Originally posted on the fediverse. Doing some PESOD.

so i was looking through the codebase of a rather widely-used C++ FOSS project. one that's known for its simplicity and minimalism, in fact!

of the things:

• the button to change the sort mode gets handled with 2 layers of string-based indirection, the first to map events from the UI, the second because the application state object only communicates through a "setProperty" method. there's no hashmap or anything internally, the method directly accesses properties. but i guess string constants are more "dynamic" than object.setSortMode
  • i should note that all of these are user-created objects. they decided to do this.
  • the second layer of indirection is through a #define constant, btw. because we want our indirection system to be kinda checked by the compiler.
  • the ui code directly contains the constants to set the sort order. i don't get why the extra layer of indirection is necessary at all.
• every single class that isn't part of an implementation of something has the class/class::Implementation pattern. every single one. including classes with no virtual methods, so what's the point. the class that stores the sort mode is patterned like you would do DI on it, but it's so tightly coupled to the UI that that would be impossible. what's the point.
  • this includes other silly things like the main window class!
• there's this somewhat silly pattern where you have a class Foo with property Foo foo_ and method initMyFoo(Foo& foo). i don't get why the method doesn't just operate on the class itself, it would be basically impossible for it to be used on anything other than this->foo_.

anyways, im not gonna name the actual software project, because i do think it's good and i dont want to publicly shit on FOSS devs unprompted. i just hate how aggressive the misapplication of patterns that could be useful is.

like, string indirection is handy! DI is handy! there are plenty of cases where "initialize this thing based on the state of this other thing" is necessary! but the overuse of these patterns where it's not necessary and likely never will be leads to an ostensibly simple program being utterly labyrinthine to get a foothold in. this is a simple program, that basically just runs a library and renders a list from it, whose code is written like it will ever do anything more than render that library's data. it's designed for "loose coupling" of things that are just never going to not be directly coupled.

it's not an underperformant application, it's not a memoryhog, it's just written to preempt a fundamental change in design and purpose that has yet to come in the last 2 decades of the software's existence. genuinely, what is the point?

so, a friend of a friend wanted a way to aggregate their posts from various sources onto their Neocities site, with no Javascript. as it turns out, there are no good services with which to do this, and little in the way of open source projects. parsing RSS/Atom isn't hard, so I threw together a PHP script that aggregates some N feeds, caches them, interleaves them, and renders them. it even lets you filter them!

it's pretty minimal overall and a large portion of the file is just the config section at the top. a running example is up on my website, with some styling to pretty it up.

I'll be hosting an install of it for said friend of a friend until they get a VPS or something, but, like a lot of the other stuff I make, this thing is in the public domain. it, and the style in-use on my homepage, are available on my cgit.

If you're looking for a similarly minimal guestbook script, I wrote one called visitors_dot_php a bit ago. I think I forgot to mention it here lol.

I'm gonna make a more proper update post sometime this month, but I figured I'd post something about this.

I've been ~aleteoryx, g'night.

This is just a delightful set of cursed programming articles. Each is, on the one hand, a shitpost reminiscent of Tom7 in terms of elaborateness, and at the same time an entertaining narrative in the strange world the author has created, where magic and programming are one and the same.

Though there is little continuity, they are listed below in published order.

Worth the read.

• Reversing the technical interview
• Hexing the technical interview
• Typing the technical interview
• Rewriting the Technical Interview
• Unifying the Technical Interview

Enjoy!

i love bash.

i really, really, love bash. well, it and all the other posix shells. i personally use zsh in my terminal, with oh-my-zsh stacked on top. for the rest of this article, i will be using bash to refer generally to the bashlike shells(so no fish).

bash is so excellent at specific types of problems, and completely god awful at almost every other kind.

bash does not do math well. bash can do integer arithmetic pretty well, but the second you need to deal with decimals bash is your mortal enemy, forcing you to invoke bc or some other scripting environment.

bash does not do parsing well! bash does simple parsing effortlessly assuming you're comfy with regexes, and most formats have some way to handle them from the command line (jq, yq, sqlite, etc.), but bash will be sure to make you unhappy the entire time for just about anything else and you will have to learn what the IFS is.

what bash excels at, though, is anything you can do from the command line. bash has quite literally effortless process management, it's the main thing you do. bash has easy file and directory handling, and pattern matching tools to make simple file searches trivial. if you live in the command line, every utility you install is at bash's disposal. on top of the wealth of standard posix tools almost every command line utility is built to serve bash's whims.

the language of a terminal is one of text. when you type a command, whatever it outputs doesn't really feel like reading or writing bytes to you, the user. it's just text appearing. text moving. text changing color. part of the brilliance of bash's UI design is that capturing the output of a command--as text--is, like so much else, trivial. i do not need to think about the true abstractions of the operating system to understand that $(date +%B)-* will, at time of writing, get me every file beginning with "February-". the brilliance of bash is that it lets the programmer think in terms of the established user interface, and not in terms of the underlying systems. the fundamental abstractions of bash bear no relation to posix or unix or linux, instead just to the posix command line.

bash is, for the experienced user, so effortless to write, because it's not really a deviation from the normal use of the command line, it's so intuitive because you are essentially just telling the computer to do what you would do. there is not, i don't think, any similar form of ui-language link for gui-based applications, nor for web-based applications. the blurred line between programming environment and interface, made possible by virtue of them both being text, gives bash a certain elegance that few other repls aspire to, much less manage.

this is, i think, the thing powershell fails at. in its attempt to be a safer, more structured bash, it loses much of the effortlessness that makes bash what it is. it is clunkier and slower, though technically more featureful. as a tool for remote windows administration, i think it succeeds fine, but i could not live in powershell the way i live in my terminal on linux. i do understand that that isn't the point, and even when windows was text-based, the scripting environment was still in desperate need of features, seen more as a tool than an interface, but i still figured it was worth mentioning.

these days, when scripting, i find myself bouncing between bash and python, depending on the task. they're both excellent at completing whatever i throw at them, and i'm happy to have such loyal and powerful tools as those, and such a wonderful interface for my day-to-day, as bash.

that's all.

