How to create a SAS dataset for each individual trading day (TAQ) data and save them to a file - sas

Question

I have the daily trading data (TAQ data) for a month. I am trying to unzip each of them.
The folder's name is EQY_US_ALL_TRADE_202107.
It has several zipped (GZ files) files for each trading day named as
EQY_US_ALL_TRADE_202210701
EQY_US_ALL_TRADE_202210702
EQY_US_ALL_TRADE_202210703 ...
EQY_US_ALL_TRADE_202210729
I want to create a SAS dataset for each individual day and save them to a file. As far as I understand, I need a do loop to go through a month of daily TAQ data and calculate the trade duration and then just save the relevant data to a file so that each saved data set would be small, and then I have to aggregate them all up. For calculating trade duration, I am just taking the difference of the "DATETIME" variable, (ex. dif(datetime))
Until now, I have been working by making one working directory (D:\MainDataset) and doing calculations in it starting with unzipping files. But it is taking too much time and disk space. I need to create separate datasets for each trading day and save it to a file.
data "D:\MainDataset" (keep= filename time exchange symbol saleCondition tradeVolume tradePrice tradeStopStock
tradeCorrection sequenceNumber tradeId sourceOfTrade tradeReportingFacility
participantTimeStamp tradeReportingFacilityTimeStamp);
length folderef $8 time $15. exchange $1. symbol $17. saleCondition $4. tradeStopStock $1.
sourceOfTrade $1. tradeReportingFacility $1.
participantTimeStamp $15. tradeReportingFacilityTimestamp $15.;
rc=filename(folderef,"D:\EQY_US_ALL_TRADE_202107");
did = dopen(folderef);
putlog did=;
/* do k = 1 to dnum(did); Use this to run the loop over all files in the folder */
do k = 1 to 3;
filename = dread(did,k);
putlog filename=;
if scan(filename,-1,'.') ne 'gz' then continue;
fullname = pathname(folderef) || '\' || filename;
putlog fullname=;
do while(1);
infile archive zip filevar=fullname gzip dlm='|' firstobs=2 obs=5000000 dsd truncover eof=nextfile;
input time exchange symbol saleCondition tradeVolume tradePrice tradeStopStock
tradeCorrection sequenceNumber tradeId sourceOfTrade tradeReportingFacility
participantTimeStamp tradeReportingFacilityTimeStamp;
output;
end;
nextfile:
end;
stop;
run;
Proc contents data = "D:\MainDataset";
run;
proc print data ="D:\MainDataset" (obs = 110);
run;

Answer 1

Create code to process one file. Probably coded as a macro that takes as input the name of the file to read and the name of the dataset to create.
%macro taq(infile,dataset);
data &dataset;
infile "&infile" zip gzip dsd dlm='|' truncover firstobs=2;
....
run;
%mend taq;
Then generate a dataset with the names of the files to read and the dataset names you want to create from them. So perhaps something like this:
%let ym=202107;
%let folder=D:\EQY_US_ALL_TRADE_&ym;
data taq_files;
length dataset $32 filename $256 ;
keep dataset filename;
rc=filename('folder',"&folder");
did=dopen('folder');
do fnum=1 to dnum(did);
filename=catx('\',"&folder",dread(did,fnum));
dataset=scan(filename,-2,'./\');
if 'gz'=lowcase(scan(filename,-1,'.')) then output;
end;
did=dclose(did);
rc=filename('folder');
run;
Now that you have the list of files you can use it to call the macro once for each file.
data _null_;
set taq_files;
call execute(cats('%nrstr(%taq)(',filename,',',dataset,')'));
run;
The body of the macro can include the code to both read the values from the delimited files and calculate any new variables you want. There should not be any need to do that in multiple steps based on what you have shown so far.
Your logic for converting the timestamp strings into time values seems overly convoluted. Just use informats that match the style of the strings in the file. For example if the strings start with 6 digits that represent HHMMSS then read that using the HHMMSS6. informat. If the filenames has digit strings in the style YYYYMMDD then read that using the YYMMDD8. informat.
Note that a text file that is compressed to 2Gbytes will generate a dataset that is possibly 10 to 30 times that large. You might want to define the individual datasets as views instead to avoid having the use that space by changing the DATA statement:
data &dataset / view=&dataset ;