On 2015-04-19 15:33, Tom Lane wrote:
>
>> It's possible that in the future someone will implement an optimization
>> like this, but I'm not aware of anyone working on that and I wouldn't
>> hold my breath.
>
> Yeah. In principle you could make the planner do this. As Adam Williams
> notes nearby, there's a problem with lack of exact consistency between
> extract() semantics and straight timestamp comparisons; but you could
> handle that by extracting indexable expressions that are considered lossy,
What about functions that are simpler such as upper()/lower()?
On 9.3, this:
`select email from users where lower(first_name) = 'yves';
is not using the index on first_name (Seq Scan on first_name). This should be
easy to implement?
--
http://yves.zioup.com
gpg: 4096R/32B0F416