> ... Therefore, we have to tell people
> to use some other API anyway. The existing tsearch2 API at least has
> the virtue of having been proven in the field over several years.
I can only speak as a moderately sophisticated end user, but ... I
think the tsearch2 API has been "proven" to alienate a lot of
potential users, myself included. If the simple things were simple,
there might be a large user base that would rebel against an API
extension, but I don't think this is the case. And I think the need
for a simpler, refactored interface to tsearch is desperate.
Granted, one can learn tsearch2 as is, but it is somewhat painful.
It isn't the sort of thing one figures out for fun and potential
future use, but probably only if one is forced to. If we (well,
"you", really) could make tsearch2 less like C++ (or OCAML or FORTH )
and more like Python, we would get a whole lot of new users of tsearch
in the process, and probably a whole of good will toward PostgreSQL.
-W