly to the record.

Yes, it seems that the comparison function support in sort() was in Python 2 but was dropped for Python 3.

Instead, it applied to the keys that are derived from record.
Besides, my impression is that in Python sorting by user-supplied
comparison function is less idiomatic than doing all heavy lifting in
user-supplied key function.

A key function can be applied once per item, while a comparison function is called once per comparison - thus key functions will be (or at least /can/ be) more efficient.

For the case, presented by Malcolm, I'd certainly do it all in key(),
without custom cmp_to_key(). May be, it's a little less efficient, but
significantly easier to comprehend.
 Back to topic:
C qsort() sucks. They forgot to provide an option for 3rd parameter
(context) in comparison callback.

It is also not guaranteed to be stable (which is important in some contexts), and it is a misnomer - it is rarely a quicksort.

Back to O.T.:
Python's sort bypasses the problem by allowing lambda as a key()
function, so it could have visibility of variables of the caller. IMHO,
it still sucks.
C++ way, where comparison can be functor, sucks far less. I find it
less than obvious, but at least all functionality one can ever want is
available.

The C++ way is also massively more efficient than C in cases where the comparison function is simple.  And it is typesafe.

 Back to topic: why C standard committee still didn't add something like
gnu qsort_r() to the standard?
 

That is a little more flexible, but it's still ugly!