You could use the %dirtree macro that I found here: https://github.com/sasutils/macros/blob/master/dirtree.sas  %macro dirtree /*--------------------------------------------------------------------------- Build dataset of files in directory tree(s) ----------------------------------------------------------------------------*/ (directory /* Pipe delimited directory list (default=.) */ ,out=dirtree /* Output dataset name */ ,maxdepth=120 /* Maximum tree depth */ ); /*--------------------------------------------------------------------------- Use SAS functions to gather list of files and directories directory - Pipe delimited list of top level directories out - Dataset to create maxdepth - Maximum depth of subdirectories to query Output dataset structure --NAME-- Len Format Description FILENAME $256 Name of file in directory TYPE $1 File or Directory? (F/D) SIZE 8 COMMA20. Filesize in bytes DATE 4 YYMMDD10. Date file last modified TIME 4 TOD8. Time of day file last modified DEPTH 3 Tree depth PATH $256 Directory name Size is not available for the directories. LASTMOD timestamp is only available on Unix for directories. Will not scan the subtree of a directory with a path that is longer then 256 bytes. For such nodes TYPE will be set to L . ----------------------------------------------------------------------------*/ %local fileref ; %let fileref=__FL__ ; %if 0=%length(&directory) %then %let directory=. ; * Setup dataset and seed with starting directory list ; data &out; length filename $256 type $1 size 8 date time 4 depth 3 path $256 ; retain filename ' ' depth 0 type ' ' date . time . size . ; format size comma20. date yymmdd10. time tod8. ; do _n_=1 to countw(symget('directory'),'|'); path=scan(symget('directory'),_n_,'|'); output; end; run; %* Allow use of empty OUT= dataset parameter ; %let out=&syslast; data &out; modify &out; retain sep "%sysfunc(ifc(&sysscp=WIN,\,/))"; retain maxdepth &maxdepth; * Create FILEREF pointing to current file/directory ; rc1=filename("&fileref",catx('/',path,filename)); if rc1 then do; length message $256; message=sysmsg(); put 'ERROR: Unable to create fileref for ' path= filename= ; put 'ERROR- ' message ; stop; end; * Try to open as a directory to determine type ; did=dopen("&fileref"); type = ifc(did,'D','F'); if type='D' then do; * Make sure directory name is not too long to store. ; if length(catx('/',path,filename)) > vlength(path) then do; put 'NOTE: Directory name too long. ' path= filename= ; type='L'; rc3=dclose(did); end; else do; * Move filename into the PATH and if on Unix set lastmod ; path=catx(sep,path,filename); filename=' '; if sep='/' then do; lastmod = input(dinfo(did,doptname(did,5)),nldatm100.); date=datepart(lastmod); time=timepart(lastmod); end; end; end; else do; * For a file try to open file and get file information ; fid=fopen("&fileref",'i',0,'b'); if fid then do; lastmod = input(finfo(fid,foptname(fid, 5)), nldatm100.); date=datepart(lastmod); time=timepart(lastmod); size = input(finfo(fid,foptname(fid,ifn(sep='/',6,4))),32.); rc2 = fclose(fid); end; end; * Update the observation in the dataset ; replace; if type='D' then do; * When current file is a directory add directory members to dataset ; depth=depth+1; if depth > maxdepth then put 'NOTE: ' maxdepth= 'reached, not reading members of ' path= ; else do i=1 to dnum(did); filename=dread(did,i); output; end; rc3=dclose(did); end; * Clear the fileref ; rc4=filename("&fileref"); run; %mend dirtree; Then use the macro to create a directory listing and delete all listed files and folders. %let topFolderToDelete=c:\temp\deldirtest; /* create table work.dirtree with directory listing */ %dirtree(&topFolderToDelete); proc sort data=work.dirtree; by DESCENDING path DESCENDING type; run; /** delete files and folders **/ /* 1. folder separator \ or / (Win or Unix/Linux)? */ data _null_; set work.dirtree; if find(path,'\') then sep='\'; else sep='/'; if not missing(sep) then do; call symputx('sep',sep); stop; end; run; /* 2. delete files and folders */ data _null_; set work.dirtree; by DESCENDING path DESCENDING type; fname='tempfile'; if type='F' then rc=filename(fname, catx("&sep",path,filename)); else rc=filename(fname, path); if rc = 0 and fexist(fname) then rc=fdelete(fname); rc=filename(fname); run;   ... View more

