Technology Tales

Finding human balance in an age of AI code generation

Recently, I was asked about how I felt about AI. Given that the other person was not an enthusiast, I picked on something that happened to me, not so long ago. It involved both Perplexity and Google Gemini when I was trying to debug something: both produced too much code. The experience almost inspired a LinkedIn post, only for some of the thinking to go online here for now. A spot of brainstorming using an LLM sounds like a useful exercise.

Going back to the original question, it happened during a meeting about potential freelance work. Thus, I tapped into experiences with code generators over several decades. The first one involved a metadata-driven tool that I developed; users reported that there was too much imperfect code to debug with the added complexity that dealing with clinical study data brings. That challenge resurfaced with another bespoke tool that someone else developed, and I opted to make things simpler: produce some boilerplate code and let users take things from there. Later, someone else again decided to have another go, seemingly with more success.

It is even more challenging when you are insufficiently familiar with the code that is being produced. That happened to me with shell scripting code from Google Gemini that was peppered with some Awk code. There was no alternative but to learn a bit more about the language from Tutorials Point and seek out an online book elsewhere. That did get me up to speed, and I will return to these when I am in need again.

Then, there was the time when I was trying to get a Julia script to deal with Google Drive needing permissions to be set. This started Google Gemini into adding more and more error checking code with try catch blocks. Since I did not have the issue at that point, I opted to halt and wait for its recurrence. When it did, I opted for a simpler approach, especially with the gdrive CLI tool starting up a web server for completing the process of reactivation. While there are times when shell scripting is better than Julia for these things, I added extra robustness and user-friendliness anyway.

During that second task, I was using VS Code with the GitHub Copilot plugin. There is a need to be careful, yet that can save time when it adds suggestions for you to include or reject. The latter may apply when it adds conditional logic that needs more checking, while simple code outputting useful text to the console can be approved. While that certainly is how I approach things for now, it brings up an increasingly relevant question for me.

How do we deal with all this code production? In an environment with myriads of unit tests and a great deal of automation, there may be more capacity for handling the output than mere human inspection and review, which can overwhelm the limitations of a human context window. A quick search revealed that there are automated tools for just this purpose, possibly with their own learning curves; otherwise, manual working could be a better option in some cases.

After all, we need to do our own thinking too. That was brought home to me during the Julia script editing. To come up with a solution, I had to step away from LLM output and think creatively to come up with something simpler. There was a tension between the two needs during the exercise, which highlighted how important it is to learn not to be distracted by all the new technology. Being an introvert in the first place, I need that solo space, only to have to step away from technology to get that when it was a refuge in the first place.

For anyone with a programming hobby, they have to limit all this input to avoid being overwhelmed; learning a programming language could involve stripping out AI extensions from a code editor, for instance, LLM output has its place, yet it has to be at a human scale too. That perhaps is the genius of a chat interface, and we now have Agentic AI too. It is as if the technology curve never slackens, at least not until the current boom ends, possibly when things break because they go too far beyond us. All this acceleration is fine until we need to catch up with what is happening.

Dealing with this Python error message on Windows: UnicodeEncodeError: ‘charmap’ codec can’t encode characters in position 56-57: character maps to <undefined>

Recently, I got caught out by the above message when summarising some text using Python and Open AI’s API while working within VS Code. There was no problem on Linux or macOS, but it was triggered on the Windows command line from within VS Code. Unlike the Julia or R REPL’s, everything in Python gets executed in the console like this:

& "C:/Program Files/Python313/python.exe" script.py

The Windows command line shell operated with cp1252 character encoding, and that was tripping up the code like the following:

with open("out.txt", "w") as file:
    file.write(new_text)

The cure was to specify the encoding of the output text as utf-8:

with open("out.txt", "w", encoding='utf-8') as file:
    file.write(new_text)

After that, all was well and text was written to a file like in the other operating systems. One other thing to note is that the use of backslashes in file paths is another gotcha. Adding an r before the quotes gets around this to escape the contents, like using double backslashes. Using forward slashes is another option.

with open(r"c:\temp\out.txt", "w", encoding='utf-8') as file:
    file.write(new_text)

Getting custom Python imports to work in Visual Studio Code

While I continue to use Spyder as my preferred Python code editor, I also tried out Visual Studio Code. Handily, this Integrated Development Environment also has facilities for working with R and Julia code as well as MarkDown text editing and adding the required extensions is enough for these applications; it helps that there is an unofficial Grammarly extension for content creation.

My Python code development makes use of the Pylance extension and it works a little differently from Spyder when it comes to including files using import statements. Spyder will look into the folder where the base script is located but the default behaviour of Pylance is that it looks in the root path of your workspace. This meant that any code that ran successfully in Spyder failed in Visual Studio Code.

