Given that the data don't include cycles, this is an elementary exercise of traversing a tree (or trees). It is a pity that we cannot pass (hash) object references to user-written functions. If it were possible, then I would have written the recursive part as a user-written function. Here, I use the link blocks instead. The link blocks allow recursive "calls" up to 10 or 11 deep by default. But this limit can be relaxed by using the stack= option to the data statement. (This option is now documented in 9.2.) You need to create and manage call stacks by yourself, though.
/* test data */
data one;
input boss $ sub $;
cards;
Jim Kate
Lisa Jim
Dave Lisa
Dave Erik
Anna Mary
Anna Ben
Erik Anna
. Dave
Mary Tom <-- new data
Mary Bill <-- new data
;
run;
/* find all the subs */
%let maxDepth = 100;
data _null_ / stack=&maxDepth;
/* main */
link setup;
do until (end);
set one end=end;
thisBoss = boss;
link find;
end;
link out;
return;
/* helpers */
setup:
/* hashes */
dcl hash in(dataset:"work.one", multidata:'y');
in.definekey('boss');
in.definedata('boss','sub');
in.definedone();
dcl hash out(ordered:'yes');
out.definekey('boss','sub');
out.definedata('boss','sub');
out.definedone();
/* const and global vars */
retain OK 0;
length thisBoss $8;
/* call stack */
array stack_boss[&maxDepth] $8 _temporary_;
array stack_sub[&maxDepth] $8 _temporary_;
array stack_rc[&maxDepth] _temporary_;
array stack_i[&maxDepth] _temporary_;
array stack_j[&maxDepth] _temporary_;
stack_top = 0;
return;
add:
out.add(key:thisboss, key:sub, data:thisboss, data:sub);
return;
find:
link push;
if missing(thisboss) | missing(sub) then goto outoffind; /* no point adding */
if out.check(key:thisBoss, key:sub)=OK then goto outoffind; /* already there*/
link add;
/* follow */
boss = sub;
sub = '';
/* find_next() gets reset as soon as another find() is called.
thus we have to call find() and the number of find_next()
for each sibling here. 😞 */
do i = 0 by 1 until(rc^=OK);
rc = in.find();
do j = 1 to i;
rc = in.find_next();
end;
if rc = OK then link find;
end;
outoffind:
link pop;
return;
pop:
boss = stack_boss[top];
sub = stack_sub[top];
rc = stack_rc[top];
i = stack_i[top];
j = stack_j[top];
top + (-1);
return;
push:
top + 1;
if top > &maxDepth then do;
put "ERROR: stack overflow";
stop;
end;
stack_boss[top] = boss;
stack_sub[top] = sub;
stack_rc[top] = rc;
stack_i[top] = i;
stack_j[top] = j;
return;
out:
rc = out.output(dataset:"work.two");
return;
run;
/* check */
proc print data=two;
run;
/* on lst
Obs boss sub
1 Anna Ben
2 Anna Bill
3 Anna Mary
4 Anna Tom
5 Dave Anna
6 Dave Ben
7 Dave Bill
8 Dave Erik
9 Dave Jim
10 Dave Kate
11 Dave Lisa
12 Dave Mary
13 Dave Tom
14 Erik Anna
15 Erik Ben
16 Erik Bill
17 Erik Mary
18 Erik Tom
19 Jim Kate
20 Lisa Jim
21 Lisa Kate
22 Mary Bill
23 Mary Tom
*/
Are you ready for the spotlight? We're accepting content ideas for SAS Innovate 2025 to be held May 6-9 in Orlando, FL. The call is open until September 25. Read more here about why you should contribute and what is in it for you!
Learn how use the CAT functions in SAS to join values from multiple variables into a single value.
Find more tutorials on the SAS Users YouTube channel.