Thomas Lockhart <lockhart@alumni.caltech.edu> writes:
>> Something else we might consider as a stopgap is to resurrect the
>> "compressed text" datatype that Jan wrote, and then removed in
>> anticipation of having TOAST.
> Naw, the above should be easier all around.
When you finish catching up on your mail, you'll find lztext is already
back in ;-). At this point, whether you change the representation is
pretty much irrelevant for rule size, I think. However, I am still
concerned by the hack I had to put into ruleutils.c to get pg_dump
to produce valid output for cases likecreate view foo as select * from int8_tbl;
See the note and code at about line 1000 of utils/adt/ruleutils.c.
Ideally we want to be able to tell from the parsetree whether the user
wrote any column aliases or not (and if possible, distinguish the ones
he wrote from any that got added by the system). So that may force
a representation change anyway.
regards, tom lane
