Is there a faster way to join using an 'or' statement?

[UniversitySas]

I have two tables that I want to join, call them ID_table and merge_table.

data Id_table;
   input Name $ Parent_name $  var1 ;
   datalines;
ABC  ABC_Co 28.6
HLE  ABC_Co 42.1
Ggl_inc google 18.8
APPLE APPL_Inc 123.4
;
run;

data merge_table;
   input Comp_Name $ var2 ;
   datalines;
ABC_Co 110
Ggl_inc 105.3
APPLE 112.1
XYZ 101
;
run;

I want sas to merge on either: Name = comp_name, or Parent_name = comp_name.

If I do this:

PROC SQL;
	CREATE TABLE merge_test AS
	SELECT l.*, r.*
	FROM Id_table AS l LEFT JOIN merge_table AS r
	ON	 (l.Name = r.Comp_Name or l.Parent_name = r.Comp_Name);
QUIT;

It works. However, my actual data set has 582k obs and 300k obs in both tables, and when I run the above command it seems to take forever and SAS times out.

Is there a more efficient way to program this?

Furthermore, I still want to keep unmerged matches in my data set, but SAS doesn't seem to be doing so in the above test sample (the values for comp_name = XYZ from merge_table are dropped). Is there  a way it is included?

thanks!

[Kurt_Bremser · Accepted solution]

Maxim 2: Read the Log.

The SQL procedure notifies you that it has to do cartesian joins, which is often a bad sign, performancewise.

Split your join into two, and concatenate the results:

proc sql;
create table merge_test as
  select l.*, r.*
  from Id_table as l left join merge_table as r
  on l.Name = r.Comp_Name
  where not missing(r.comp_name)
union
  select l.*, r.*
  from Id_table as l left join merge_table as r
  on l.Parent_name = r.Comp_Name
  where not missing(r.comp_name)
;
quit;