On Thu, Jul 22, 2004 at 02:24:17PM -0700, Bill Moseley wrote:
> No. The way the wild card index works is just like the index in the
> back of your text books -- arranged alphabetically. Imagine someone
> asking you to look up in that index all words that end in "ing" --
> the words are not organized that way.
Well, by "brute force" I meant, I'd be happy to code up a hack just for
myself that would examine every word in the index and check their endings,
for example. Roughly were in the code would I look for doing that kind of
thing?
Or, worst case, is there a way I can get swish-e to dump the list of all
words associated with a certain meta name?
> You are more patient than I. Is 6 days acceptable? That's not your
> average web site's worth of pages to search.
Well, it's mostly in the preprocessing. :) It's a one-time indexing for
this data, but it's the searching we'll be doing a lot of.
> without thinking about meta names (or structure or word position).
> Then once it finds that word it walks though a list of data saying
> what meta ID it's associated with and picks only the ones that match
> the meta you are asking to search.
Hm, I've worked around this by doing a*, b*, ... z* all OR'd together. :)
--
Kees Cook
Open Source Development Lab
kees@osdl.org
Received on Thu Jul 22 14:40:43 2004