Re: pg_dump of only the structure from a client such as ruby

Perry Smith <pedz(at)easesoftware(dot)com> writes:
> I find it odd that you are resistant to the idea. To me, the
> advantages are clear if it can be done without a tremendous amount of
> work.

Well, it can't. pg_dump is an enormously complicated and frequently
changed bit of code, and so you really really don't want to be copying
it into other clients.

For a long time there's been occasional talk of refactoring pg_dump into
a library and wrapper program such that the library could be used by
other clients. But that would be a large amount of work in itself,
especially if you hoped to design a library API that was rich enough
that it'd be meaningfully more flexible than pg_dump itself is. Aside
from the sheer work involved, there's the danger of introducing bugs
into what is certainly a critical part of our infrastructure. So it's
never gotten further than "wouldn't it be nice" discussions.

regards, tom lane

PS: *please* see if you can get them to take out the usage of pg_dump's
-i switch. Having code invoke that blindly borders on criminal
negligence.