Hi,
Thanks for the reply's. I was tempted to accept the Rodoslaw Smogura proposal. There will be about 100 websites to capture data on daily basis. Each website adds per day(average) 2 articles.
Thomas talked about the noSQL possibility. What do you think would be better? I have no experience in noSQL and that could be a weakness.
Best Regards,
André
On Mon, Jan 3, 2011 at 11:58 AM, Thomas Schmidt
<postgres@stephan.homeunix.net> wrote:
Hello,
Am 03.01.11 12:46, schrieb Radosław Smogura: I can propose you something like this:
website(id int, url varchar);
attr_def (id int, name varchar);
attr_val (id int, def_id reference attr_def.id, website_id int references website.id, value varchar);
If all of your attributes in website are single valued then you can remove id from attr_val and use PK from website_id, def_id.
Depending on your needs one or many from following indexes:
attr_val(value) - search for attributes with value;
Probably you will use 2nd or 3rd index.
Example of search on website
select d.name, v.value from attre_def d join attr_val v on (v.def_id = d.id) join website w on (v.website_id = w.id)
where d.name = 'xxxx' and w.url='http://somtehing'
Your schema is nice for specific querying, but might blow up if lots of data is stored in the database (joins, index-building might be time consuming).
On the other hand, google put some effort into their "BigTable" http://en.wikipedia.org/wiki/BigTable for storing tons of data...
Thus - it all depends on the usage :-)
