Re: Performance comparison: direct select or regex filter ?

>> To your point, though (the reason I started writing this email):  
>> there's not
>> too much motivation for implementors to try to detect and optimize  
>> this kind
>> of edge case, because nobody should be writing a query in the  
>> second form :)
>
> There are reasons for wanting to see all strings that start a
> particular way...

I was actually referring to the "regex identical to a string match"  
problem, which I think the original mail intended: it's not really  
worth adding an optimization check for regex(... "^foo$"), because the  
user really shouldn't write that! I agree with you abuot prefix search.

>> One thing I've experimented with is detecting some kinds of regexes  
>> and
>> compiling them into free-text index operations. That's likely a  
>> middle
>> ground.
>
> I've considered the same, but so far I've been of the opinion that
> there are a lot of other areas I need to improve performance in first.
> :-)

Heh, ain't that always the way? :)

Received on Thursday, 29 April 2010 19:20:48 UTC