Pass array into each object in JSON file. (Proc JSON SAS) - arrays

Question

I'm trying to export a dataset to a JSON file. With PROC JSON every row in my dataset is exported nicely.
What I want to do is to add an array into each exported object with data from a specific column.
My dataset has structure like this:
data test;
input id $ amount $ dimension $;
datalines;
1 x A
1 x B
1 x C
2 y A
2 y X
3 z C
3 z K
3 z X
;
run;
proc json out='/MYPATH/jsontest.json' pretty nosastags;
export test;
run;
And the exported JSON object looks, obviously, like this:
[
{
"id": "1",
"amount": "x",
"dimension": "A"
},
{
"id": "1",
"amount": "x",
"dimension": "B"
},
{
"id": "1",
"amount": "x",
"dimension": "C"
},
...]
The result I want:
For each id I would like to insert all of the data from the dimension column into an array so my output would look this this:
[
{
"id": "1",
"amount": "x",
"dimensions": [
"A",
"B",
"C"
]
},
{
"id": "2",
"amount": "y",
"dimensions": [
"A",
"X"
]
},
{
"id": "3",
"amount": "z",
"dimensions": [
"C",
"K",
"X"
]
}
]
I've not been able to find a scenario like this or some guidelines on how to solve my problem. I hope somebody can help.
/Crellee

Answer 1

There are other methods for json output, including
hand-coded emitter in DATA Step
JSON package in Proc DS2
Here is an example of a hand-coded emitter for your data and desired mapping.
data _null_;
file 'c:\temp\test.json';
put '[';
do group_counter = 1 by 1 while (not end_of_data);
if group_counter > 1 then put #2 ',';
put #2 '{';
do dimension_counter = 1 by 1 until (last.amount);
set test end=end_of_data;
by id amount;
if dimension_counter = 1 then do;
q1 = quote(trim(id));
q2 = quote(trim(amount));
put
#4 '"id":' q1 ","
/ #4 '"amount":' q1 ","
;
put #4 '"dimensions":' / #4 '[';
end;
else do;
put #6 ',' #;
end;
q3 = quote(trim(dimension));
put #8 q3;
end;
if dimension_counter > 1 then put #4 '}';
put #2 ']';
end;
put ']';
stop;
run;
Such an emitter can be macro-ized and generalized to handle specifications of data=, by= and arrayify=. Not a path recommended for friends.

Answer 2

You can try concatenating / grouping the text before calling proc json.
I don't have proc json in my SAS environment, but try this step and see it works for you:
data want;
set test (rename=(dimension=old_dimension));
Length dimension $200. ;
retain dimension ;
by id amount notsorted;
if first.amount = 1 then do; dimension=''; end;
if last.amount = 1 then do; dimension=catx(',',dimension,old_dimension); output; end;
else do; dimension=catx(',',dimension,old_dimension); end;
drop old_dimension;
run;
Output:
id=1 amount=x dimension=A,B,C
id=2 amount=y dimension=A,X
id=3 amount=z dimension=C,K,X