Greg Stark <gsstark@gmail.com> writes:
> On Jun 3, 2011 4:20 PM, "Tom Lane" <tgl@sss.pgh.pa.us> wrote:
>> I'm inclined to write this off as "so don't do that". There's nothing
>> that pg_dump can do to make this work: it has to use the USING syntax
>> for the join, and that doesn't offer any way to qualify the column name
>> on just one side.
> There's nothing stopping us from adding a nonstandard syntax to cover
> precisely the information needed to resolve this case when dumping.
> For example we could support USING (a.a=b.a) or ON (a.a=b.a as a)
1. "Nonstandard syntax" is widely seen as "vendor lock-in". I don't
think that people would appreciate such a fix, especially for an issue
so obscure that we've never seen it before.
2. I don't believe your proposal covers all cases. For instance, there
are cases where there is no valid qualified name for a column, ie, it's
a merged column from an alias-less JOIN. (The existence of such cases
is another reason why USING sucks, but I digress.)
> We could use it only in this case where there's ambiguity too so it wouldn't
> clutter people's dumps.
No, because the problem case is where ambiguity gets added after the
fact.
regards, tom lane
