Mark Summerfield schrieb: > On 14 Sep, 21:23, James Stroud <[EMAIL PROTECTED]> wrote: >> Mark Summerfield wrote: >>> I guess I'll have to rename my module (although unfortunately, my book >>> has just gone into production and I have a (simpler) example of what I >>> considered to be an ordered dict, so I will be adding to the >>> terminology confusion). That notwithstanding, given that it is a >>> sorteddict, how is the API? >> I must think the API good because I have been implementing, in parallel >> with this discussion, my own "OrderedDict" with a very similar API (this >> is part of a larger project where I recently found the need to have a >> well-implemented ordered dict). The only real omission I see is to allow >> instantiating a "sorted dict" with an optional cmp function--to allow >> the generalization of such features as case-independence, etc. > > I tend to create different orderings by munging the keys rather than > by having optional cmp functions (as you'll see in the sorteddict.py > docstring). I've now put sorteddict on PyPI (with docs & tests): > http://pypi.python.org/pypi/sorteddict
The advantage of a passed cmp-function is that the context of usage hasn't to be aware of any change in semantics. Yet for better performance, the cmp-function might be a tad to slow, better use a key-function like for sorted/sort. Additionally, I'd change the parameter names of the keys()-method to start/end and make more clear if end is inclusive or not. The first/secondindex-names are pretty horrible IMHO, because the essentially enumerate parameters. Who wants to work with a method foo(arg1, arg2, arg3) ? :) Diez -- http://mail.python.org/mailman/listinfo/python-list
