How do I efficiently get the first n rows of a query from

Community Support
1

I want to get for example first 10 rows from partitioned table in specific date (date is a partition column). The solution is

10# select from trade where date = 2014.01.02

but this query has the same usage of memory as

select from trade where date = 2014.01.02

(I used \ts and get same values for both cases)

so I hope, theres more efficient solution. Any ideas, please?

Thanks Pavel

2

This isn’t necessary on 64bit, but on 32bit this should reduce your mem load; using functional select, load a column at a time. Here using trade table, last date. Use parse"select …." to figure out how to form functional select.

q)flip raze {flip 10#?[trade;enlist(=;`date;2014.01.02);0b;(enlist x)!enlist x]}each cols trade

3

select from trade where date=2014.01.02,i within (1;1+10)
or look at 

http://code.kx.com/wiki/DotQ/DotQDotind

Thanks,

Sean

4

address-space/mapping is the problem
hence Charlie’s suggestion

10#..

or

..i<10

does not matter

Cheers,

  Attila

  Attila

5

I can’t use i<10 nor select [10] becase it is partitioned table. It tried .Q.ind before, but I waited for resuresult almost half an hour.Selecting by date is quite complicated with .Q.ind and not efficient (searching for starting index of the particilar date).

6

With 32bit, for ease of querying, you probably don’t want to go above ~500MB per partition.
See http://code.kx.com/wiki/JB:KdbplusForMortals/segments#1.4_Segments

7

Thank you for all responses. So it looks, there is no less memory consuming solution, the only way is to keep partiones “small” enough.

Dne pátek, 23. kv?tna 2014 19:29:22 UTC+2 Charles Skelton napsal(a):

With 32bit, for ease of querying, you probably don’t want to go above ~500MB per partition.
See http://code.kx.com/wiki/JB:KdbplusForMortals/segments#1.4_Segments