result=[]
things=listAllThings()
for t in things:
   if t.startswith(key): res.append(t)
for t in things:
   if key in t: res.append(t)

Going through the list twice makes sure I get all things that match well first (i.e. the start with the string I complete for), and then things matching less well later (they only contain the string).

Of course the world has made up a far better data-structure for indexing prefix’es of string, namely the trie, or prefix tree. James Tauber had already implemented one in python, and kindly made it available. His version didn’t do everything I needed, so I added a few methods. Here is my updated version:

http://gromgull.net/2009/11/trie.py

Enjoy!

So — Koble lets me create a page for a concept and then associate links with this page, in fact, it even has a bookmarklet for easily creating new things from pages or associating page to existing things. Sounds like anything you know? A bit like tagging with delicious or bibsonomy? Oh yes. So in the off periods of not using Koble I used bibsonomy for all my links, and I definitely use bibsonomy for keeping track of publications that I read skim the abstract and conclusion of. However, when it comes to writing notes about what I read I need the wiki. Sigh. So I bit the bullet, picked up the python showel and dug into the bibsonomy API, and a few hours later I have:

At first I worried about identity — what if the OnlineLearning of Koble is no the same as the OnlineLearning tag on bibsonomy? What if I have SemanticWeb vs. semantic-web? Maybe I should let users (read “myself”) selectively associated different things with different tags?

Then I learned to stop worrying and just get on with my life and now I show any links/publications where the tags matched if they are the same once all non-letters have been removed and the rest converted to lower-case.

Now I’ve fixed my tool — now nothing can stop me from actually doing some work.