Wow, makes your head spin!

I am about to start a project, and although my mySql is OK, I can't get my head around what required for this:

I have a table of web addresses.

id,url
1,http://www.url1.com
2,http://www.url2.com
3,http://www.url3.com
4,http://www.url4.com

I have a table of users.

id,name
1,fred bloggs
2,john bloggs
3,amy bloggs

I have a table of categories.

id,name
1,science
2,tech
3,adult
4,stackoverflow

I have a table of categories the user likes as numerical ref relating to the category unique ref. For example:

user,category
1,4
1,6
1,7
1,10
2,3
2,4
3,5
.
.
.

I have a table of scores relating to each website address. When a user visits one of these sites and says they like it, it's stored like so:

url_ref,category
4,2
4,3
4,6
4,2
4,3
5,2
5,3
.
.
.

So based on the above data, URL 4 would score (in it's own right) as follows: 2=2 3=2 6=1

What I was hoping to do was pick out a random URL from over 2,000,000 records based on the current users interests.

So if the logged in user likes categories 1,2,3 then I would like to ORDER BY a score generated based on their interest.

If the logged in user likes categories 2 3 and 6 then the total score would be 5. However, if the current logged in user only like categories 2 and 6, the URL score would be 3. So the order by would be in context of the logged in users interests.

Think of stumbleupon.

I was thinking of using a set of VIEWS to help with sub queries.

I'm guessing that all 2,000,000 records will need to be looked at and based on the id of the url it will look to see what scores it has based on each selected category of the current user.

So we need to know the user ID and this gets passed into the query as a constant from the start.

Ain't got a clue!

Chris Denman