https://stackoverflow.com/questions/30641876/monetdb-sql-method-to-locate-or... i'm thinking i can do this with some costly self-join, but i'd appreciate any other eyes on the problem
Maybe I’m thinking in a too simple way, but if you store the intermediate results in a table, you can use LIMIT 1: CREATE TABLE t1 (svcmon int); -- assume svcmon is an INT insert into t1 SELECT svcmon FROM person_table AS z WHERE a.yr = z.yr AND a.person_id = z.person_id ORDER BY abs( z.svcmon - a.svcmon ) LIMIT 1; Using TEMP TABLE might give you a bit speed up, but then you need to put them in one transaction. Jennie
hi Jennie, thanks for spending time on this! won't LIMIT only give one record? i need one record per event.. so i think the "insert into" command would need to be repeatedly called until every single event got stored (which would take forever)? sorry if i'm missing something On Wed, Jul 15, 2015 at 6:20 PM, Ying Zhang <Y.Zhang@cwi.nl> wrote:
Opes, I see. I’m sorry. Then LIMIT 1 is not what you need. MonetDB doesn’t have anything directly assemble TOP 1. Your use case sounds a bit similar to another use case on the mailing list, about computing the OHLC chart. It’s a bit too late to think of the exact query, but how about using PARTION BY, then ROWNUMBER, en select the row with number 0 (or is 1 the starting value?)?
Anthony, could you possibly share the full schemas of your "event_table" and "person_table", including primary keys and foreign keys? Thanks! Stefan ----- On Jul 16, 2015, at 1:57 PM, Anthony Damico ajdamico@gmail.com wrote:
-- | Stefan.Manegold@CWI.nl | DB Architectures (DA) | | www.CWI.nl/~manegold/ | Science Park 123 (L321) | | +31 (0)20 592-4212 | 1098 XG Amsterdam (NL) |
hi Stefan, i just made up those column names as a simple fake example. i'm not sure why schema would help here? it's mostly just a re-statement of the original question i linked to in my post..? i guess i could make you some fake data to play with, but i'm confused why this query requires table structure? sorry if i'm overlooking something silly On Fri, Jul 17, 2015 at 9:22 AM, Stefan Manegold <Stefan.Manegold@cwi.nl> wrote:
Hi, making up my suggestions given my understanding of the desired semantics: for the orignal question: create table a (id int, sales int, date timestamp); create table b (id int, goal int, date timestamp); select a.*, b.* from a,b, (select a.date as a_date, max(b.date) as b_date from a,b where b.date < a.date group by a.date) as ab where a.date = ab.a_date and b.date = ab.b_date; for the question below: create table event_table (yr int, person_id int, svcmon int, payload string); create table person_table (yr int, person_id int, svcmon int, payload string); select * from (select e.yr, e.person_id, e.svcmon as e_svcmon, e.payload as e_payload, p.svcmon as p_svcmon, p.payload as p_payload, row_number() over (partition by e.yr,e.person_id order by abs(e.svcmon - p.svcmon) asc) as pos from event_table e , person_table p where e.yr = p.yr and e.person_id = p.person_id) as ep where pos = 1; knowing actual schemas would help to understand whether "each event" is identified by yr,person_id (as I assume above) or by, say, (yr,person_id,svcmon) (in which case e.svcmon should be added to the partition-by clause). knowing actual schemas might also help to pull the projection out of the inner query, as thus shrinking the intermediate result size(s) ... Best, Stefan ----- On Jul 17, 2015, at 3:27 PM, Anthony Damico ajdamico@gmail.com wrote:
-- | Stefan.Manegold@CWI.nl | DB Architectures (DA) | | www.CWI.nl/~manegold/ | Science Park 123 (L321) | | +31 (0)20 592-4212 | 1098 XG Amsterdam (NL) |
that's amazing, thank you!! i've posted here as well- https://stackoverflow.com/questions/30641876/monetdb-sql-method-to-locate-or... On Fri, Jul 17, 2015 at 10:15 AM, Stefan Manegold <Stefan.Manegold@cwi.nl> wrote:
This is probably more of a question than an answer, but - why wouldn't this work?: CREATE TABLE nearest_matches AS ( SELECT a.* , ( SELECT min(abs(z.svcmon - a.svcmon)) FROM person_table AS z WHERE a.yr = z.yr AND a.person_id = z.person_id ) AS nearest_month FROM event_table AS a ) WITH DATA (... that is, aggregation instead of TOP 1.) Eyal On 04/06/2015 14:39, Anthony Damico wrote:
hi Eyal, that does tell you the minimum distance, but not the month which is the minimum distance, i believe? so if your month is july, you might know the nearest month is august or june, but not which one. On Fri, Jul 17, 2015 at 10:39 AM, Eyal Rozenberg <eyalroz@technion.ac.il> wrote:
because one wants the actual z.svcmon value in the result, not it's abs.distance to a.svcmon ;-) ----- On Jul 17, 2015, at 4:39 PM, Eyal Rozenberg eyalroz@technion.ac.il wrote:
-- | Stefan.Manegold@CWI.nl | DB Architectures (DA) | | www.CWI.nl/~manegold/ | Science Park 123 (L321) | | +31 (0)20 592-4212 | 1098 XG Amsterdam (NL) |
participants (4)
-
Anthony Damico -
Eyal Rozenberg -
Stefan Manegold -
Ying Zhang