Re:Following a Select Statement Through Postgres Internals
by pat_shaughnessy on Monday October 13, 2014 @07:43PM
(#8451245)
Oh absolutely. I thought about saying that, and just ran out of energy before posting the article this morning.
The reason I used this particular SQL statement as an example is that ActiveRecord (a Ruby ORM) generates it from a fairly simple, common Ruby expression. I suppose ActiveRecord could be improved to drop the sort when you know there's one one possible match.
Re:Following a Select Statement Through Postgres Internals
by talaketu on Monday October 13, 2014 @09:07PM
(#8451429)
Unless "name" is unique, there could be multiple matches. Asking for the "first" demands an ordering.
Perhaps (a) the user doesn't care what ordering is used, and/or (b) (such as in your example in the preceding post) hasn't specified an order (and apparently PK ordering is defaulted to PK - a model option?). It would be error-prone for ORM to take (b) as implying (a).
Re:Following a Select Statement Through Postgres Internals
by pat_shaughnessy on Monday October 13, 2014 @09:15PM
(#8451455)
Yes, good point. Rails has a lot of conventions like this - that "first" implies ordering by primary key (by default). I just imagine that most Rails developers don't think about sorting and its implications when they ask for the first record.
Re:Following a Select Statement Through Postgres Internals
by empthought on Tuesday October 14, 2014 @06:07AM
(#8452688)
Ordering by ID can be used in keyset pagination as well as limit/offset: http://use-the-index-luke.com/no-offset
Just drop sort clause. Returning the user with the lowest id seems like an unusual use case.
Imagine searching for a name in a phone book and only finding the oldest person with that name, or the person with the lowest phone number.
More common cases are checking whether a user exists, which doesn't require a sort, or finding all the users with a particular name, which I guess could be paged using limit and offset.