* Micah Stevens <[EMAIL PROTECTED]>: > What kind of speed do you need?
I was aiming at something that would run in 2 hours or so. Currently, it takes around 3 - 4. > I had to go to temp tables for a logging application, but not until my > table got upwards of a million records or so.. For the numbers your > quoting, it should be pretty quick unless your engine needs > optimization.. Well, actually... The index currently has around a million records, with approximately 350 words per resource id, several thousand resource ids per app_id.... you get the idea. I've optimized my indices, and I've got appropriate indices for the selection criteria I'm using. As I've noted, I also realize that this is a one shot deal -- after the initial link table is created, I won't be needing to recreate it every time I have data that changes -- I'll only be operating on deltas from here out, and those can be done quite quickly. Anyways, it's a moot question now -- I've got the data in, and I'm pretty happy with the results I've seen. > On Friday 11 February 2005 11:56 am, Matthew Weier O'Phinney wrote: >> * Micah Stevens <[EMAIL PROTECTED]>: >> > Sounds like a self join should work wonders. >> > >> > I didn't test this, but the idea should work: >> > >> > >> > select t1.app_id as a1, t2.app_id as a2, t2.word, t2.score >> > from tablename as t1 >> > left join tablename as t2 >> > on t1.resource_id = t2.resource_id and t1.app_id != t2.app_id >> > group by word >> > order by word >> > >> > voila, no temp tables. Am I missing something? > > >> Speed. :-) > > >> I had tried this as well. The temporary tables really do offer much >> better performance. The issues I was having were (1) bad resultsets >> (I've now got that fixed) and (2) speed. I still don't have (2) >> completely fixed, and it may be something I can't fix. > > >> > On Thursday 10 February 2005 07:56 pm, Matthew Weier O'Phinney wrote: >> >> I have a table which contains the following: >> >> id (primary key, auto incrementing) >> >> app_id (integer, foreign key) >> >> resource_id (integer, foreign key) >> >> word >> >> score >> >> >> >> (This is a search index.) I want to find all resource_ids from one >> >> app_id that match resource_ids in another app_id by word. >> >> >> >> I have created a temporary table 'tmp1' that contains all resource_ids >> >> from the second app_id (the one whose resources I wish to retrieve). I >> >> am then looping through all resource_ids in the main table with the >> >> first app_id, and doing the following: >> >> >> >> * Creating a temporary table tmp2 with a single column 'word' >> >> populated by the words associated with resource_id in the main >> >> table >> >> * Selecting all distinct resource_ids from tmp1 INNER JOIN'd on tmp2 >> >> on the word field >> >> >> >> The issues I'm running into are that (1) each resource_id cycle takes a >> >> good amount of time, and (2) I seem to be getting either too many >> >> resource_ids or not enough. >> >> >> >> (1) may be something I just have to deal with. As it is, I'm planning on >> >> running the full indexing once, and then doing incremental updates, so >> >> it may not be that big of an issue (unless it takes too much time to >> >> create the initial index). As for (2), unfortunately, I'm not sur ehow >> >> to really trouble shoot the issue. I know, for instance, that in once >> >> case, I have a list of 343 words that generates a list of ~12,000 >> >> resource_ids (of a possible 18,000) -- but I don't quite know how to >> >> spot check 300 values to be certain that this is reasonable. >> >> >> >> In a previous incarnation of the script, I was looping through each word >> >> of each resource_id and then selecting out of tmp1 based on the single >> >> word value. The results were very different (very few matches), and, >> >> again, the script ran long. >> >> >> >> Any ideas? -- Matthew Weier O'Phinney | WEBSITES: Webmaster and IT Specialist | http://www.garden.org National Gardening Association | http://www.kidsgardening.com 802-863-5251 x156 | http://nationalgardenmonth.org mailto:[EMAIL PROTECTED] | http://vermontbotanical.org -- PHP Database Mailing List (http://www.php.net/) To unsubscribe, visit: http://www.php.net/unsub.php