I've been doing some research work using the GiST indexes,
but I persistently develop a problem where the system doesn't
make use of the indexes during the execution of a query. If
I use the examples provided here:
http://wit.mcs.anl.gov/~selkovjr/pg_extensions/
For instance, and I place an elog( DEBUG, "functionname" )
in each of the GiST accessor functions, I can witness when
the database is making use of the index. During the construction
of the index, I never have a problem, although during query
execution, it seems that my indices aren't getting used at
all, and the database is simply searching through all of
the entries in the database.
This is a terribly frustrating problem that I encountered
once before, but which mysteriously went away after fiddling
with the problem for a while. This time, the problem isn't
going away, however. When I trace through the postgres
application I can see that it at least examines the opclass
for my specialized data types, and detects that there exists
an index that could be used, but it seems to decide not to
make use of it regardless.
Is there an easy way that I can force the use of an index
during a query?
-David
----------------------[=========]------------------------
David T. McWherter udmcwher@mcs.drexel.edu
vdiff
=====/vee'dif/ v.,n. Visual diff. The operation offinding
differences between two files by {eyeball search}. Theterm
`optical diff' has also been reported, and is sometimes more
specifically used for the act of superimposing two nearly identical
printouts on one another and holding them up to a light to spot
differences. Though this method is poor for detecting omissions in
the `rear' file, it can also be used with printouts of graphics, a
claim few if any diff programs can make. See {diff}.