I built a webcomic searching tool! It's a simple booru-style positive/negative tag search, and the actual catalog doesn't contain that many comics right now, but that's besides the point.

The actually interesting thing is the context, as this was launched as a part of Alexandra's Marigold Town! It's half webhost, half roleplay, with the idea being that each site is a location in Marigold Town. Certainly not a new idea, but anything making the internet a bit of a weirder place is good in my book. The town's mayor is an owl also? Rad. If you've got the sitebuilding bug and some free time you should totally check it out! Also if you want to just dick around online!! There's an apothecary with docs on 19th century and before medicines! A coffee shop with cute little pixel arts to download!! Lots of other cool stuffs!!!

My contribution was Marigold Town Comics, Games, and More. Creative name, I know. Right now it's just a JavaScript toggle for future ventures, and the aforementioned index.

As is starting to become customary, here's a brief technical breakdown:

The actual comics are entered into a big long YAML file. Unfortunately there's no strictYAML port for PHP(I'll probably end up writing one lol), but I know the quirks of it enough to stay safe. I prefer YAML for its general readability, without the verbosity of something like JSON. On pageload, the catalog checks the YAML file's modtime against the modtime of a Sqlite DB file. If the DB is older, it gets regenerated from the YAML file. This is done just because, as more comics get loaded in, having an index saved to disk will be significantly faster than YAML parse -> Hashmap gen -> Hashmap search. Plus it makes the code nicer.

This project contains my most advanced SQL work to-date. The queries are a bit bulky, but I figure keeping everything as 2 or 3 queries will be faster than constantly making Sqlite library calls from PHP. It also makes the code more concise, which is a plus in my book. I did take care to document their workings for future me, so that they're not too unweildly.

That's the only particularly complicated part, from there it's just standard webdev stuff. That help menu is by-far the sexiest CSS I've ever put into prod though.

Short-term Future Plans

• I need to sort out a method for getting user contributions, like adding tags or whole webcomics. This should either be via e-mail, a form, or both.
• Start adding the networks I'm aware of.
  • SpiderForest
  • Xepher
• Continue to add various comics from across the 'net!
• Attempt to put up an archive of gopus.xepher.net

Long-term Future Plans

• This site will eventually evolve into a more complicated web experience! There's gonna be characters and a dialog system and probably some other stuff, too!!
• Spoilers!

Alright! I'm actually pretty happy with where my site is now. It's not gonna support mobile anytime soon, and I'm fine with that. It's got an aesthetic^tm.

Site areas are basically done!

I've finally gotten together a links page proper, and im super happy with it so far. I've tried to make sure the page loads as fast as possible on everything, and some testing proves this, as it's still getting around 20s to full load in dial-up simulators(significantly faster to become navigable). It is, by far, the heaviest part of the site, simply due to the sheer number of people's little buttons I've included. All of this to say: I'm doing my part!.

This isn't the only area I've recently finished, and I think I'm basically done with the core site layout for now. Everything else will end up accessible through the downloads page, explore page, or homepage, although the PHP that renders the header has room for new links just-in-case.

A Discovery:

I've also discovered that the platform, in its infinite wisdom, includes a nojs way to snap scrolling, and am now planning to try doing mandatory monospaced scrolling. I would absolutely love to get the screen to, at least mostly, snap to individual rows of text. I need to see how it plays with screen readers is the main issue. I'm pretty sure everything works fine now, but I have yet to do particularly extensive testing. To anyone reading: has the MacOS screen reader changed all that much in the last 8 years? The only Apple device I have to test with is slowly approaching middle school, and I want to be sure it's accurate for testing purposes.

Final things:

Added a www.aleteoryx.me redirect subdomain, just in case, and a noenc.aleteoryx.me subdomain, which serves the main site with no HTTPS redirection. I still need to fix the odd embed or iframe that breaks security policy, but that's a problem for later.

Plans:

I'd like to add a guides.aleteoryx.me, where I can put out help for all sorts of things. This would probably end up being similar to LandChad but with a far broader scope of topics. I have an image in my head of what this will end up visually looking like, and I'm excited to start work on it.

Might play around with some sort of virtual tenancy/shared unix setup? Need to harden the server more, but mod_userdir should handle the bulk of the site hosting for me. It would probably just be for me, my partners, and some friends.

More work on personal 88x31 buttons! I've already obsoleted my current one after subtly changing the link colors for accessibility reasons! I'm probably going to play with a design like wings.nu's button design.

Find a way to reintegrate eyes to the new site! I simply must return to my bullshit, and said bullshit is being a scopophobe's death sentence!!!