Adding custom noise pattern support basically caused a 7x increase in cloud generation times as the algorithm produces cooler clouds but is more computationally expensive.

So I added an LRU cache that stores the sampler result to speed up cloud generation.

I tried to use fancier methods such as remembering previously computed points but I wasn’t able to get it working.

With the current approach, there’s a decent performance improvement at least, 140 ms -> 50 ms.

Why did I even do this a performance difference this tiny does not matter at all, the cloud generation runs on a separate thread anyway

I spent the past 8 hours fixing bugs. Like I don’t think I did anything else.

The main issue I was dealing with was the release of Create Aeronautics, an insanely cool mod that essentially just adds realistic physics to Minecraft.

The issue was that it was on 1.21.1, and the 1.21.1 version of the mod is Qendolin’s multiple-year-old version and has a ton of bugs.

And that’s really bad because my development environment is broken for older versions. I essentially had to print debug by building JARs -> copying them over to my PrismLauncher instance (I symlinked it later, saved a bit of time) -> then launching the game, letting it crash, then reading PrismLauncher logs. Was a ton of overhead, but fortunately the fix was simple so it didn’t take forever.

Some bugs I ran into:

Clouds would render over DH LODs. The fix was implemented already but had to be manually configured to turn on for a new version of DH, so I had to do that and also add a fallback for certain DH versions.

Oh and also the render phase name changed for some reason (startDrawing -> setupRenderingPhase). This just causes a crash when drawing clouds. Was annoying as I had to add some code to print the method names on crash, since I didn’t have a debugger or anything to work with.

Clouds also only rendered on a corner of the screen. This is intended and only works on certain shaders, the bug was that the shader preset would change without the user doing so due to a bug. I have created an FAQ, but I still need to reproduce the bug somehow.

That was an epic rant of pain. Next up, the cloud generation times increased by 7x because I added a better algorithm in the latest version. I’m gonna try to optimize it.

Oh hey this is a cool cloud noise pattern

(I can’t look at IRL clouds normally anymore 😭)

huh how has it been 8hrs since my last devlog??? I felt like I didnt do anything

I’ll do a devlog about what I’ve done in 15 mins

I completely forgot to devlog

So once again, here’s one giant devlog covering everything I did over the past few days.

• Sampler visualizers: Allows me to create an image with the cloud noise patterns to get an idea of what the cloud generation would look like. Attached as the first image is the original noise pattern.
• Multiple noise support: Cloud generation uses PerlinSimplexNoise (a Minecraft-exclusive combination of Perlin and Simplex noise) to generate clouds. This can be further configured with Perlin Noise Octaves. I added support for a generator to use multiple noise handlers with different octaves, using a simplex noise pattern to pick the noise generator to use, and interpolate the result between the two closest noise generators. The end result is a nice, smooth pattern (second image) that allows for a lot of diversity in cloud generation.
• Fill cloud void options: The bottom of a cloud is empty, but to fix that I added the bottomSparsity option that would mirror the top cube under the bottom cube, filling in the cloud and making it look like a blob instead of a thin sheet. Third image: cloud with empty bottom, 4th image: cloud with filled bottom.
• Noise preset options: Not quite done yet, but once finished it would let you build your own custom noise pattern. 5th image: a preview of the noise options

And lastly I’ve also included a quick timelapse of the clouds moving by. Each cloud pattern is different from the previous due to the new noise options. Hopefully you will enjoy!

This week is LockIn week 3, and I’m going to be submitting this project for it. Here’s what I’ve done so far:

• Get rid of lunar client sucks message: Qendolin (the original author) added the message; he had a hard time debugging with lunar client when it caused a crash with the mod. I’ve gotten rid of it for now because Lunar seems to work fine right now.
• Use world time for cloud gen: Added a “Time Source” option to the options menu, with the default being World time. As shown in the attached video, this makes the clouds follow the world time and be affected by /time set commands.
• Use world seed for cloud gen: The client doesn’t directly receive the world seed, but an obfuscated version of it is sent upon client login for use within the biome manager. I’m using the same seed to produce deterministic cloud patterns based on world seed (previously was hardcoded with seed 1337).

