Sort in a data step to get more compact code

[SasStatistics]

I often find myself creating a data-set and then sorting it since for instance I want to do an visual inspection of my data. 

For example, with pseudo code it looks like: 

EDIT: When creating the data set, I do some other things such as keeping some columns and create a new one. 

* Creating my data-set; 

data MyData;
 set SomeDataSet;
 keep colA colB;
 NewCol = ColA - ColB;
run; 

* Sorting my data-set;

proc sort data = MyData; 
   by ColumnA; 
run; 

Now, this quickly becomes alot of code, can this be done more compactly? For instance sort the data already in the data step? Any other methods to recommend? 

Thanks. 

[andreas_lds]

The first step is not necessary at all, just use

proc sort data= SomeDataset out = MyData; 
   by ColumnA; 
run; 

[SasStatistics]

Thanks, please view my edit, my pseudo code example was bad.

[PaigeMiller]

I think your idea of "compact code" and my idea of "compact code" are probably different.

If you need to do a subtraction in a DATA step, and then a sort, then these are two different processes. You can't do them both in one step * so you are left with the need to do these two different processes in two steps. NOTE: there are times when you can do two different processes in one step, this isn't one of them.

* — I suppose you could use hash objects to do all of this in one data step, if you want to use an advanced method.

[andreas_lds]

---

@SasStatistics wrote:
Thanks, please view my edit, my pseudo code example was bad.

---

Yes, it was 😉

As @PaigeMiller pointed out: one could use a hash-object to do the sorting in the data step, but you would have to write more than three lines of code to get the same result, so using easy-to-read steps is by far more desirable than having one big step.

[Kurt_Bremser]

You could do a sort in a data step on the fly, but it can only work as long as the complete dataset (including the B-tree) can be kept in memory, and it would be A LOT MORE code.

If you want to avoid storing an intermediate dataset, you can use your data step to define a view and use that as input in the PROC SORT, but you will still have to write the code (even a little more, as you need the /VIEW= in the DATA statement).

[Astounding · Accepted solution]

If you switch from a DATA step to PROC SQL, you can do it in one step.  My SQL is rusty at best so you will need to test for syntax errors:

proc sql noprint;
create table MyData as
  select ColA, ColB, NewCol as ColA - ColB,
  from SomeDataSet,
  order by ColA;
quit;