set, dict and other structures

[Jeremy Bowers]

On Tue, 01 Feb 2005 00:43:21 +0000, Raymond Hettinger wrote:
> My guess is that there will be two issues.  One is that no one
> implementation of graphs seems to satisfy all users.  The second is that
> only a small fraction of Python users need for graph support (there is
> probably a much greater demand for combinatorics, numeric/numarray, or
> financial modules).  If the appeal is not broad, it has little chance of
> making it into the standard library.

This reminded me of a group of people that were supposed to have started
in on this:

http://www.python.org/moin/PythonGraphApi

It's too soon to call it dead, but I don't see a success pattern for it;
I'd say they foundered on trying to shoot for the moon in advance, instead
of starting to write code and incrementally improving it. (Hint, hint, if
any of you are reading this.)

Trying to create a graph library that meets any significant percentage of
the needs of the people who would use it is quite non-trivial. It's not
impossible, but it's a lot lot lot lot lot of work. I remember counting at
least 48 distinct types of graphs when I enumerated my concerns to that
project and I'm sure there are a few more dimensions that didn't make it
into that list (the Wiki talks about "hierachial" graphs which adds
another dimension I didn't count, for instance, and I'm sure there are
more...).