The Hermit Hacker <scrappy@hub.org> writes:
> On 27 Nov 1998, Tom Ivar Helbekkmo wrote:
>> See http://glimpse.cs.arizona.edu/ for a powerful inverted indexing
>> engine and various related software.
> Just curious, but other then specialized applications like
> Glimpse, does anyone actually support/do this?
I dearly love Glimpse. (Sample things I use it for: rooting through
nearly 10 years worth of archived email; finding all references to a
particular name in the Postgres sources, almost instantly; ditto for the
even larger Ptolemy sources; looking for files that I can't remember
where I put ... it's great. And aren't the Postgres mailing list
archive indexes Glimpse-driven?)
I don't currently have any databases that could benefit from full-text
indexes. But I can think of applications where it'd be important,
particularly after we get rid of the limit on tuple sizes so that it
becomes reasonable to put fair-size chunks of text into database
entries. For example: would it be useful to put my email archive into
a Postgres database, one message per tuple? Maybe ... but if I can't
glimpse it afterwards, forgetaboutit.
You could probably glue something like this together from existing
spare parts, say by running a nightly cron job that dumps out the
text fields of your database for indexing by Glimpse. But it wouldn't
be integrated into SQL --- you'd have to query the index separately
outside of SQL, then use the results to drive a query to fetch the
selected records.
A seamless integration would make Glimpse indexes be a new type of
index associated with a new match operator, something likecreate index index1 on table using glimpse
(text_field);select* from table where glimpse(text_field, 'pattern');
I have no idea how hard that would be...
regards, tom lane