You MUST (and I mean MUST) read the documentation supplied in ASC_NTS.pdf, as it contains the essentially important information about the columns contained in the data files. Following this document, and my previous experience with the DEMO file, I wrote this code and ran it after downloading the file to my University Edition:
libname myfold '/folders/myfolders';
data myfold.drug19q4 (compress=yes);
infile '/folders/myfolders/DRUG19Q4.txt' dlm='$' dsd truncover firstobs=2;
input
primaryid :$10.
caseid :$8.
drug_seq
role_cod :$2.
drugname :$30.
prod_ai :$50.
val_vbm :$1.
route :$10.
dose_vbm :$30.
cum_dose_chr
cum_dose_unit :$7.
dechal :$1.
rechal :$1.
lot_num :$20.
_exp_dt :$10.
@
;
if length (_exp_dt) = 4
then do;
if _exp_dt > "19"
then exp_dt = mdy(1,1,input(_exp_dt,4.));
else exp_dt = mdy(input(substr(_exp_dt,1,2),2.),1,2000+input(substr(_exp_dt,3,2),2.));
end;
else if length(_exp_dt) = 6
then exp_dt = input(cats(_exp_dt,'01'),yymmdd8.);
else exp_dt = input(_exp_dt,yymmdd8.);
drop _exp_dt;
format exp_dt e8601da10.;
input
nda_num
dose_amt :$30.
dose_unit :$7.
dose_form :$30.
dose_freq :$4.
;
run;
which imported the file successfully, with only 2 NOTEs for really hare-brained date entries that do not match the doc at all.
The resulting dataset file has a size of 215MB, so should be easy to handle in UE. The compress option resulted in a 55 % decrease, so even the uncompressed file would not be a showstopper.
Note that I used a library in the shared folder, so the data is stored outside the VM, which further alleviates any storage problems.
You should now inspect the data to see how the strings are filled; if the maximum length of contents is smaller than the used informat, you can adapt the informat; if strings of a length corresponding to the informat length are found, the informat needs to be increased, until you get a few blanks left with even the largest input string.