The way around this was to add the required location using the python.analysis.extraPaths setting for the workspace. That meant opening Settings by navigating to File > Preferences > Settings in the menu system and entering python.analysis.extraPaths into the search box. That took me to the section that I needed and I then clicked on Add Item before entering the required path and clicking on the OK button. That was enough to fix the problem and all worked as it should after that.

Using multi-line commenting in Perl to inactivate blocks of code during testing

Recently, I needed to inactivate blocks of code in a Perl script while doing some testing. Since this is something that I often do in other computing languages, I sought the same in Perl. To accomplish that, I need to use the POD methodology. This meant enclosing the code as follows.

=start

<< Code to be inactivated by inclusion in a comment >>

=cut

While the =start line could use any word after the equality sign, it seems that =cut is required to close the multi-line comment. If this was actual programming documentation, then the comment block should include some meaningful text for use with perldoc. However, that was not a concern here because the commenting statements would be removed afterwards anyway. It also is good practice not to leave commented code in a production script or program to avoid any later confusion.

In my case, this facility allowed me to isolate the code that I had to alter and test before putting everything back as needed. It also saved time since I did not need to individually comment out every executable line because multiple lines could be inactivated at a time.

SAS Data Step Hash Objects and Memory

Using hash objects in SAS data step code offers some great advantages from the speed point of view; having a set of data in memory rather than on disk makes things much faster. However, that means that you need to keep more of an eye on the amount of memory that’s being used. The first thing is to work out how much memory is available, and it’s not necessarily the total amount installed on the system or, for that matter, the amount of memory per processor on a multiprocessor system. What you really need is the number, in bytes, that is stored in the XMRLMEM system option and here’s a piece of code that’ll do just that:

data _null_;
mem=getoption('xmrlmem');
put mem;
run;

The XMRLMEM is itself an option that you can only declare in the system call that starts SAS up in the first place, and there are advantages to keeping it under control, particularly on large multi-user servers. However, if your hash objects start to exceed what is available, here’s the sort of thing that you can expect to see:

ERROR: Hash object added 49136 items when memory failure occurred.
FATAL: Insufficient memory to execute data step program. Aborted during the EXECUTION phase.
NOTE: The SAS System stopped processing this step because of insufficient memory.
NOTE: SAS set option OBS=0 and will continue to check statements. This may cause NOTE: No observations in data set.

Those messages are a cue for you to learn to keep those hash objects and to only ever make them as large as your memory settings will allow. Another thing to note is that hash objects are best retained for rather fixed data volumes instead of ones that could outgrow their limits. There’s a certain amount of common sense in operation here, but it may be that promoters of hash objects don’t mention their limitations as much as they should. If you want to find out more, SAS has a useful paper on their website and their Knowledge Base has more on the error messages that you can get.

Selecting SAS code in the Program Editor on UNIX

Here’s a possible bugbear with programming using the SAS Display Manager in UNIX, selecting sections of code and running them. In the installations that I have encountered, the mouse selection is not retained, so the code selection cannot be run. There is a fix for this that is not the most obvious. Going to the Preferences dialogue box (Tools > Options > Preferences… from the menu bar) and selecting the Editing tab brings up the screen below:

Ensuring that “Automatically store selection” is switched off, as shown above, will allow one to select and submit sections of code from a SAS program like what is normal practice with Windows SAS. Though it isn’t an obvious solution, it does the trick for me.

Another way to look through WordPress code

At times in the past, I have complained about there not being sufficient documentation of the actual code used in WordPress itself. True, there is the venerable Codex but that can be incomplete and any extra help is always useful. Peter Westwood (a.k.a westi) has generated documentation from the code itself and put it up on the web for all to peruse. While we are on the subject of information on the more advanced aspects of WordPress, blogs by Mark Jaquith and Andrew Ozz are also worth a look. As the cliché goes, every little helps…

Some SAS Macro code for detecting the presence or absence of a variable in a dataset

Recently, I needed to put in place some code to detect the presence or absence of a variable in a dataset, and I chose SAS Macro programming as the way to do what I wanted. The logic was based on a SAS sample that achieved the same result in a data step and some code that I had for detecting the presence or absence of a dataset. Mixing the two together gave me something like the following:

%macro testvar(ds=,var=);

%let dsid=%sysfunc(open(&ds,in));
%let varexist=%sysfunc(varnum(&dsid,&var));
%if &dsid > 0 %then %let rc=%sysfunc(close(&dsid));

