Quentin Tracker

Re: SAS Macros: Error Handling Best Practices

Quentin — Fri, 18 Sep 2026 20:50:54 GMT

We’re so excited you’re coming to BASUG to present in person at our annual SAS Blowout on Oct 30! We’re just putting the finishing touches on the official announcement. Registration opens Tues Sept 22. www.basug.org/events.

Join us to see presentations by @CarleighJoC, Charu Shankar and Jane Eslinger.

Re: How to get 2-D heatmap/histogram?!

Quentin — Sat, 05 Sep 2026 20:59:29 GMT

https://www.google.com/search?q=sas+heatmap

Re: Using PROC COMPARE with Case Insensitivity

Quentin — Thu, 03 Sep 2026 19:23:27 GMT

Very cool @Ksharp . One of the things I love most about SAS is that I've been using it for (gasp) 30 years and yet there are still parts of even Base SAS that I've never used!

Re: Using PROC COMPARE with Case Insensitivity

Quentin — Wed, 02 Sep 2026 17:32:46 GMT

Great post. PROC COMPARE is one of my favorite procs! Only suggestion for a small improvement to consider would be to consider creating the second dataset where everything is upcased as view, instead of a new dataset. If the data is big, sometimes views can be more efficient as they don't require storage space.

Also in a post like this where a PROC COMPARE is the point, introducing the idea of views could be a distraction for readers unfamiliar with the concept.

Re: How to measure the concaveness ?!

Quentin — Sat, 29 Aug 2026 10:53:13 GMT

Please share your code, and mark it as the accepted answer.

Re: How to measure the concaveness ?!

Quentin — Fri, 28 Aug 2026 16:08:50 GMT

Before asking for a macro, you need to define how you want to measure "overall concaveness". You can take the second derivative of a function to measure the local curvature, integrate over a range to get the average concavity... If you're looking for general approaches to the question, I would google / ask AI. It's not a macro question until you have code that you want to generate.

Re: Why I believe coding education must evolve for the age of AI

Quentin — Thu, 27 Aug 2026 14:11:08 GMT

Great article. My only quibble, I would have an even stronger title. I believe AI will revolutionize coding education (and also coding itself, and also education itself).

One small example (from a 30 year SAS user). For years I've wanted to learn python, but haven't really wanted to put in the time, and haven't found business cases where I'll actually build with Python. This summer, my company has a student intern, who's using Python to build some ML models for a pilot project.

So now I'm reviewing their Python code (in VS code) and I ask github copilot about python's scoping/namespace rules. And github copilot basically says to me "Because you're a SAS programmer [which it remembered from previous conversations] you're thinking of SAS macro language scoping rules which work like X. But Python scoping works like Y." And proceeded to explain Python to me using comparisons to similar constructs in SAS. It took by breath away.

A second example, I've played with building tests suites to cover my utility macros. But I'm not a "real" programmer, unit testing is not something that was taught in my psychology undergrad classes. But with Copilot, I can write a macro myself, and then ask copilot to create a test suite, and then ask copilot to run the test suite and tell me if it passes. So this is closer to vibe coding the test coverage, but I'm reading the code, to learn about how you can develop a test suite in SAS. Pretty amazing.

I've been slow to adopt AI myself. In the past couple months I've been doing AI-assisted coding with SAS and Python (VS code extension for SAS), and full vibe coding of a couple personal iPhone apps at home.

Bill Gates is right. AI (and robots) are going to be a true revolution to not only computing, but also society as a whole.

Re: Graphics ODS Destination

Quentin — Mon, 24 Aug 2026 20:54:12 GMT

Agree for ease of use for students (and me : ), I could imagine an ODS destination like:

ODS GRAPHICSONLY imagefmt=png imagename="MyGraph" gpath='~';

Or just adding a gpath option to the ODS GRAPHICS statement:

ods graphics on / imagefmt=png imagename="MyGraph" gpath='~' ;

On reflection, it's weird to me that there is no gpath option for the ODS GRAPHICS statement. It feels like that is a missing option that would be nice.

I rarely try to make stand-alone image files. But when I do, I often end up doing the same dance I did today, cuz I don't remember the ODS listing trick.

Ooh, I have an idea @Statsbob :

Since your students are using EG and the SAS Report destination, you could configure EG to set gpath to their home directory. If I do this:

Then you could write graphs to your home directory with:

ods graphics on / imagefmt=png imagename="MyGraph";
proc sgplot data=sashelp.class;
    scatter x=height y=weight;
run;

This isn't the easy flexibility you want, but it could be helpful for students who just want to see the image files they create.

