Re: fulltext searching via a custom index type

Eric Ridge <ebr@tcdi.com> writes:
> I assume the doc chapter on Page Files and the various storage-related 
> README files are good places for more information.  Any other tips or 
> pointers?

Not right offhand, but feel free to ask questions when you get stuck.
Also don't forget that there's a wealth of info in source code comments;
you should always try looking for existing code that does something
similar to what you want to do.

BTW, one problem with using the hash AM as a model is that it's not yet
WAL-aware.  The btree AM is the only one that's really WAL-ified yet,
so you need to look at that if you want to think now about how to make
your code safe for WAL.  I think you'd probably be okay to postpone this
consideration for later, but keep in mind that all your operations that
change the contents of index files should be divisible into bite-size
operations that can be logged as WAL entries.  Each "bite-size
operation" has to leave the index in a legal state, too, so there's a
limit as to how small you can subdivide.
        regards, tom lane