Re: I/O support for composite types

Greg Stark <gsstark(at)mit(dot)edu> writes:
> Following this path, perhaps the array i/o syntax should be changed to
> use []s

I would think about that if there weren't compatibility issues to worry
about, but in practice the pain from such an incompatible change would
vastly outweigh the benefit.

> and the keyword ARRAY should likewise be optional in the array constructor.

Not sure this is syntactically feasible, or a good idea even if it is
possible to get bison to take it --- it might foreclose more useful
syntactic ideas later on. (I wouldn't think that omitting ROW is a
good idea either, but the spec says we have to.)

regards, tom lane