About JackHamilton

JackHamilton · ‎06-13-2019

After a few reboots for other reasons, SAS appears for the moment to be working in both Jupyter Notebooks and Atom. It also works in Jupyter Lab. It may depend on the order in which programs are installed.

JackHamilton · ‎06-12-2019

Interestingly, the problem is now solved for Jupyter but the sas_kernel no longer works in Atom. It shows up in the kernel list, but doesn't actually run. This must the Open Source equivalent of Microsoft's DLL Hell.

JackHamilton · ‎06-12-2019

Update: Although the "where" command indicated that "pip" should be finding the main Anaconda pip, I looked around to see what other copies there might be on the system. A surprising number, it turned out. I got rid of all of them, re-installed the sas_kernel, and it worked! I think the pip it was finding was from the Atom editor, because uninstalling that was the last thing I did before the sas_kernel started working again. Atom didn't seem to be in the system path, but Windows has many ways to find executables. Now to re-install Atom and see if anything stops working again. Thanks for your help.

JackHamilton · ‎06-10-2019

Hello, I uninstalled Miniconda, sas_kernel, and saspy. As far as I could tell, everything that's disk-based was deleted or moved to a backup directory. I then installed the full Anaconda and issued these commands: pip install sas_kernel conda install -c conda-forge saspy jupyter kernelspec list The commands appeared to execute successfully, but the result from the last command was: Available kernels: python3 C:\Users\c449630\AppData\Local\Continuum\anaconda3\share\jupyter\kernels\python3 Note: no SAS kernel shown. I then started a Jupyter notebook. In the first cell I put import saspy saspy That printed <module 'saspy' from 'C:\\Users\\c449630\\AppData\\Local\\Continuum\\anaconda3\\lib\\site-packages\\saspy\\__init__.py'> I had saved a copy of my working sascfg_personal.py file, and I copied it into the directory shown. In the next cell I entered: %%SAS proc print data=sashelp.class (obs=1); run; And the result was: Using SAS Config named: winiomsolaris SAS Connection established. Subprocess id is 12040 Out[2]: The SAS System Obs Name Sex Age Height Weight 1 Alfred M 14 69 112.5 So saspy is installed and configured well enough to start a SAS session on the server and talk to it. Any ideas? Why is Jupyter not finding the SAS kernel? The kernel also doesn't show up in Jupyter Lab.

JackHamilton · ‎06-03-2019

Does a SAS cell work in Jupyter Notebook? I used to have the same problem in Lab, but I can't tell you how I fixed it other than that I reinstalled and updated everything.

JackHamilton · ‎05-31-2019

You could do the rankings in a single pass with no explicit sorting using hash objects or arrays in a data step. I wonder which of the three methods would be faster. Hash objects might be very fast with the right programming.

JackHamilton · ‎05-30-2019

So you want <= 2 years to have passed between the earliest treatment date and the latest discharge date? Or between the latest treatment date and the earliest admit date? Also, "year" is not a well-defined term. Depending on how you define it, two years could be anywhere between 366 days and 1,095 days (look at the different definitions of year in the documentation for the YRDIF function, and that doesn't cover all the possibilities). Because SAS dates are measured in days, it would be better to say something like "within 730 days". 24 %put DAYS=%sysevalf(%sysfunc(yrdif('01jan2000'd, '31dec2002'd)), integer); DAYS=2 25 26 %put DAYS=%sysevalf('31dec2002'd - '01jan2000'd); DAYS=1095

JackHamilton · ‎05-29-2019

It doesn't have to pass a restriction to the database; all it has to do is to stop when it has enough records. How many records it brings back and then throws away is likely to depend on buffer sizes and other options. My experience with Oracle is that it does not read the entire data set when you have specified OBS=. Something I would like to know is how to discover how many records were transferred through SAS/Access. If you are reading a partitioned table and specify SASTRACE=',,t,' you will get a display in the log showing the number of records processed by each thread. But not if it's a non-partitioned table.

JackHamilton · ‎05-20-2019

I hadn't though of using DROP=_ALL_. That's a very clean and extensible solution.

