Nested do loop in marco generating empty datasets

[ChristianWI]

Background: I am conducting a sensitivity analysis where I am looping over combinations of two parameters A and B, where A=1,2 and B=1 to 9. I have saved down datasets for each of the 18 combinations of the two parameter values and I now would like to write a macro the generate summary statistics, by group, for 6 output variables I am interested in.

Problem: If I limit B to B=1 to 8 I get the 16 desired output datasets, however if I loop over B=1 to 9 I inexplicably get 21 output datasets, where the the output for the output for some iterations is duplicated, and trying to open the duplicates produces an error: "Cannot open dataset, file does not exist".

Although I do ultimately get the desired output, there must be something going wrong to produce the additional duplicate datasets.

Example code:

data data;
input group $ var1-var6;
datalines;
group1 1 2 3 4 5 6
group2 7 8 9 10 11 12
;
run;

%macro example;
%do i = 1 %to 2;
	%do j = 1 %to 9;
		%let case = A&i._B&j.;
		data case_A&i._B&j.;
		set data;
		case = "&case.";
		run;
	%end;
%end;
%mend;
%example

Any help would be greatly appreciated.

[ChrisNZ]

I doubt you create additional tables, check your log to verify. If not, then it looks like a display problem by the SAS client. In this case, you should contact SAS tech support.

[Sajid01]

I reran your code without any duplicates and I didn't find any duplicate datasets as you have mentioned. see it below. ( I executed it on SAS On Demand for Academics)

I have attached the log.

Needless to say that filesystems do not allow files of exactly the same name.

[ChrisNZ]

> Needless to say that filesystems do not allow files of exactly the same name.

I wonder what SAS sees if you have identical names but different cases under Linux/Unix. Like:

test.sas7bdat

Test.sas7bdat

SAS names are supposed to be lower case, so does SAS ignore the other files?

[Sajid01]

Unix/Linux are case sensitive.
For example test and Test are not the same.

[ChrisNZ]

> For example test and Test are not the same.

Exactly. Hence my question.

[Kurt_Bremser]

At least with 9.4M7 on AIX and EG 7.15, only .sas7bdat files in all lowercase will appear in the library list.

But IIRC, there was a time when SAS would list all files with extension .sas7bdat as datasets (including those that had a capital letter before the extension), but would have trouble opening those that were not all lowercase.

[ChristianWI]

Thanks for this. I did check the log and only 18 datasets were recorded as having been generated which added even more to my confusion. I forgot to mention this in my post, but it sounds like you are correct in that is is most likely a display problem.

[Tom]

As has been mentioned UNIX filesystems are case sensitive.  So a file named CASE_A2_B9.sas7bdat is not the same file as one named case_a2_b9.sas7bdat.

SAS on unix will only use lowercase letters in the names of files used to store SAS datasets.  So if you run this program:

data copy;
  set sasuser.CASE_A2_B9;
run;

It will look for a file name case_a2_b9.sas7bdat in the directory pointed to by the SASUSER libref.

Using a quoted physical name to reference the dataset in your SAS code will not help.  If you write code like:

data copy;
  set "~/sasuser/CASE_A2_B9.sas7bdat";
run;

Under the hood SAS will create a new libref pointing to the ~/sasuser/ directory and reference the dataset the same way as if you had use two level name to begin with.