Re: [HACKERS] Number of index fields configurable

Bruce Momjian <pgman@candle.pha.pa.us> writes:
> I will tweek the code to properly check for trailing numbers.  Right now
> multiple spaces cause problems, and trailing numbers are ignored.  With
> oidn, we can get away with trailing zeros because an oid of 0 is
> invalid, but with int2n, a zero is valid, so I think we can't just ignore
> extra trailing zeros.  We can pad with zeros, however.  Comments?

For the primary use of these things, which is attribute numbers in
pg_index, padding or dropping zeroes is correct behavior --- unused
positions in the vector will have zero values, same as for the oid
vector.  I think it's OK to define the type's behavior suitably for
the system's use, because it's not intended as a general-purpose user
type; users oughta be using int2[].  (Really, the only reason we have
these types at all is that we depend on having compile-time-constant
field sizes in the system catalogs that are accessed via
include/catalog/'s struct declarations...)
        regards, tom lane