Re: Graphics ODS Destination

Quentin — Mon, 24 Aug 2026 19:29:17 GMT

Yes, it's one of the macro vars EG makes. I tend not to use them. I would use "%sysfunc(pathname(work))" for portability. But students might find &sasworklocation easier.

Re: Graphics ODS Destination

Quentin — Mon, 24 Aug 2026 18:59:36 GMT

Thanks @PaigeMiller, you are correct. Works on linux too. I was blindly following the log, which doesn't write a note about creating the .png file. But when I looked in the directory, there were about 50 png files waiting form me from all my "failed" attempts.

You can do stuff like:

ods HTML path=&sasworklocation gpath='~';
proc sgplot data=sashelp.class;
    scatter x=height y=weight;
run;
proc sgplot data=sashelp.class;
    vbar age/response=height;
run;
ods html close;

Re: Graphics ODS Destination

Quentin — Mon, 24 Aug 2026 18:40:59 GMT

I found this blog post by @ChrisHemedinger : https://blogs.sas.com/content/sasdummy/2015/11/23/ods-statement-options-to-change/

For a minute I thought again that I could do it with:

ODS HTML5 (ID=EGHTML) gpath="~";
ods graphics on / imagefmt=png imagename="MyGraph" ;

proc sgplot data=sashelp.class;
    scatter x=height y=weight;
run;

But it looks like you are not able to change the value for gpath after opening a destination, as he notes in the blog for the path option. Maybe Chris will have another idea.

I also tried forcing the issue by closing and reopening just the HTML5 destination, but still couldn't make it work:

ODS _ALL_ CLOSE;

ODS HTML5 FILE="%sysfunc(pathname(work))/foo.htm" GPATH="~" ;
ods graphics on / imagefmt=png imagename="MyGraph" ;

proc sgplot data=sashelp.class;
    scatter x=height y=weight;
run;

ods _all_ close ;

which I find confusing.

Re: Graphics ODS Destination

Quentin — Mon, 24 Aug 2026 17:53:33 GMT

How odd. I tried over-riding GPATH for the ODS HTML5 statement generated by EG, but had no luck. Maybe I'm doing something dumb? Seems like it should be easy.

7         FILENAME EGHTML TEMP;
18         ODS HTML5(ID=EGHTML) FILE=EGHTML
19             OPTIONS(BITMAP_MODE='INLINE')
20             %HTML5AccessibleGraphSupported
21             ENCODING='utf-8'
22             STYLE=HTMLBlue
23             NOGTITLE
24             NOGFOOTNOTE
25             GPATH=&sasworklocation
26         ;
NOTE: Writing HTML5(EGHTML) Body file: EGHTML
27         
28         *add gpath ;
29         ods HTML5(ID=EGHTML) gpath="~";
30         
31         ods graphics on / imagefmt=png imagename="MyGraph" ;
32         
33         proc sgplot data=sashelp.class;
34             scatter x=height y=weight;
35         run;

NOTE: PROCEDURE SGPLOT used (Total process time):
      real time           0.09 seconds
      cpu time            0.04 seconds
      
NOTE: There were 19 observations read from the data set SASHELP.CLASS.

Code runs without errors, but no .png file created.

Like you said, if I add ODS listing statement, it works fine:

29         ods listing gpath="~";
30         
31         ods graphics on / imagefmt=png imagename="MyGraph" ;
32         
33         proc sgplot data=sashelp.class;
34             scatter x=height y=weight;
35         run;

NOTE: PROCEDURE SGPLOT used (Total process time):
      real time           0.13 seconds
      cpu time            0.06 seconds
      
NOTE: Listing image output written to /home/Quentin/MyGraph9.png.
NOTE: There were 19 observations read from the data set SASHELP.CLASS.

Re: Graphics ODS Destination

Quentin — Mon, 24 Aug 2026 14:10:10 GMT

That's interesting. I think most of the modern SAS IDEs (e.g. EG, studio, etc) pretty much assume that you have one destination turned on (typically HTML). So I don't think a user would have to turn on the listing destination just to get a png file or whatever. Wouldn't they just code something like:

ods html gpath="Q:\junk";
ods graphics on / imagefmt=png imagename="MyGraph" ;

proc sgplot data=sashelp.class;
    scatter x=height y=weight;
run;

But I can see how to a student, having ODS graphics destination (with options for PNG/JPG/SVG etc) might be a more obvious implementation.

