Man, nobody has posted since last year...

[Duthomhas]

😃










Frist!

[zapshe]

Maybe this will be the year I start learning to program in C++

[Duthomhas]

It’s interesting how little C++ I’ve been using lately. Most of my programming time lately (which isn’t much) has been in Python (making add-ons for Blender).

Latest offering:
Shell script to identify .blend file version
https://blenderartists.org/t/shell-script-to-identify-blend-file-version/1625311

[zapshe]

I haven't been using C++ too much either. C# and some other languages, but my work generally doesn't benefit from C++ much anymore. No one is writing complex apps and such in C++.

I use Blender mainly for designing parts to 3D print and stuff, very low-power Blender user compared to many. Most complex things I've done was a chess board (Code Geass themed hehe) and a gearbox (which was surprisingly efficient).

[Duthomhas]

Don’t write it off — C++ is still widely used in a lot of very heavy applications and industries, showing up in places you would never imagine it to be, and not just for maintenance of old projects.

[zapshe]

Well, where ever C++ is, it's not in the AI training I'm doing much. Lately, training has been using AI via CLI to fix or add features to Github repos and see how they do.

They do really good. Claude's top model is absolutely great. ChatGPT's top model is astounding.

But yea, most repos are written in Python, Typescript, etc.. There's been a gross shift of creating GUI apps that are essentially just a browser. Slap a cross-platform GUI together with Typescript - seeing that a lot with these Github repos.

[Ganado]

Ddog wrote:

Latest offering:
Shell script to identify .blend file version
https://blenderartists.org/t/shell-script-to-identify-blend-file-version/1625311

Just wanted to say, that must be one of the most thorough installation instructions I've ever seen. I wish all GitHub readme's were that clear.

edit: s/throughout/thorough/

[Duthomhas]

Heh, I usually try to make things a whole lot more one-clickety... but I guess I just didn’t want to bother for that one. Didn’t stop me from writing a half-done blag, though...

Either way, thanks for the compliment. 😊


_{I should probably go back and make it easier.}

[kbw]

I'm still using C++ as my primary programming language.

Performance matters in my area.

[Duthomhas]

(Zapshe uses Python, etc for the machine learning work he’s doing... Wait until he learns how most of the Python modules used to implement that are written in C++... AFAIK)

[zapshe]

(Zapshe uses Python, etc for the machine learning work he’s doing... Wait until he learns how most of the Python modules used to implement that are written in C++... AFAIK)

Well I personally haven't had to code much. It's mostly checking the work of the AI's code. And in that case, mostly Githubs repos which end up being Typescript/Javascript, but Python and other languages are in there too.

I'm always a C++ lover. I just haven't really touched in... a while.

Some of my work isn't even programming related at all, but writing really complex engineering scenarios for the AI to figure out.

They probably think they've already saturated AI training data with good programming examples enough.


I know Python uses C++ for a lot for backend, and I don't feel like I've been far from C++ though since I use C# for some personal projects which feels similar.


I just don't know how much longer a simple human like me will be useful for AI training.

[Duthomhas]

—sells soul for lucre—
—waiting to be reaped—

lol


But hey, don’t put yourself down. You've got a good job.
Humanz rule!




_{^{All hail our robot overlords!}}

[kbw]

Checking the code generated by a LLM is a full time job in itself.

[Ganado]

I love and hate C++. I think it's ridiculously complex and I haven't been keeping up with the latest standards, but I keep coming back to it.

[kbw]

Can't disagree with that.

The organisation of the development of the language encourages a lot of navel gazing and fame seekers. It's very different from community driven languages.

Most of the stuff in there obscure and niche, and in the general case, inapplicable to the real world.

Occasionally you see brilliant libraries like simdjson, but for a language who's design goal was to create libraries, it's disappointing.

[zapshe]

—sells soul for lucre—
—waiting to be reaped—

Ha! Exactly. Honestly, pays well and the only job I could get in this economy.. Programmers aren't really in style anymore.

I was ready to go down some other paths to making money...

Checking the code generated by a LLM is a full time job in itself.

Literally.

I love and hate C++. I think it's ridiculously complex and I haven't been keeping up with the latest standards, but I keep coming back to it.

Just when I think I'm a C++ pro, I find some C++ code that I can't begin to comprehend lmao.

[Duthomhas]

I figure it is worth posting information that is cool or new or useful, so:

https://eater.net/quaternions

Wanna understand the beasts? You’ll never find a better resource.
(And, for the first time, I think I’m getting a good grip on how they work.)

[Ganado]

Wow, I didn't know Grant Sanderson made those demos (with Ben Eater making the WebGL for it, it sounds like). Cool stuff. Still is going over my head at times.

[Duthomhas]

I know, right?

I’m currently designing a 3D a bubble level that actually keeps its bubble upright and level for you.

For context, the vial lies centered along the Y axis, with +Z as “up”. Hence, rotating the thing around the X axis tilts it left and right, Z spins it, and Y should be ineffectual (because that bubble should always be UP).

In order to make the bubble stay up, I have to calculate the amount of Y spin necessary to counteract anything the user does around that Y axis. (The technique is called "billboarding" actually.)

Currently I have to use a separate object positioned directly above the center of the level and a rotation constraint on the Y axis to keep the thing properly up.

But I figure I ought to be able to compute it without the separate object using quaternions. If the imaginary “axis”, if you will, is positioned along the Euler Y axis, then 2*arcsin(W) ought to be the amount to rotate the Euler Y back to “up”.

[EDIT: that was for the Blender swing & twist’s `W`, IIUC; for a quaternion, the equation is asin(2*(W*Y - X*Z)), jsyk]

But there is something somewhere I am not understanding about quaternions, or that the Blender devs didn’t get right (unlikely), or that Blender’s “swing and twist” rotation mode does not line up with my conceptual understanding of how it is computing its data, or...

The result is that I can only get that bubble up -most- of the time and -mostly up- the rest of the time...

So using a separate object it is for now...