topic Re: Proc SQL Left join and order by in SAS Programming

Proc SQL Left join and order by

LineMoon — Wed, 16 Sep 2015 21:28:58 GMT

Hello,

I want to write this

proc sql;

create table test as

select t1.*, t2.v

from t1 left join t2

on t1.w=t2.w;

quit;

Is there any mather , if I do not use " order by ...", an other mean, sas can give any false result ?

if we do not write

proc sql;

create table test as

select t1.*, t2.v

from t1 left join t2

on t1.w=t2.w

order by m1,

quit;

If yes, please can you give some exemples ?

Re: Proc SQL Left join and order by

hbi — Wed, 16 Sep 2015 22:02:31 GMT

The presence or lack of an "order by" statement in SQL will not affect the number of observations in the output dataset.

The above statement would be true, in a join even if you "ORDER BY" a variable that is not in the output dataset.

Even if you added something weird, like " ORDER BY monotonic(), ranuni(123) ", it should not affect the number of observations in the output dataset.

Re: Proc SQL Left join and order by

ballardw — Wed, 16 Sep 2015 22:04:14 GMT

I am not sure that I understand what you mean by "false result";

Currently the implementation of proc sql without an order by variable will tend to order data by the values of the varaibles as they appear on the Select clause. The order of set T1 in your example may not be the order it had going in.

Re: Proc SQL Left join and order by

art297 — Wed, 16 Sep 2015 22:35:23 GMT

Your code works as is! The only problem might be if v and m1 are in both datasets. However, the result of either run wouldn't be false in such cases, just not what you might have expected.

Re: Proc SQL Left join and order by

PGStats — Thu, 17 Sep 2015 01:38:45 GMT

It might only be a transcription typo but the comma after m1 should be a semicolon. Otherwise, proc SQL will be looking for a variable named quit.

Re: Proc SQL Left join and order by

LineMoon — Thu, 17 Sep 2015 19:54:56 GMT

you say "Currently the implementation of proc sql without an order by variable will tend to order data by the values of the varaibles as they appear on the Select clause"

You mean that : sas will order the final data by the variables in select clause ?

Thank you