Beefy Boxes and Bandwidth Generously Provided by pair Networks Frank
Perl-Sensitive Sunglasses

Re^3: Exact words in super search (MySQL)

by tye (Cardinal)
on Nov 05, 2005 at 01:03 UTC ( #505922=note: print w/ replies, xml ) Need Help??

in reply to Re^2: Exact words in super search
in thread Exact words in super search

Actually, the code to handle \b was written long ago but it isn't available for public use because it requires MySQL regexes which are huge CPU hogs. It used to be that Saints could use MySQL regexes for searching on node titles, but that feature was doomed to eventually go away (which finally happened as a prelude to introducing the new levels) from the beginning, because it still required too much DB server CPU for it to be used except rarely (it offered a superb denial-of-service attack vector).

sauoq shows the best that is available in Super Search and, unfortunately, MySQL makes it unreasonable to provide anything much more useful. The only way MySQL provides for matching whole words is regexes (too much CPU) or "full-text searches" (which we tried for a while but they had their own problems, the worst but not only one being that they couldn't be prevented from taking the DB server to its knees if your search matched too many nodes). It is too bad that MySQL's LIKE operator doesn't allow something as simple as '%[^a-z]map[^a-z]%', which is a fairly common feature IME.

There are some fairly minor improvements that would be possible such as the ability to not ignore letter case, the ability to anchor searches, the ability to "OR" terms in a single search, or the ability to use concat(' ',title,' ') like '% map %' (which is used if you search for a one-letter word in the simple title search). But none of those are a huge "win" over what we already have.

The best answer (other than somehow getting MySQL to upgrade to one of the newer regex engines that aren't CPU hogs) is probably to allow post-filtering of search results using limited Perl regexes. The trick is that good keywords need to be provided in the SQL for the search to be effective. So I can't just let the user search for /\bmap\b/, I have to have them search for LIKE '%map%' in the SQL and post-filter on /\bmap\b/. Perhaps that would be done automatically by having a field that searches for "full words"...

- tye        

Comment on Re^3: Exact words in super search (MySQL)
Select or Download Code

Log In?

What's my password?
Create A New User
Node Status?
node history
Node Type: note [id://505922]
and the web crawler heard nothing...

How do I use this? | Other CB clients
Other Users?
Others rifling through the Monastery: (5)
As of 2014-04-20 12:05 GMT
Find Nodes?
    Voting Booth?

    April first is:

    Results (485 votes), past polls