join tables on like with index - postgresql

Question

I have a table of URLs (domains and pages)
URLs
-----
url_id
url
I have a list of domain names, that I want to see if are contained in the URLs table.
so if I have a domain in my list:
http://stackoverflow.com
I want it to match the URLs.url record of:
https://stackoverflow.com/question/230479
https://stackoverflow.com/question/395872364
etc
The URL table is quite large, 10million+ and will grow
The list of domain names I want to test will vary between 1-10k
Currently I am creating a temp table of the list of domains, then joining to the URLs table to find all URLs that match
SELECT * from URLs
JOIN tmp_table_domains on tmp_table_domain.domain like URLs.url || '%'
I have indexed the URLs.url and the tmp_table_domain.domain, with the thinking that indexing will work as the wild card is on the right.
However, EXPLAIN ANALYSE doesn't show any index being used. An old post mentioned that postgres 8.x cannot like join with index, but I could find nothing else to back this up or alternatives or whether it applies to newer versions
If it helps, my postgres is 9.1. If upgrading will fix this, that is fine, only reason haven't upgraded is not been any reason to that I am aware of
Edit_1
this is a first database project have worked on and am learning it all as I go along
I don't mind ripping out all of the above and using whatever works better, whether that is a temp table / array / better query
edit_2
GroupAggregate (cost=1429152.90..1435118.48 rows=340890 width=44) (actual time=157905.450..157905.609 rows=27 loops=1)
-> Sort (cost=1429152.90..1430005.13 rows=340890 width=44) (actual time=157905.425..157905.451 rows=29 loops=1)
Sort Key: task_items.task_item
Sort Method: quicksort Memory: 29kB
-> Nested Loop (cost=14210.95..1387337.41 rows=340890 width=44) (actual time=18216.187..157905.055 rows=29 loops=1)
Join Filter: ((task_items.task_item)::text ~~ ((tmp_domains.domain)::text || '%'::text))
-> Hash Join (cost=14210.95..194126.53 rows=14066 width=44) (actual time=452.262..7953.639 rows=13737 loops=1)
Hash Cond: (task_items.task_id = tasks.task_id)
-> Seq Scan on task_items (cost=0.00..170062.71 rows=2589924 width=48) (actual time=0.019..4480.360 rows=2575206 loops=1)
Filter: (task_item_status_id = 2)
-> Hash (cost=14205.68..14205.68 rows=421 width=4) (actual time=440.409..440.409 rows=171 loops=1)
Buckets: 1024 Batches: 1 Memory Usage: 7kB
-> Seq Scan on tasks (cost=0.00..14205.68 rows=421 width=4) (actual time=101.491..439.821 rows=171 loops=1)
Filter: ((account_detail_id = 695) AND (base_action_type_id <> ALL ('{1,3,4}'::integer[])))
-> Materialize (cost=0.00..109.70 rows=4847 width=32) (actual time=0.002..4.924 rows=4536 loops=13737)
-> Seq Scan on tmp_domains (cost=0.00..85.47 rows=4847 width=32) (actual time=0.010..5.851 rows=4536 loops=1)
Total runtime: 157907.403 ms
The actual query is a bit different to the simplified explanation above.
task_items has just under 7million rows
and the tmp_domains has 4,500
tl;dr
so to summarise. What is the best way to partial match a list of strings to a column

Answer 1

A few months back Peter Eisentraut published the pguri extension which can greatly simplify your work. It is currently only source code so you'd have to build the library code, which is very easy on any Linux box, then place the files in the PG installation directory and finally CREATE EXTENSION in your database. After that you can do simple queries like:
SELECT *
FROM urls
JOIN tmp_table_domains d ON uri_host(d.domain::uri) = uri_host(urls.url::uri);
Note that this will also match between different schemes, so an http:// domain will match the corresponding https:// url. If you do not want that, then also join on uri_scheme() for both domain and url.
Indexes will work on the text data type that the functions of the extension returns. If your database uses UTF-8 encoding you should create your index somewhat like this:
CREATE INDEX url_index ON urls (uri_host(url::uri) text_pattern_ops);
And then also for your domain names table.
You can ALTER TABLE urls ALTER COLUMN url SET DATA TYPE uri so you can forego the casts.