Hi guys,
I am working on a problem, that I would like to see if anyone might be able to help me with.
I have a simple rails application that uses a controller to find a record in a database. I grab the path that the client is trying to access, and then I look in a database to see if I have a record that matches this path and return an appropriate response.
The table looks like this (simplified example):
The challenge here is that there's no DB (that I'm aware of) that can meaningfully index regexes, meaning that every query against the table is going to be a full table scan. How bad that is depends on how big your table is:
- if it's just a couple dozen records, it won't be too slow. You may want to just load all the objects and scan them on the Ruby side, as the implementation will be much more straightforward and performance won't be an issue.
- if it's tens of thousands of records, you're probably screwed. DB- side lookups will still be slow, and loading everything into memory will also be slow.
The MySQL manual mentions that you can use a DB column for the pattern in a REGEXP expression, so that might be looking into.
--Matt Jones
able to insert a regular expression, that I will then use to match against the path. That way having a record value like /\/users\/(\d*)/ to match a client request path like /users/1234.
> path | response | ------------------------------ > /\/users\/(\d*)/ | { users } |
I first looked athttp://www.postgresql.org/docs/8.3/static/functions-matching.html (section 9.7.2), but that is the reverse problem
I'm not sure what problem you're trying to solve with an approach like this, but a quick test confirms that a query such as:
select response from paths where '/users/1234' similar to path
would actually work. Of course, it will require setting up a custom sql condition, but that's not really a huge deal. As for the previous reply, a sequential table scan on 10k records isn't a big deal for PostgreSQL on any modern machine, in fact I rarely even think about indices until tables are at least approaching 10k records and usually well beyond.
Jim Crate
Postgres can handle this kind of lookup, though it is more complex than the normal situation. Postgres uses customisable indexes, so you can specify things exactly as you need for specific lookups.
Thanks guys...! I am going to give this a try! I should of course just have tried this out. Will give it a go.
I am not going to be anywhere near 10k rows, but probably rather a few hundred rows. So table scan won't be that big a deal, and this is never going to be a performance critical system, but rather one used in test environments.
Thanks again!
/JP
I'd be interested in getting a bit more information about what sort of custom SQL condition or customized index is required to get this to work.
Thanks... ESB
> I'm not sure what problem you're trying to solve with an approach like > this, but a quick test confirms that a query such as:
> select response > from paths > where '/users/1234' similar to path
> would actually work. Of course, it will require setting up a custom > sql condition, but that's not really a huge deal.
I'd be interested in getting a bit more information about what sort of custom SQL condition or customized index is required to get this to work.
In Rails 2.x, you would do something like:
Path.find(:first, :conditions => ['? similar to path', accessed_path])
I haven't had to do custom SQL conditions in Rails 3.x yet so I can't help with that. I'm also not sure what kind of index you'd use to query like that, the PostgreSQL mailing list would be a better place to ask.
Jim Crate