Extend observations for all years in sequence - sql

Question

I have 2 sets.
First one is big (~1000k rows), it contains patient observation data grouped by observation year, from, lets say 2000 to 2005. In this set there are some patients that contain observations for all years (or should I say for each year in sequence), and there are some that has, for example, observations for year 2002-2003 only.
The second set contains only sequence of years from 2000 till 2005, 6 rows.
What I want to have is a table with the data from set 1 for each patient, but extended so that for each patient I would see observations for each year from set 2, and if there were not any observation for particular year in set 1, the empty rows should be added or emptyness (or better "-") in the data column only.
For example set 1 could be:
patient_id | obs_year | data
a 2000 10
a 2001 12
a 2002 13
a 2003 9
a 2004 1
a 2005 6
bb 2002 100
bb 2003 110
Set 2 is like:
year |
2000
2001
2002
2003
2004
2005
So what I want in result ideally would be like this:
patient_id | obs_year | data
a 2000 10
a 2001 12
a 2002 13
a 2003 9
a 2004 1
a 2005 6
bb 2000 -
bb 2001 -
bb 2002 100
bb 2003 110
bb 2004 -
bb 2005 -
I should also mention that I do this job in SAS, so SQL query or SAS script (or both )solutions are welcomed.

Answer 1

Dedup your patient_id from set 1 in a sort. Merge this onto set 2 to give every patient_id against the years, then merge this back onto set 1 by patient_id and year to give your output. Anywhere that patient_id and year do not match will be blank as in your desired output

Answer 2

Another option is PROC FREQ with sparse, which produces a line for every possible combination whether they appear or not. This works if you don't have any legitimate zeroes in the data; if you do and care that they're different from missing, this won't work.
proc freq data=have noprint;
weight data;
tables patient_id*obs_year/missing sparse out=want(rename=count=data keep=count patient_id obs_year);
run;
Then you need to convert 0 back to missing, if you care about the difference (presumably in the next step, if there is one).
A similar approach that is closer to the desired results is proc tabulate with printmiss, which works similarly to sparse:
proc tabulate data=have out=want(keep=patient_id obs_year data_sum rename=data_sum=data);
class patient_id obs_year;
var data;
tables patient_id,obs_year*data*sum='data'/printmiss misstext='.';
run;
That actually does get you missing values properly.