Question

I have two tables that look like this:

node:
NID    Name
1      Chicago
2      New York
3      LA

url_alias:
src    dst
user/1    user/Bob
user/2    user/Mike
node/1    node/chicago
node/2    node/new-york
node/3    node/LA

etc

Basically, the url_alias table has alias urls that my webserver redirects to. Both nodes and users have ‘ids’ as seen in the url. I am trying to join my node table to the url_alias table and match on the stuff after the ‘/’ (the node id). So far I have this code, but it only uses substring and not like so it also matches any users that end in the correct number:

select `node`.`nid` AS `nid`,`node`.`name` AS `name`,`url_alias`.`src` AS `srcurl`,`url_alias`.`dst` AS `dsturl`
from ((`node` join `url_alias` on ((`node`.`nid` = substring(`url_alias`.`src`,6)))))
order by `node`.`nid` desc

(actually the query is a bit longer as it also joins another table, which is working, but I have edited it for simplicity)

This query returns results like this:

nid    name      srcurl    dsturl

1      chicago   node/1    node/chicago
1      chicago   user/1    user/Bob
2      new-york   node/1    node/new-york
2      new-york   user/1    user/Mike

Is there a way to use “LIKE ‘node/%'” within my query, or something similar to only grab ‘nodes’ before substring runs its matching?

Answer 1

the on portion of a join is essentially just a where clause. Anything you can specify in a where will work in an ON, so…

JOIN x ON somefield LIKE 'node%' AND node.nid = SUBSTRING(src, 6)

is perfectly acceptable. Note that since you’re using the substring operation, use of indexes will not be possible – indexes do not work on derived data. You would be far better off normalizing your data and splitting that ‘id’ portion into its own field.