%if &varexist gt 0 %then %put Info: Variable &var is in the &ds dataset;
%else %put Info: Variable &var is not in the &ds dataset;

%mend testvar;

%testvar(ds=dataset,var=var);

What this does is open up a dataset and look for the variable number in the dataset. In datasets, variables are numbered from left to right with 1 for the first one, 2 for the second and so on. If the variable is not in the dataset, the result is 0 so you know that it is not there. All of this is what the VARNUM SCL function within the SYSFUNC macro function does. In the example, this resolves to %sysfunc(varnum(&dsid,var)) with no quotes around the variable name, akin to what you would do in data step programming. Once you have the variable number or 0, then you can put in place some conditional logic that makes use of the information, like what you see in the above simple example. Of course, that would be expanded to something more useful in real life, but I hope it helps to show you the possibilities here.

Getting Pylance to recognise locally installed packages in VSCode running on Linux Mint

When using VSCode on Linux Mint, I noticed that it was not finding any Python package installed into my user area, as is my usual way of working. Thus, it was being highlighted as being missing when it was already there.

The solution was to navigate to File > Preferences > Settings and click on the Open Settings (JSON) icon in the top right-hand corner of the app to add something like the following snippet in there.

"python.analysis.extraPaths": [
"/home/[user]/.local/bin"
]

Once you have added your user account to the above, saving the file is the next step. Then, VSCode needs a restart to pick up the new location. After that, the false positives get eliminated.

HennessyBlog theme update

Over the weekend, I have been updating the theme on my other blog, HennessyBlog. It has been a task that projected me onto a learning curve with the WordPress 2.1 codebase. Thus, I have collected what I encountered, so I know that it’s out there on the web for you (and I) to use and peruse. It took some digging to get to know some of what you find below. Since any function used to power WordPress takes some finding, I need to find one place on the web where the code for WordPress is more fully documented. The sites presenting tutorials on how to use WordPress are more often than not geared towards non-techies rather than code cutters like myself. Then again, they might be waiting for someone to do it for them…

The changes made are as follows:

Tweaks to the interface

These are subtle, with the addition of navigation controls to the sidebar and the change in location of the post metadata being the most obvious enhancements. “Decoration” with solid and dashed lines (using CSS border attributes rather than the deprecated hr tagset) and standards compliance links.

Standards compliance

Adding standards compliance links does mean that you’d better check that all is in order; it was then that I discovered that there was work to be done. There is an issue with the WordPress wpautop function (it lives in the formatting.php file) in that it sometimes doesn’t add closing tags. Finding out that it was this function that is implicated took a trip to the WordPress.org website; while a good rummage in the wp-includes folder does a lot, it can’t achieve everything.

Like many things in the WordPress code, the wpautop function isn’t half buried. The the_content function (see template-functions-post.php) used to output blog entries calls the get_content function (also in template-functions-post.php) to extract the data from MySQL. The add_filter function (in plugin.php) associates the wpautop function and others with the get_the_content function to add the p tags to the output.

To return to the non-ideal behaviour that caused me to start out on the above quest, an example is where you have an img tag enclosed by div tags. The required substitution involves the use of regular expressions that work most of the time but get confused here. So adding a hack to the wpautop function was needed to change the code so that the p end tag got inserted. I’ll be keeping an eye out for any more scenarios like this that slip through the net and for any side effects. Otherwise, compliance is just making sure that all those img tags have their alt attributes completed.

Tweaks to navigation code

Most of my time has been spent on tweaking of the PHP code supporting the navigation. Because different functions were being called in different places, I wanted to harmonise things. To accomplish this, I created new functions in the functions.php for my theme and needed to resolve a number of issues along the way. Not least among these were regular expressions used for subsetting with the preg_match function; these were not Perl-compliant to my eyes, as would be implied by the choice of function. Now that I have found that PCRE’s in PHP use a more pragmatic syntax, there appeared to be issues with the expressions that were being used. These seemed to behave OK in their native environment but fell out of favour within the environs of my theme. Being acquainted with Perl, I went for a more familiar expression style and the issue has been resolved.

Along the way, I broke the RSS feed. This was on my off-line test blog so no one, apart from myself, that is, would have noticed. After a bit of searching, I realised that some stray white-space from the end of a PHP file (wp-config.php being a favourite culprit), after the PHP end tag in the script file as it happens, was finding its way into the feed and causing things to fall over. Feed readers don’t take too kindly to the idea of the XML declaration not making an appearance on the first line of the file. Some confusion was caused by the refusal of Firefox to refresh things as it should before I realised that a forced refresh of the feed display was needed. Sometimes, it takes a while for an addled brain to think of these kinds of things.