I’ve heard that using LIKE is very slow, but I see it being used a lot
in examples, blogs etc. Is it really that bad? Since Rails doesn’t
directly support Fulltext search, this is the easiest way to get
searching done, right? Or are there any other easier ways? I’m using
acts_as_indexed right now, but it still doesn’t do what LIKE does.
On Fri, Mar 13, 2009 at 12:42 PM, Mike C [email protected] wrote:
I’ve heard that using LIKE is very slow, but I see it being used a lot
in examples, blogs etc. Is it really that bad? Since Rails doesn’t
directly support Fulltext search, this is the easiest way to get
searching done, right? Or are there any other easier ways? I’m using
acts_as_indexed right now, but it still doesn’t do what LIKE does.
The problem with the sql “LIKE” is that it can render the use of an
index impossible. Imagine a search like:
@users = User.find(:all, :conditions => “first_name like ‘%ILL%’”)
The database will have to find BILL, GILL, JILL, JILLIAN, etc. If you
have a
few hundred or a few thousand users that may not be such an issue.
Millions
of users makes this a different issue altogether. A full table scan is
needed, and what will you do with all those rows?
The point is this, when you are tempted to use LIKE be sure the data set
searched and returned are small and that you try to other ways to
constrain
the search.
Cheers–
Charles
Thanks, that makes things clearer. But how big is big? I don’t plan on
my app being hugely popular, but would thousands of entries be ok? Is
there an area where LIKE starts to be really slow?
On Fri, Mar 13, 2009 at 1:11 PM, Mike C [email protected] wrote:
Thanks, that makes things clearer. But how big is big? I don’t plan on
my app being hugely popular, but would thousands of entries be ok? Is
there an area where LIKE starts to be really slow?
Depending on the speed of your hardware and the performance of your
database
I would look for other ways to search if you have more than a 1000 rows
or
two. I have a postgres database here with an accounts table with 1810
rows.
A LIKE search for names like '%ce% returned no rows in 25 ms. I have a
jobs
table with 5.3 million rows and the same search took 28 seconds to
return no
rows.
Of course, if your database caches searches and you do the same
search repetitively they will get faster and faster. I guess I would
not
worry about a few thousand rows. Get you app working, passing all tests,
then worry about optimizations.
Cheers–
Charles
Thanks, that makes things clearer. But how big is big? I don’t plan on
my app being hugely popular, but would thousands of entries be ok? Is
there an area where LIKE starts to be really slow?
If it’s a blog, I wouldn’t think twice about using LIKE. That is, I
can’t see a blog being big enough for it to become a problem.
However, as other folks have said it definitely can be the source of
problems down the road.
At that point you should look at one of the full text search engines
like ferret, sphinx, solr, etc.
Also, just so you’re aware LIKE in mysql is case-INSENSITIVE. LIKE in
PostgreSQL is case-SENSITIVE.
So when you’re searching for names and are happy doing a LIKE ‘%ill%’
to find those names below, keep in mind that won’t work in
PostgreSQL. PostgreSQL has an ILIKE for that, but then that’s not
supported by MySQL. PostgreSQL also has some regex pattern matching
functions that are faster, but again I believe are postgres only.
Anyway, just something to keep in mind if you think your app might be
run with different backends.
-philip
Likes are always slow if you must do a suffix match ( “%something” ),
as no index can handle that. But if you’re doing a prefix match (
“something%” ) it won’t be that bad.
And there’s no “how big is big”. You can have a slow app with a
thousand rows and a hundred queries per second, and you can have a
fast one with a million rows and a query per second. It all depends on
what you’re doing.
Get a book on database optimization, all big databases have one, then
you’ll undertand what’s fast and what’s slow about your database.
Maurício Linhares
http://alinhavado.wordpress.com/ (pt-br) | http://blog.codevader.com/
(en)
Thanks for the replies. Another thing I see are some search plugins
that use LIKE in them predominantly and that’s also why I was
wondering if it was ok. searchlogic was one I was looking at and so
was search_do.