Check all text files have same number of fields when reading multiple datasets in using wildcard

[bh11]

Hi!  I am reading in several text files using a wildcard in the infile statement and I include additional statements to store the name of the file in a separate variable.  Each text file includes the variable names in the first row.

DATA work.test
     LENGTH filename txt_file_name $256;
     RETAIN txt_file_name;
     INFILE "C:\samplepath\*.txt" 
          EOV = eov
          FILENAME = filename
          DLM = ","
          MISSOVER
          DSD
          LRECL = 32767;
     INPUT@;
          IF _N_ = 1 OR EOV THEN DO;
               txt_file_name = scan(filename, -1,"\");
               EOV = 0;
               DELETE;
               END;
          ELSE
       INPUT .....

The files that I am reading in 'should' all have the same structure.  Is there a way to test within this data step that the files being read in all have the same number of variables? I am trying to avoid reading them all in separately and checking that way.  I was thinking that replacing the MISSOVER option with the STOPOVER option might work, but I get an error message "INPUT statement exceeded record length."  Any help is greatly appreciated.  Thanks!

[SASJedi · Accepted solution]

Try something like this:

/* Make flat files to play with */
title; footnote;
ods csvall file="C:\temp\t1.txt";
proc print data=sashelp.class(where=(Sex='F')) noobs;
run;
ods csvall close;
ods csvall file="C:\temp\t2.txt";
proc print data=sashelp.class(where=(Sex='M')) noobs;
run;
ods csvall close;
data _null_;
   file "C:\temp\t3.txt" dsd ;
   set sashelp.class (obs=3);
   if _n_=1 then put '"Name","Sex","Height","Weight"';
   put Name Sex Height Weight;
run;

/* Proof of Concept */
DATA work.test;
     LENGTH filename txt_file_name _header _last_header $ 256;
     RETAIN txt_file_name _header _last_header ;
     drop _:;
     INFILE "C:\temp\t*.txt" 
          EOV = eov
          FILENAME = filename
          DLM = ","
          MISSOVER
          DSD
          LRECL = 32767;
     INPUT@;
          IF _N_ = 1 OR EOV THEN DO;
               txt_file_name = scan(filename, -1,"\");
               EOV = 0;
               _header=_infile_;
               if _last_header ne _header and _n_ ne 1 then do;
                  PUT "ERROR: File structure for " txt_file_name " has changed.";
                  PUT "ERROR- New file structure:      " _header;
                  PUT "ERROR- Previous file structure: " _last_header;
                  stop;
               end;
               _last_header=_header;
               DELETE;
          END;
          ELSE INPUT Name:$15. Sex:$1. Age Height Weight;
run;

[bh11]

I tested this using sample datasets I created where there was a variable removed, a variable added, and a variable name changed. Once I adjusted the LENGTH statement (the _header and _last_header variables were very long for my datasets) this worked perfectly.  Thanks!