Start out by making your data intelligent. Right now it's outright "stupid". See Maxim 33.
Values that deal with dates and times should be stored as such, so the mountain of SAS functions, formats and other tools for such values can be used.
Therefore t1 will look like this:
data t1;
input anomes yymmdd10.;
format anomes yymmn6.;
datalines;
2018-10-01
2019-02-01
;
Next, we don't need t2 at all, as SAS provides a format that converts dates to quarters on the fly.
t3 will look like this:
data t3;
input quarter yymmdd10. car :$10.;
format quarter yyq6.;
datalines;
2019-01-01 BMW
2019-01-01 MB
2019-01-01 Porsche
2019-01-01 Tesla
2019-01-01 Jaguar
2019-04-01 Nissan
2019-04-01 Ferrari
2019-04-01 Audi
2019-04-01 VW
2019-04-01 Mazda
;
The code for the join will be this:
%let cur_qu=%sysfunc(today(),yyq6.);
proc sql;
create table want as (
select t1.anomes,t3.car
from t1 left join t3
on put(anomes,yyq6.) = put(quarter,yyq6.)
where t3.quarter ne .
union all
select a.anomes,b.car
from (
select t1.anomes
from t1 left join t3
on put(t1.anomes,yyq6.) = put(t3.quarter,yyq6.)
where t3.quarter = .
) a,
t3 b
where put(b.quarter,yyq6.) = "&cur_qu"
)
order by anomes
;
quit;
The first select finds the matches, the second finds the non-matches and joins them to the current quarter.
As you can see, I added a second date in t1 to illustrate that the code works for both possibilities.
The resulting table will be this (result of proc print):
anomes car
201810 Nissan
201810 Ferrari
201810 Audi
201810 VW
201810 Mazda
201902 Porsche
201902 Jaguar
201902 MB
201902 BMW
201902 Tesla
Note how data steps with datalines are used to create example data. This is the required way to post data here so that answers can be quickly found and tested.
