Some comments on random TODO entries:
* Allow INET subnet tests using non-constants
This should say "Allow ... to be indexed" as it's otherwise a nonissue.
* ARRAYSo -Allow arrays to be ORDER'ed
Although Joe implemented ordering operators, he didn't get around to
adding MIN()/MAX() support for arrays. Seems like if you can do '<'
then you should be able to do MIN().
* Allow CREATE INDEX zman_index ON test (date_trunc( 'day', zman ) datetime_ops) fails index can't store constant
parameters
This is done as of 7.4.
* Improve handling of index scans for NULL
I think this should read "Allow use of indexes to search for NULLs".
ISTM "improve" implies that there's something there to improve...
* CURSORo Allow BINARY option to SELECT, just like DECLARE
This is either done or rendered irrelevant, depending on your point of
view, by the V3 protocol.
o Add SET SCHEMA
What is this supposed to do (and how's it different from SET SEARCH_PATH)?
* -Support statement-level triggers (Neil)
If we're going to remove this as done, we ought to put in a separate
TODO item to allow such triggers to get at the sets of modified rows.
People seem to think that that's a standard part of the capability.
* Wire Protocol Changeso Add optional textual message to NOTIFYo Allow fastpast to pass values in portable format
Those are both done as far as the protocol is concerned, and therefore
should be moved to some other category. The fastpath issue probably
should be under libpq, since it's only a matter of redesigning the API
that libpq exposes for this feature.
o Special passing of binary values in platform-neutral format (bytea?)
That's done, period.
regards, tom lane