Jing Wang <jingwang...@gmail.com> writes:
> [ support_CURRENT_DATABASE_keyword_v4.7.patch ]
TBH, I think we should reject this patch. While it's not huge,
it's not trivial either, and I find the grammar changes rather ugly.
The argument for using the feature to fix pg_dump issues has evaporated,
but I don't see anything in the discussion suggesting that people see
a need for it beyond that.
I particularly object to inventing a CURRENT_DATABASE parameterless
function. That's encroaching on user namespace to no purpose whatever,
as we already have a perfectly good regular function for that.
Also, from a user standpoint, turning CURRENT_DATABASE into a fully
reserved word seems like a bad idea. If nothing else, that breaks
queries that are relying on the existing current_database() function.
The parallel to CURRENT_ROLE is not very good, because there at least
we can point to the SQL spec and say it's reserved according to the
standard. CURRENT_DATABASE has no such excuse.
regards, tom lane