JackHamilton · ‎05-17-2019

NOBS= is OK if you just created the data set yourself in the same program, so you know it hasn't been edited. If you've set POINTOBS=NO, you can use NOBS= but not POINT=. REUSE=YES causes POINTOBS to be set to know. You might want to use POINTOBS=NO because, according to the documentation, writing a compressed data set with POINTOBS=YES increases CPU usage by 10% when creating a compressed data set. There's a function, ATTRN, which gives more information about the number of observations in a data set, and potentially overcomes the problem of deleted observations. https://documentation.sas.com/?docsetId=lefunctionsref&docsetTarget=p12uchp7hm5h2zn1om2ut816af7h.htm&docsetVersion=9.4&locale=en

JackHamilton · ‎05-17-2019

The monotonic function can be useful, but it is not documented, and is not guaranteed to return sequential numbers with no gaps starting from 1. That's not promised by the definition of monotonic: "varying in such a way that it either never decreases or never increases". In his paper on adding sequence numbers, http://support.sas.com/resources/papers/proceedings14/1277-2014.pdf , Howard Schreier says: "The catch is that the MONOTONIC function is not documented. Organizations may have policies prohibiting the use of undocumented features. Even in the absence of such a policy, a coder may be reluctant to rely on the function. In developing examples to illustrate the usage of the MONOTONIC function, the author encountered some surprising and disconcerting results. In some cases, series returned by MONOTONIC began not with 1 (one), but rather with higher integers. Unfortunately, it seemed not possible to produce such results consistently, or to experimentally determine just what circumstances cause the surprising and suspicious offsets."

JackHamilton · ‎05-17-2019

As ballardw notes, the question of whether you want the next to last record is physical order or in sorted order is an important one. The approach using NOBS= is potentially dangerous. It works only for native SAS data sets (or data step views) that have not been edited. It potentially does not work for data sets in a database, or transport data sets, or data sets that have been edited. Using the data above 83 data _null_; 84 i = nobs-1; 85 set emp nobs=nobs point=i; 86 put (_all_)(=); 87 stop; 88 run; EMP_ID=107 Name=G Salary=30000 Sal_Date=30NOV2020 but 90 proc sql; 91 delete from emp 92 where emp_id = 107; NOTE: 1 row was deleted from WORK.EMP. 93 quit; NOTE: PROCEDURE SQL used (Total process time): real time 0.00 seconds cpu time 0.00 seconds 94 95 data _null_; 96 i = nobs-1; 97 set emp nobs=nobs point=i; 98 put (_all_)(=); 99 stop; 100 run; EMP_ID=. Name= Salary=. Sal_Date=. i=7 nobs=8 EMP_ID=. Name= Salary=. Sal_Date=. _ERROR_=1 _N_=1 This problem is a variation on calculating lags and leads. See this 2019 SAS Global Forum paper for some ideas: https://www.sas.com/content/dam/SAS/support/en/sas-global-forum-proceedings/2019/3699-2019.pdf

JackHamilton · ‎05-15-2019

I often use CARDS4 (4 fewer characters to type than DATALINES4) but I've never used the LINES alias. I suppose it's been around for 30 years and I'd just never noticed it. I think DATALINES is newer than CARDS, but I don't remember when I first saw it.

JackHamilton · ‎10-23-2018

Perhaps your Windows file association for .ova files has gotten messed up. Or perhaps mine has.

JackHamilton · ‎10-23-2018

Works for me in Chrome 70 under Windows 10.

Re: Using FINDW in a DO loop to flag keywords in a string

Re: Using FINDW in a DO loop to flag keywords in a string

Re: Using FINDW in a DO loop to flag keywords in a string

Re: SAS macros in Git

Re: How to control the number of lines when rolling the mouse wheel on...

Re: Automatically calculate yearly average

Re: How to anonymize the Data (not masking)

Re: SAS RTDM: Custom Details Length

Re: SAS 9.4 Base and STAT for Windows under Parallels on a Mac with an...

Re: How to get table size using proc sql

Re: Using FINDW in a DO loop to flag keywords in a string

Re: How to control the number of lines when rolling the mouse wheel on...

Re: CALL SYMPUT vs CALL SYMPUTX

Re: Import metadata from qlikview to SAS

Re: Personal Login Manager not updating metadata

Re: SAS Environment Manager - Verificar versão

Re: Make year,month and day functions work on datetime

Re: Using FINDW in a DO loop to flag keywords in a string

Re: Can I put the BY-group value into the ExcelXP sheet name?

Re: Using FINDW in a DO loop to flag keywords in a string

Re: Problem with SAS kernel in Jupyter on Windows

Re: Problem with SAS kernel in Jupyter on Windows

Re: Problem with SAS kernel in Jupyter on Windows

Problem with SAS kernel in Jupyter on Windows

Re: No Kernel, WinError 2 for SAS Notebook in JupyterLab

Re: Discovering America's Most (and Least) Dangerous Jobs with SAS Uni...

Re: Indication treatment within the last two years

Re: RESULT SET processing

Re: How to Identity 2nd Last Record

Re: How to Identity 2nd Last Record

Re: How to Identity 2nd Last Record

Re: How to Identity 2nd Last Record

Re: SAS Tip: DATALINES statement overview

Re: Can't download SAS University Edition.

Re: Can't download SAS University Edition.

SUGA

SAS Analytics Explorers