Conkey VS https-bot

Most of my programs were written for my own use, including:

• A keyboard layout to type numerous non-English letters, punctuation marks and mathematical symbols, originally for Windows but subsequently ported to Linux and Mac [https://github.com/bradrn/Conkey]

• A ‘sound change applier’ for my hobby of language construction, to simulate the process of historical sound change [https://bradrn.com/brassica/]

• A small browser extension to save the full text of all webpages I visit, and a local client to search the database [not open-sourced, apologies!]

The first two have gained a few other users since being released, but I’m pretty sure I’m still the one who uses them the most!

I made my own crossplatform multilingual layout [0]. Although it’s based on QWERTY, it shouldn’t be hard to remap the Linux and Mac versions to any other base layout, since they’re autogenerated from the Windows version.

[0] https://github.com/bradrn/Conkey

The biggest one for me is undoubtedly my custom keyboard layout Conkey [0], which I use constantly (including for typing this very comment). I hate the way the base US layout tends to get distorted in other keyboard layouts with good support for non-ASCII characters, so Conkey had the explicit goal of retaining that basic unshifted layout. I’ve also ended up porting Conkey to Mac and Linux — and given that I’m slowly switching from Windows to Linux, at least the Linux ports have ‘scratched my own itch’ too, which is nice.

Also, I made a utility to archive the full text of every website I view and store it in a SQLite database for searching. It’s proven pretty useful when I want to find something I saw a while ago and then forgot. (I haven’t attempted to open-source it, though — it consists of three entirely separate components, two of which were a pain to set up. I must try to get it into a more usable state one of these days.)

What else… my sound change applier [1], perhaps? Not that I use it very much, because I only need it on those occasions when I want to do some conlanging, which I haven’t had much time for recently. Actually, sound change appliers strike me as being very much a ‘scratch own itch’ type of project in general… sometimes it feels like every conlanger has written their own, and no two can agree on a nice design. Everyone just has their own unique preferred way of doing things.

[0] https://github.com/bradrn/Conkey

[1] https://github.com/bradrn/brassica

For a ~50-year-old program, Emacs’s support for multilingual input — and really, it’s all-round flexibility — continually amazes me! For myself I prefer my own custom keyboard layout [0], because it works outside Emacs too, but I’d happily use Emacs’s own input methods if that would be sufficient.

(In fairness, I have found one weak spot, namely font support… I’ve used ‘unicode-fonts’ [1] with some success, but reportedly it doesn’t work with the latest Emacs. Ah well, it’s at least fairly rare that this becomes a problem in practice.)

[0] https://github.com/bradrn/Conkey

[1] https://github.com/rolandwalker/unicode-fonts

To port my keyboard layout [0] to OSX, I used ‘osxkb’ [1], which outputs an OSX keyboard layout bundle given a simple textual specification file. It was originally created specifically to port Conkey to OSX, but should be entirely usable for other purposes as well.

[0] https://github.com/bradrn/Conkey

> But instead, we're dealing with Latex - a language that overcomplicates the most basic features such as fonts, tables and special characters.

I can’t really argue with the rest of your post, but in my experience this is incorrect. Fonts and special characters are both trivial if you use XeTeX, and tables, though slightly clumsy, are still pretty easy. As an example, see the documentation I wrote for https://github.com/bradrn/Conkey, which makes extremely heavy use of all three features. (As documentation for a keyboard layout, it uses characters from pretty much every corner of Unicode, and accompanying tables of many shapes and sizes to show how to type these characters; I needed to use Gentium in order to render all these characters, with Times New Roman as a fallback. I found that LaTeX could ably handle all of these complecations.)

It's inspired by this comment I made: https://news.ycombinator.com/item?id=26599746.

I actually saw several comments with HTTP URL posted, and that was the only one I bothered to comment on. So I thought that this is something better suited for bots than human.

I hacked this together over yesterday and today: https://github.com/fishy/https-bot.

Basically it uses the Firebase API (https://github.com/HackerNews/API) to find comments with HTTP URLs in them, try the HTTPS version, compare the contents, post back a comment if the contents are more than 95% similar.

The "95% similar" part was actually the first part I wrote in the code. At first I tried a few existing go packages implementing diff/lcs, but most of them was quite slow and does a lot of allocations when I'm comparing two randomly generated 10KiB blobs, so I wrote my own (https://pkg.go.dev/github.com/fishy/https-bot/similarity), which is optimized for space (it does almost no allocations), and it's also faster because allocations are slow. (I know this is an unfair comparison that most of the existing implementations need to give you an output that can be used to reconstruct the two blobs back, so at least some of their allocations are required and unavoidable)

I also wrote a bug that it would find the same HTTP url in every run and post the same comment over and over again. My apologize to dang or whoever dealt with it (or maybe the system is good enough that it blocked those repetitive comments automatically).

In the end it successfully made 6 comments across ~4 hours (not including the repetitive ones). All of those comments are flagged (likely due to hn policy), https://news.ycombinator.com/item?id=26692588 is the only one that's still visible to other users at the time of writing, if you are curious.

I just killed it completely from the request of dang. Although it only lived for a few hours, it's still a fun exercise. Maybe I'll convert it into a reddit bot next? Who knows.

Thank you for the feedback. I added FAQ to the README and answered this question: https://github.com/fishy/https-bot/blob/main/README.md#this-...