: use of cou time for each marchsearch usage. I can't install marcsearch in
: large archive (10000 mails) because it need 30" cpu on a P200 pro linux
: machine. Any solution to make Marc-Search work faster ?
Split the archive into smaller by-month archives, and let marc-search
give your users the ability to select which months they want to
search. I use archives of the following form:
archivename/1997-11/
archivename/1997-10/
and so forth. Searches initiated from html files within 1997-11 will
be limited to those files; searches initiated from filess in
archivename (the index of months) will present the user with a
multi-select list so that (s)he can choose to search in November,
October and August, but not September. The config files for
marc-search also let you define "human readable" names for these
Year 2000-proof directory names, for example, Juillet 1997 or November,
1997.
The only other solution would be to build real search indexes (like
Glimpse does) beforehand so that marc-search doesn't have to open all
of those files. If there were sufficient demand for this, I'd make
plans to do it, but I suspect most people who need this kind of
performance would prefer a C program like Glimpse because of its speed
and because it's useful in other contexts as well. As a line-matching
grep-derivative, glimpse is not especially intelligent when it comes to
full body searches, but it's free. ;-)
Eric
--
Eric D. Friedman
friedman(_at_)uci(_dot_)edu