Going forward please create a new question for new requests. It keeps the Q&A leaner and cleaner and though makes it easier to find answers searching the forum.   The parquet libname engine is only available under Viya. You need first to copy your .sas7bdat files to the Viya side. Below sample code demonstrates how to convert a .sas7bdat file to the Parquet file format. The libname parquet engine has imho still some quirks (like when it comes to deletion of the files) which I've already raised with SAS Tech Support. Right now it just needs some extra coding for some operations. options fullstimer msglevel=i; /* create directories for demo */ %let sv_dlcreatedir=%sysfunc(getoption(dlcreatedir)); options dlcreatedir; libname _dir "&_userhome/demo"; libname _dir "&_userhome/demo/source"; libname _dir "&_userhome/demo/target"; libname _dir clear; options &sv_dlcreatedir; /* assign source library with .sas7bdat files */ libname source "&_userhome/demo/source"; /* create sample source data */ data source.class(index=(name sex)); set sashelp.class; run; data source.cars; set sashelp.cars; run; /* assign target library for parquet files */ libname target parquet "&_userhome/demo/target"; /* copy data from source to target */ proc datasets lib=source nodetails; copy in=source out=target ; quit; /** housekeeping: remove demo folder and all its content **/ %macro housekeeping(); /* delete SAS files in source */ proc datasets lib=source kill; quit; /* delete source folder */ filename source "%sysfunc(pathname(source))"; data _null_; rc=fdelete("source"); run; filename source clear; libname source clear; /* delete parquet files in target */ proc sql; create table work.file_list as select l.libname, lowcase(l.engine) as engine, lowcase(r.memname) as memname, l.path from dictionary.libnames l inner join dictionary.tables r on l.libname='TARGET' and l.libname=r.libname ; quit; data _null_; set work.file_list; fname='tempfile'; length fully_qualified_path $2000; fully_qualified_path=cats(path,'/',memname,'.',engine); rc=filename(fname, fully_qualified_path); if rc = 0 and fexist(fname) then rc=fdelete(fname); rc=filename(fname); run; /* delete target folder */ filename target "%sysfunc(pathname(target))"; data _null_; rc=fdelete("target"); run; filename target clear; libname target clear; /* delete demo folder */ filename demo "&_userhome/demo"; data _null_; rc=fdelete('demo'); run; filename demo clear; %mend; /* %housekeeping(); */ If you run above code as-is you should see the following in SAS Studio:   Please note Parquet files are immutable meaning Write Once. You can't insert, delete or update a parquet file. You can only fully replace it. Also consult the docu. It appears there are libname options allowing you to read/write parquet files directly from/to S3. STORAGE_PLATFORM= LIBNAME Statement Option In my mind a data flow under Viya could look like: Upstream source data (any format but could be parquet). Data manipulation ("ETL") mainly under Compute (your "SAS9.4" under Viya).  Intermediary data still .sas7bdat. Endpoint of process for load into CAS or for downstream non-SAS applications potentially parquet and potentially written to S3 (only if table doesn't require further change later on). Load into CAS for analysis and reporting I also suggest you search the SAS Communities with keyword "parquet". There are many useful Library Articles available.   Also read this discussion. It appears that currently when using the parquet engine with the compute libname statement a single file gets created but when using the cas caslib statement then multiple files get created under a folder with the name of the parquet file. The libname and caslib can read each others structures but I guess this could create some challenges when it comes to managing the output under S3. From how I understand things right now one best sticks to a design where a specific set of parquet files only ever gets created either under Compute with a libname or under CAS with a caslib. I'm not sure but I assume that creating the parquet file in chunks will better support multi-threading for reading the data into CAS (or writing from CAS to parquet).     Here and example how the parquet file looks like when created via caslib/proc casutil. You reference such a structure in code still as <caslib>.class_casutil.     ... View more

Below uses amended sample data so it contains at least one case with 4 repeated values. I've also added a variable visit_date. I do hope that your real data contains sufficient data to ensure your analysis runs against data in the required sort order. It would also help if you show us the desired result as this will clarify which rows you would want flagged (all of them or only from the 4th repetition onwards).  data have ; infile datalines dsd dlm=' ' truncover; input patient cpevent$ vstest$ vsren vsorresu$; visit_date='01jan2024'd+_n_; format visit_date date9.; cards ; 100 base pr 78 min 100 base pr 78 min 100 scrn hr 72 /min 100 week1 sbp 70 mmhg 100 week7 dbp 110 mm 100 week21 weight 75 kg 100 fwp height 120 kg 100 base pr 78 min 100 base4 pr 78 min 100 base1 pr 78 min 100 base6 pr 78 min 100 base pr 78 min 100 scrn hr 79 /mn 100 week1 sbp 70 mmhg 100 week7 dbp 110 mm 100 week21 weight 80 pounds 100 fwp height 75 kg ; data want; set have; by patient visit_date; if not first.patient and lag(vsren)=vsren and lag(vsorresu)=vsorresu then occurrence_cnt+1; else occurrence_cnt=1; multi_flg= (occurrence_cnt>=4); run; proc print data=want; run; ... View more