topic Re: proc sql left join vs proc sql inner join in SAS Procedures

proc sql left join vs proc sql inner join

dataMart87 — Mon, 02 May 2016 21:08:16 GMT

If a LEFT JOIN and an INNER JOIN return the exact same rows, which is more efficient to use?

For example,

data one;
	mrn='1111';cpt='aaa';output;
	mrn='2222';cpt='bbb';output;
	mrn='3333';cpt='ccc';output;
	mrn='4444';cpt='ddd';output;
run;

data two;
	mrn='1111';age=22;output;
	mrn='2222';age=44;output;
	mrn='3333';age=66;output;
	mrn='4444';age=11;output;
run;

proc sql noprint;
	create table blah1 as
		select
			a.mrn,
			a.cpt,
			b.age
		from
			one a
			inner join two b
				on a.mrn=b.mrn;

	create table blah2 as
		select
			a.mrn,
			a.cpt,
			b.age
		from
			one a
			left join two b
				on a.mrn=b.mrn;
quit;

Re: proc sql left join vs proc sql inner join

Reeza — Mon, 02 May 2016 21:13:21 GMT

Use the one that makes logical sense. In this case you have a 1 to 1 match. If you always expect this use an inner match, if you can't trust this rule and always want the information from a specific table use the right/left join instead.

Re: proc sql left join vs proc sql inner join

LinusH — Tue, 03 May 2016 04:59:52 GMT

To extend @Reeza answer - if your logic holds for inner joins - those are better performing than outer joins generally speaking.

Re: proc sql left join vs proc sql inner join

dataMart87 — Fri, 06 May 2016 23:31:35 GMT

Thanks. I always thought INNER JOINS were not as efficient as a LEFT JOIN or a RIGHT JOIN because INNER JOINS check rows common to both tables, whereas LEFT JOINS or RIGHT JOINS just check for rows on either the left-hand or right-hand table.

I'm curious what the exceptions are to this rule, if any?