Scott David Daniels wrote: > Delaney, Timothy (Tim) wrote: >> Scott David Daniels wrote: >>> For 2.3: (using DSU -- Decorate, Sort, Undecorate) > ... >>> lst = ['1', '1.2', '1.12', '1.1', '3.1'] >>> decorated = [(numparts(txt), txt) for txt in lst] >>> decorated.sort() lst[:] = [txt for code, txt in decorated] >> >> Slightly better to do:: >> ... >> lst = ['1', '1.2', '1.12', '1.1', '3.1'] >> decorated = [(numparts(txt), i, txt) for i, txt in >> enumerate(lst)] decorated.sort() lst[:] = [txt[-1] for d >> in decorated] > > In the particular case given, I don't think it matters.
True. > In general, however, I agree the latter is a better idea. Just > so someone hopping in here is clear about what is better, and > explicitly _not_ because I think Tim misunderstands the issue, Hey! I resent that! I misunderstand lots of things! <wink> > The latter version, in the case of a match on the "decorated" > version compares first the "key", and then, if the keys match, > it compares the position in the list. In no case does it actually > compare the original list elements. The 2.4 version (just using the > key= form of sort) does the same kind of thing; it avoids comparing > the element and only compares keys. The reason you care about this > is that is can be arbitrarily expensive to compare elements, or the > elements themselves may be "incomparable" (think of complex numbers). Yep - I really should have explained the reasoning better. I guess having been involved in the 2.4 sort discussions on python-dev, it kinda seems obvious to me why only comparing the key is an important feature. Tim Delaney -- http://mail.python.org/mailman/listinfo/python-list
