About Kurt_Bremser

Kurt_Bremser · ‎05-05-2014

With the filename url access method, you can specify user= and pass=. Alternatively you could try to use wget to download to a local file and read that.

Kurt_Bremser · ‎05-05-2014

Depending on the setting of system option VALIDVARNAME, the new variables will be named _0 and _1 (VALIDVARNAME=V7) or 0 and 1 (VALIDVARNAME=ANY), respectively. A variable named 0 needs to be adressed as '0'n in the following code, which may be a hassle.

Kurt_Bremser · ‎05-05-2014

In addition to Ksharp: If SAS encounters an error (or already has an error condition set if you are in batch mode), the data set will not be replaced. But: If you test-run such code, you will be forced to rerun it from the start each time you need to control an intermediate stage of the data set. That's why I let all my data sets in a batch job have unique names.

Kurt_Bremser · ‎05-05-2014

When you have a chain of "else if"s where you evaluate the same function in every "if", the function is evaluated once for every "false" you get. With a select(), the function is evaluated exactly once. You could avoid that by assigning the result of the function to a variable and comparing against that. OTOH, a "select" enables a compiler (and SAS turns into a compiler when encountering a data step) to do certain optimizations which usually speed up at runtime. And, it is that much easier to read and conveys the intention of the programmer better. Readable code is a must for proper maintenance.

Kurt_Bremser · ‎04-30-2014

I quota reeza's post: Also, be careful with the in: as the matches may not be exactly what you want.

Kurt_Bremser · ‎04-30-2014

PS: You could even try to automate that from within SAS, but it requires shell scripting on the UNIX side. You can use the X statement in an rsubmit block to execute a shell script that does the fuser/ps evaluation and then writes the result to a file. A data step can then be used to read that file and put the contents into the SAS log, or use call system to do the kills, if you want to be that nasty.

Kurt_Bremser · ‎04-30-2014

How about this: data diags_78060 ..... diags_78791 ; set patients; array a_dx dx_code1-dx_code41; do i = 1 to dim(a_dx); select (a_dx(i)); when ('78060') output diags_78060; ..... when ('78791') output diags_78791; otherwise; end; end; run;

Kurt_Bremser · ‎04-30-2014

Everytime you have a succession of else if ..., you might consider using select: select (put(type,$diagx.)); when (1) output want1; ... when (7) output notfound; otherwise output error; end; It's easier to read, and more efficient, because it has to do the put function only once.

Kurt_Bremser · ‎04-30-2014

rsubmit is used when you want to run code on the server from a local SAS session over SAS/CONNECT. That's what led me to believe you are using CONNECT. If you have options comamid tcp; and signon servername; somewhere in your code (or in the autoexec.sas), then you are using SAS/CONNECT. When the server part of a SAS/CONNECT setup is on UNIX, the connection is established via the telnet protocol (see wikipedia about that). You can use the same credentials to manually establish a telnet session by using "telnet servername" in the "start - run" dialog of windows (or just enter telnet servername in a cmd window). Your example for the fuser command looks right. fuser gives you process number(s); ps -fp processnumber will show you the program name and the user who started the process. To kill another user's process, though, you will need root (superuser) privileges.

Kurt_Bremser · ‎04-30-2014

Since it is a UNIX system, you can use the fuser command to get the id of the process that has an open file handle to the file. Given that you work with SAS/CONNECT, you should be able to telnet to the server.

Kurt_Bremser · ‎04-30-2014

Does the patient ID change if the same person is admitted repeatedly? If not, you may need to include the actual date in the data set. Then try the following: proc sort data=HAVE; by PATIENT_ID DATE; run; data WANT; set HAVE; retain FIRST_OBS; if DAY = 0 then FIRST_OBS=LIPID_LEVELS; run; If, instead, patients get a new PATIENT_ID on each admission, change DATE in the sort to DAY;

Kurt_Bremser · ‎04-30-2014

If you have login access to the command line, there is no, _no_, NO!!!!! reason to prohibit xcmd. Stick that to your "administrators". x starts a shell that differs from a "normal" login shell only in its environment (working directory, some environment variables), but not in its privileges. Any havoc you can wreak on the system with the x command is even easier done from the CLI. Instead of bringing forth strawman arguments, they should harden their system by proper granting of rights, having quotas, workload management and so on. "My" users are reigned in by quota management in locations where they have write access (or I set up dedicated file systems that don't bother me if full), restricting the number of concurrent SAS sessions (so they don't cause excessive paging by running >30 SAS sessions at once, one of the idiots actually did that one time), and proper management of access rights to resources. I actually encourage them to use ssh and the x command to make their days easier, and I never had problems from that. If your admins think that only some users should have xcmd (and CLI access) enabled, then they only need to define an additional workspace server that only the privileged users have access to.

Kurt_Bremser · ‎04-29-2014

As Jaap said, the limit is the size of the command line buffer, which is usually 4K. Since the shell uses this buffer to expand the wildcards, gzip *.sas7bdat may lead to "parameter list too long". My version hands the * over to the find command, which does not expand the list, but instead matches the pattern against any file it finds. The files are then gzip'd one by one, a method that scales indefinitely, for all purposes. If you have to operate under a noxcmd restriction, you may still be able to ssh to the server and submit the command there. If I can solve a problem with a single shell command, I prefer that every time to a lengthy SAS code. Many problems can be solved with a hammer, but sometimes the screwdriver is better.

Kurt_Bremser · ‎04-29-2014

Is it possible to have completely identical records in one of the datasets?

Kurt_Bremser · ‎04-29-2014

The question is, should dataset result have x records where x is the sum of all records of one,two,three, regardless of the contents, or should records be eliminated if two identical records are found in different datasets?

Online Status	Offline
Date Last Visited	Monday