To do list for this week:
[x] get rid of lunar client sucks message

[x] use world time for cloud gen

• options: clock time, game time, renderer time

[x] use world seed to salt cloud gen

[] sodium extra integration

[] no clouds in desert

[] backport to older version

[] fill cloud void?

[] work with world height datapacks

Did a huge storage refactor.

The main problem with the current code is that upon any action that changes options, a callback must be called to update the field in storage. A similar callback must be implemented for each option upon load. This process is quite tedious, and it also makes the codebase littered with storage calls.

I fixed this by using one central appState variable that defines a schema for the save format, and used a $effect to sync the storage whenever certain variables in appState changed.

But this kinda broke everything, so I had to go around fixing bugs for multiple hours.

Anyway, if you’re reading this, here’s a short userscript to add spacing between the paragraphs of a devlog:

import { pa, pn, pq, ps, pt, pv } from "@page-proxy/pp";

// ==Page Proxy==
// @title Add space after paragraph
// @website https://*.hackclub.com/projects/*
// @description Add space after paragraph to devlogs and ship messages
// @author orangishcat
// @grant run-on-page-load
// ==/Page Proxy==

// ==Selectors==
ps.injectCSS(`
div.post__body p
{
margin: 0.85rem 0;
}
div.post__body li
{
  margin: 0.5rem 0;
}
`);
 
// ==/Selectors==

And here it is in Tampermonkey format:

// ==UserScript==
// @name Add space after paragraph
// @match https://*.hackclub.com/projects/*
// @description Add space after paragraph to devlogs and ship messages
// @version 0.1.0
// @author orangishcat
// @run-at document-start
// @require https://orangishcat.github.io/page-proxy/pp/pp.min.js
// ==/UserScript==

// ==Selectors==
ps.injectCSS(`
div.post__body p
{
margin: 0.85rem 0;
}
div.post__body li
{
  margin: 0.5rem 0;
}
`);
// ==/Selectors==

I’ve attached a side-by-side comparison.(left = before, right = after).

The cursed Neoforge build script makes me lose two hours of my life again. Turns out Neoforge doesn’t use include, it uses jarJar instead. Without it, the dependencies are correctly loaded in the development environment, but bundled incorrectly when building the JAR.

-implementation("com.github.bawnorton.mixinsquared:mixinsquared-neoforge:0.3.7-beta.1")
+implementation(jarJar("com.github.bawnorton.mixinsquared:mixinsquared-neoforge:0.3.7-beta.1"))

Error screen looks something like this:

At least it told me it had something to do with MixinSquared. Better than other mixin-related issues I’ve had in the past that just gave me an arbitrary mixin number like mixin_1793.

Added a settings tool panel.

This controls settings for scripts running on the current tab (top) and also extension settings (bottom). Currently the only extension setting is showing the help tool. Since I’m running out of space and the help tool isn’t all that useful once you’ve used the extension for more than like 5 minutes, it’s the first one to be able to be turned off.

The keybinds are now a bit weird though. Originally it was Shift+1 to Shift+5 from left to right, and now it still is, but the help tool has no keybind since it can’t be turned off, and shifting the keybinds based on whether the help tool appears or not would be a bad idea.

I also want to add some labels differentiating tab settings vs extension settings, but right now I still haven’t figured out how to make them look good.

I also added a @version metadata field that is preserved upon Tampermonkey export. Defaults to 0.1.0 if not present.

Released v0.3.8 today. Just some polish and QOL features.

• Script minification (using terser and css-tree) to minify exported scripts
• Disable all grants button: prevents all scripts from using grants temporarily (basically preventing any scripts from running on page load, as that’s the only grant)
• Export tool UI polish. There’s only a small difference but I think it makes the UI clearer.
• More internal refactors, again. I have to do this often as AI writes code pretty quickly, but I think that’s just a byproduct of faster development. If I wrote everything by hand, I would still have to do refactors anyway, but instead I would just put them off for eternity.

I released v0.3.7, making some convenient API changes, bug fixes, and most importantly, finally getting export to Tampermonkey to work. The current setup is to get Tampermonkey to load the pp library from a url hosted on the demo page for my extension.

