topic Taking too long to show the results in New SAS User

Taking too long to show the results

Sandeep77 — Tue, 12 Jul 2022 14:13:38 GMT

Hi Experts,

I am running the below code where I am added tran_code details to a imported file (June_exception). The codes looks alright and simple but it is running since the last 3 hours. Is anything I am missing in the code which is taking long to show the results?

proc sql;
create table initiative_release as
select a.*,
b.tran_code
from June_Exceptions as a
Inner join
p2scflow.debt_trans as b on a.debt_code=b.debt_code;
quit;

Please check and reply. Thank you

Re: Taking too long to show the results

Reeza — Tue, 12 Jul 2022 16:54:01 GMT

How big are the files? Are they both SAS data sets or is one on a server?

Re: Taking too long to show the results

ballardw — Tue, 12 Jul 2022 22:40:43 GMT

I have a strong suspicion that your are getting way more output records than you expect. Consider this small example doing an inner join on two data sets with 5 records each.

data a;
   input code value1;
datalines;
1  1
1  2
1  3
2  11
2  22
;

data b;
   input code value2;
datalines;
1  15
1  25
1  35
2  111
2  222
;

proc sql;
   create table example as
   select a.*, b.value2
   from a 
        inner join
        b
        on a.code=b.code
   ;
quit;

There are not 5 output records there are 13 because of multiple values of the code variable in each set. 3 values of 1 in each means there are 9 output records for code=1. So if you have any duplicates of debt_code in any set you get multiple output records for each match. With n duplicates of the the code value in one set and m duplicates in the second set you get n * m records for each code.

You may want to run proc freq on the debt_code variable in your two sets and consider just how many matches that will be.

Re: Taking too long to show the results

Kurt_Bremser — Wed, 13 Jul 2022 05:29:52 GMT

Get to know your data (Maxim 3).

How many observations are in both datasets
Are there multiples of the keys in both datasets (leading to a cartesian join)
What are the variable attributes (should COMPRESS be used)

Since you use single level dataset names, all I/O concentrates on WORK, which can cause lots of latencies on spinning metal disks. So the physical structure of your available storage may allow spreading the load.