From: Greg Ewing [mailto:greg@cosc.canterbury.ac.nz]

pinard@iro.umontreal.ca (=?iso-8859-1?q?Fran=E7ois?= Pinard):

...

The advantage is that the `.sorted()' method fits well within how Python has evolved recently, offering more concise and legible writings for frequent idioms.

To support specialised non-in-place sorting algorithms, it could check whether its argument has a sorted() method, and if not, fall back on the general implementation.

Hmm - this actually suggests a couple more magic methods:

__sort__ __isort__

corresponding to "sort a copy" and "sort in-place".

Defining the rules for how these would be called requires a bit more thought however. Do you want a sort() function to prefer __sort__ or __isort__?

def sort (seq, in_place=1):

if in_place: return seq.__isort__()

try: return seq.__sort__() except: pass

seq = list(seq) seq.sort() return seq

So - if an in-place sort is specified, try to do one, throwing an exception if it's not possible. Otherwise sort a copy.

This would allow a generic mechanism for objects to ort copies of themselves, rather than blindly changing them to a list.

Would two methods be better for in-place and copy sort?

Tim Delaney