Re: pgsql-server: Some preliminary documentation for composite-type stuff.

Alvaro Herrera <alvherre(at)dcc(dot)uchile(dot)cl> writes:
> Nice work. This is a cool unexpected step forward in functionality.

As recently as Thursday I didn't think this would get done for 7.5,
but I wanted to nibble off a few rough edges, and after a while there
weren't any left.

Or almost ... I just noticed that there's no convenient way to assign to
a subfield in UPDATE. You can hack around it with something like
UPDATE mytab SET myfield = ROW(myfield.a, newval, myfield.c)
but my goodness that's painful. AFAICS the SQL99 spec gives license
to write
UPDATE mytab SET myfield.b = newval
and I think I will look into making that happen tomorrow. We already
have a solution in place for assigning to an array element, and this
doesn't seem much different.

regards, tom lane