With the current code, I think the easier change might be to add an imagepath option to the ODS GRAPHICS statement.

The good thing is with AI code assistants, students should be much faster at figuring how to do something in SAS. Much faster than reading through paper manuals or posting on SAS-L, back in my day, walking up a hill in the snow both ways to get to school...

Re: Macro-Dash - a SAS Platformer Game - https://dash.sasjs.io

Quentin — Mon, 24 Aug 2026 13:29:21 GMT

Thanks Allan, love it.

This shows how much younger you are than me. You think of Super Mario Brothers, I would think of doing Donkey Kong or Lode Runner.

Maybe you should post an index to all the SAS games available. I know you've made a few. I remember @ChrisHemedinger created a SAS version of Wordle, and in the comments of his blog post https://blogs.sas.com/content/sasdummy/2022/03/17/wordle-sas/ mentioned that someone had created a SAS version of Castle Wolfenstein! I wonder if anyone ever tried to build a SAS chess engine...

Re: Scenario: Police Raid Home of Innocent Person

Quentin — Mon, 24 Aug 2026 13:13:26 GMT

TBH, this is probably unrealistic but I'm not sure I want police using AI for criminal investigations at all. I'm a data person, and love AI/ML, but the more I learn about Flock cameras etc the idea of predictive policing is pretty scary to me.

I won't put my tinfoil hat all the way on, but for anyone interested, Flock cameras are automated license plate recognition (ALPR) cameras, leased by cities and states from a private company. Some home owners' associations and private companies lease them as well. The cameras take an image of every car that passes by, use AI to read the license plate as well as bumper stickers and other identifying characteristics, and store the license plate, image, and other metadata in a national database searchable by local, state, and federal law enforcement, including ICE. No warrant required for a search. It's literal big brother government surveillance, storing data on the movements of every car in America.

If you want to learn more, see deflock.org.

Sorry for speaking too close to politics for a SAS forum, but I think it's on-topic for this sort of ethical AI question. I'm not sure any police use of AI is ethical. Certainly using AI for mass surveillance is not ethical.

Re: Claude Skills Explained

Quentin — Fri, 21 Aug 2026 14:48:45 GMT

Thanks for this! I was just on LinkedIn this week, asking a SAS friend basically "what's a skill, how do I use it?" This is very timely introduction for me.

Re: Strange behavior when data are written and re-read directly from an Excel file

Quentin — Thu, 20 Aug 2026 14:44:27 GMT

Thanks @Chevell_sas I knew you'd come through for me. : )

But still, it feels somewhat bug-ish to me. Without adding the tagattr (which is of course always a good thing to do when writing to Excel), PROC REPORT or ODS EXCEL destination is writing a numeric value to the xml file with lots of trailing blanks. Excel has no problem reading the file and showing the data.

Then the same file that works for Excel, is mis-read by the XLSX engine, because the XLSX engine is confused somehow by the Excel / XML file that has a bunch of trailing blanks in the numeric field.

So in my mind, likely it's a bug in the XLSX engine which can't properly read data from a numeric field which 200 trailing blanks, even though it is allowed by Excel.

Or could be a bug in PROC REPORT or ODS EXCEL, if it is not allowed by the XLSX/openXML spec to create this kind of kind of numeric field with 200 trailing blanks.

Re: PhilaSUG & NJSUG Joint In-Person Meeting 2026

Quentin — Thu, 20 Aug 2026 13:46:24 GMT

I'm sure it will be a terrific meeting. Wonderful to see PhilaSUG is back, and NJSUG has grown so quickly! I'm sure this will also help spread the word that SEASUG will be just outside Philadelphia this year! https://sesug.org/SEASUG2026/index.php Great things happening in the SAS Philly world!

Re: SAS Support Communities metrics and their trends

Quentin — Tue, 18 Aug 2026 16:59:57 GMT

That's interesting, I didn't realize SO had come around after being so anti-AI. I'm surprised they even decided to make an MCP server to provide an AI interface. I would have assumed OpenAI, Anthropic et al had already hoovered up everything from stackoverflow and used it in their training. I guess an MCP server does make it easy for a company to make a chatbot that looks only at SO, or only at approved tags from SO, or whatever.

Re: SAS Support Communities metrics and their trends

Quentin — Mon, 17 Aug 2026 12:59:18 GMT

Just for fun, I looked at the Stack Overflow SAS volume over time:

https://data.stackexchange.com/stackoverflow/query/1956163