Peter Eisentraut <peter_e@gmx.net> writes:
> What's the deal with the line type?
> It's installed in the catalogs and listed in the documentation,
> varyingly as not implemented or not fully implemented, but all the
> support functions throw an error. Is there any known list of things
> that would need to be done to make it fully implemented? Or should we
> just get rid of it?
Tom Lockhart seems to have lobotomized line_in and line_out in commit
402b47cffafcbbc66af8ee6b6340b4db09f00a7b. A look in the archives
suggests it was because of my complaint the day before:
http://archives.postgresql.org/pgsql-hackers/1998-08/msg00346.php
I'm not sure there's anything wrong with line_in, but line_out is still
obviously broken. However, there are a couple dozen other functions
taking/returning "line", and in general those don't look significantly
more bogus than any other part of geo_ops.c.
I do notice this comment above line_in:
** Lines are not intended to be used as ADTs per se,** but their ops are useful tools for
otherADT ops. Thus,** there are few relops.
but that argument seems specious from here. Even if it's mostly a
support type, being able to get at it would be useful for
testing/debugging the other operations, I think.
I'd vote for fixing the I/O functions and adding some docs and
regression tests ... but on the other hand, I'm not volunteering
to do the work.
regards, tom lane
