Robert Haas <robertmh...@gmail.com> writes:
> On Dec 29, 2010, at 11:16 AM, Tom Lane <t...@sss.pgh.pa.us> wrote:
>> I really really dislike the notion of a "verification scan": it's
>> basically work that is going to be useless if it fails.
> I think it has potential in cases like text to xml. In that case it'll
> either work or fail, with no possibility of requiring a do-over. Scanning
> the whole table is a whole lot cheaper than rewriting it.
I don't believe avoiding the write part (but not the read part, nor the
XML syntax verification part) is a sufficiently compelling argument to
justify having that code path. There are not enough distinct datatypes
sharing binary representations to make this a worthwhile thing to worry
over.
Basically, I believe that the only use-case that will have more than
epsilon number of users is "I want to make this varchar(5) into
varchar(10), or possibly text". We can fix that case without adding a
boatload more code that we'll have to maintain.
I do have some interest in the idea of having a type-specific function
that can recognize no-op typmod changes, but I would envision that as
being an expression evaluation optimization: let the planner throw away
the call to the length-checking function when it isn't going to do
anything. It's not by any means only useful in ALTER COLUMN TYPE ---
and in fact probably doesn't even need any bespoke code there, if we put
it into expression_planner() instead.
regards, tom lane
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
