179193 - anyexact should use IN, not OR

Status: RESOLVED FIXED

(Bugzilla :: Query/Bug List, defect)

Description

[Bradley Baetz (:bbaetz)]

mysql optimises |foo IN (a,b,.....)| into a binary search. For OR, it tests each
one individually.

For the set of 'all duplicates in bugzilla', the difference between the two for
5 runs (of duplicates.cgi, see the patch on bug 176599) is 1min 3 sec vs 8
seconds. OK, this is probably several thousand bugs, but still...

I'm going to attach a patch which fixes this.

I know that pg barfs when it gets severl thousand things in an IF; if dkl can
come up with a ballpark numebr, we can split this into multiple INs.

Comment 1

[Bradley Baetz (:bbaetz)]

Attachment: fix

Comment 2

[Bradley Baetz (:bbaetz)]

Comment on attachment 105675 [details] [diff] [review]
fix

BTW, do I need the if, or can I be guarnateed that this routine will always be
called with at least one value in the list?

Comment 3

[Gervase Markham [:gerv]]

Comment on attachment 105675 [details] [diff] [review]
fix

r=gerv, although doesn't $term start off a "", obviating the need for the
"else"? Otherwise, looks sane to me.

Gerv

Comment 4

[Dave Miller [:justdave]]

a= justdave

Comment 5

[Bradley Baetz (:bbaetz)]

gerv: I don't know, hence my question. That code is Fun....

I'll check this in when I wake up 

Comment 6

[Bradley Baetz (:bbaetz)]

Attachment: take 2

OK, you're right - $term is set to undef before calling. I dropped the else,
and also dropped the if, since we skip out earlier earlier if trim($v) eq ''.

Comment 7

[Gervase Markham [:gerv]]

Comment on attachment 105720 [details] [diff] [review]
take 2

r=gerv.

Gerv

Comment 8

[Bradley Baetz (:bbaetz)]

fixed.

dkl, if that pg bug isn't fixed in 7.3, then we can split this up as discussed
in comment 0