Execution speed question
Suresh Pillai
stochashtic at yahoo.ca
Mon Jul 28 03:50:54 EDT 2008
On Fri, 25 Jul 2008 09:22:06 -0600, Matthew Fitzgibbons wrote:
> As for different data structures, it largely depends on how you need to
> access the data. If you don't need to index the data, just loop through
> it, you might try a linked list. The performance hit in (2) is coming
> from the list del; using a linked list would make the removal constant
> rather than O(n), and may even execute faster than (3) as well.
>
> -Matt
Yes, this was my first inclination. So my question, as alluded to in my
original post, is whether there are C compiled modules for linked lists,
doubly linked lists, ordered lists ... (the standard data structures)
somewhere, to get the extra performance out of them.
With python we have all built up creative ways of using the native
structures for efficiency reasons. This project was the first time (due
to its extreme use of resources) that I've had to worry about these
minute considerations of native vs new structure but also take into
account native vs python level construct vs compiled module.
[P.S. The linked list does compare well with (3) as expected.]
More information about the Python-list
mailing list