The pp library is built into JS via a script which I run manually every time I release a version, and it literally just injects globals into the script. The setup is mildly cursed but I can’t think of anything better right now. Surely this won’t just remain in the codebase forever, just like all my other temporary fixes…

Look! I’m running a Page Proxy script in Tampermonkey!

Painfully updated documentation 😭

This involved manually screenshotting each component and cropping them and filling the unwanted edges in with transparency.

And no, “Capture node screenshot” in DevTools does not set the contents behind the node to transparent, so the background is captured inside of the screenshot. And it doesn’t work in an extension’s sidepanel either.

Did a lot of polish to the website. Also added a nice undo button that pops from the recording stack and selects the element that was selected in the previous stack.

Finally added support for multiple scripts running on the same tab! Previously, the editor would just load the script that was created earlier, which was really annoying as you would have to navigate to a different tab to edit another script. Now there is just a nice dropdown.

Of course, I also updated the documentation and fixed bugs, as usual.

Oh no I forgot to devlog 😭

It’s been 5 days since Minecraft 26.1 released. Here’s what I’ve done so far.

• Cloud pointiness option: A value from 0 to 5, controlling whether the cloud is flat (0), super pointy like a mountain (5), or anywhere in between. The option doesn’t work the greatest so I might have to refine or tweak it in the future.
• Fixed Iris compatibility and shader support. I added the Iris compat back after it broke, and fixed a bug where Iris was using a different view bobbing matrix, causing a weird visual effect where the clouds would not be affected by view bobbing and move side to side relative to the world.
• Fixed shader preset dropdown rendering behind text, mutating option text, etc. This bug has been in the codebase for a long time, so I’m glad it’s finally fixed.
• Fixed clouds not being affected by view bobbing and nausea effects. This was a similar fix to the Iris bugfix I did. Same cause; matrix stack being captured too late.
• Clean up build.gradle (kinda)
• Lots of code quality improvements under the hood.

Here’s some screenshots with the fixed shader support.

Added a feature list to the demo! (which took forever, as usual.) Also made a lot of fixes to the website UI, enforcing consistent margins and padding and whatnot, also cleaning up some of the website code.

Time really flies when Minecraft modding; digging through source code is not fun! I felt like I barely did anything and I was already way past 10 hours.

After previously helping port the mod to Minecraft 1.21.11, I now have to port it to 26.1.

Now starting from 26.2 Minecraft will be moving to Vulkan, so I wouuld have to rewrite the mostly-OpenGL renderer in Blaze3D (Mojang’s rendering abstraction), and later port the shaders to be Vulkan-compatible.

Fortunately 26.1 is still OpenGL only and I still can procrastinate a little bit before having to do all that.

There are a lot of files changed…

Here’s a screenshot of the mod in action.

I finally have a landing page demo now. It took forever but hopefully it was worth it?

Recording tool, probably the most important of them all. Allows you to convert actions you would normally do in DevTools into code.

Also reorganized the entire project, splitting many files.

Lots and lots of polishing and fixes. There’s much more documentation and help links now, so hopefully the product will be a lot easier to use.

Made a few tweaks to documentation and UI.

I worked on documentation, export tool, demo website, and lots more; mostly polishing and stuff for the past few days.

This was supposed to be an easy project, and now I have 40 hours on it…

Just a few more final tests, and I’ll get a first ship out to get some payout. Features are limited but I don’t have time for everything.

The extension is mostly ready for a first ship. Also Monaco editor yay!

The website still has a long way to go, though…

Added this panel that allows you to match elements in many ways

UI implementation is finally done, I can finally begin implementing functionality

Only took nearly 20 hours

I wanted most of the functionality to be in the website, but CORS just doesn’t work like that. Oh well, I guess I’m developing a Chrome/Firefox extension now (using WXT).

Here’s the design.

Main app layout implemented and finalized, would look something like this

Designed the main app in Figma. Progress is slow, but I hope I’ll get better at it.

Designing is actually kinda fun. Previously I made websites by jumping straight into the coding without planning, and the end result was not what I had imagined at all. Using Figma though, this design already looks way better than what I had originally planned, and unlike before, I am actually excited this time to implement this design in code.

Haven’t written much code yet, but I have created some basic wireframes and components