I've been working away at something like this for years. After a couple of
years of studying the code and understanding how things like Gold Feather
work I have shifted my focus to some more core ( I would argue ) useful
problems to do with things like MFPR / Parallaelization. Where I got up to
was a web based front end that hands out constructive geometry tasks to a
back end cluster and does some pipe line type stuff and some other
optimizations. It was definitely faster but I have shelved it as I have
come to see my implementation as flawed because of the web stuff involved.
So I have quit stuff like Node and the Three.js and chucked all that out.
Instead I've successfully managed to build the very basics of the interface
in web assembly. That is Scintilla. The Qt basics of the Qt stuff. But I
have halted this because as stated above I just come to see that some more
core approaches need a rethink. And those core problems are not really
OpenSCAD problems. OpenSCAD does what it was designed to do and in
retrospect I think it is pretty good / excellent code. Just trying to say
this path has been well walked already and this is what I have learned.

On Wed, Mar 10, 2021 at 9:41 PM irevdev via Discuss <
discuss@lists.openscad.org> wrote:

--
Damien Towning
CEO / Developer
http://makertron.io

Not true, you can rotate around Y by holding shift when dragging with the
mouse.

On Wed, 10 Mar 2021 at 11:00, Damien Towning connolly.damien@gmail.com
wrote:

On 10.03.21 11:40, irevdev via Discuss wrote:

Indeed. I'm even more sad the emscripten based solution died.
I really would like to see that happening at some point again.
Not as replacement for the Desktop version but as useful extra
way of generating (simpler?) models.

This may also open the option of defining versions to run the
scripts against. I really love the way https://godbolt.org/
does that for a huge list of compilers.

Yes, it should in normal cases. I wonder if it would help to
have some defined resolution limit to prevent trivial attacks
just setting $fn to huge values. It probably will still need
some time threshold.

I don't know too much about Three.js, but there's a branch
working on using a newer CSG code path in CGAL which is much
faster in most cases. So maybe there's more chance generating
the actual mesh.

Yep, I certainly do :-). I'll think about options to maybe
also put up a related topic for Google Summer of Code which is
just starting this years instance.

ciao,
Torsten.

Sorry I'm a bit slow to replying again to this thread. I was head down trying
to get the coordinate systems synced between OpenSCAD and Three.js. I've had
success, definitely still problems which I've got some questions about but
really happy with the feel of the UI so far. Even though it's not real time
I think it's a really good compromise. There's a video of where its up to in
this PR https://github.com/Irev-Dev/cadhub/pull/238 or again you can play
with it at https://cadhub.xyz/dev-ide.

Rotation seems to be working perfectly, and I've got the fov matched between
the two as well. What's not working is when you "move" the model (with right
click drag) the axes get separated. I thought this might just be a scaling
issues, but nothing I've tried is working with the exception of using very
low values of the camera dist ie
--camera=${px},${py},${pz},${rx},${ry},${rz},5  the last number 5 or less,
but doing so seems to effect the distance at which objects remain in view,
so it's really not practical to use such low numbers, anyone have any
insight or ideas with this?

years of studying the code and understanding how things like Gold Feather
work I have shifted my focus to some more core ( I would argue ) useful
problems to do with things like MFPR / Parallaelization. Where I got up to
was a web based front end that hands out constructive geometry tasks to a
back end cluster and does some pipe line type stuff and some other
optimizations. It was definitely faster but I have shelved it as I have come
to see my implementation as flawed because of the web stuff involved. So I
have quit stuff like Node and the Three.js and chucked all that out. Instead
I've successfully managed to build the very basics of the interface in web
assembly. That is Scintilla. The Qt basics of the Qt stuff. But I have
halted this because as stated above I just come to see that some more core
approaches need a rethink. And those core problems are not really OpenSCAD
problems. OpenSCAD does what it was designed to do and in retrospect I think
it is pretty good / excellent code. Just trying to say this path has been
well walked already and this is what I have learned.

A full web-deployable openscad, sounds epic.
Currently my approach is to have a very light touch, wrapping openscad with
some web technologies to get it on the web with minimal effort and no
changes to openscad, I'm optimising for being able to easily upgrade with
new versions of the software, thanks for the insight.

mouse.

Oh okay, good to know.

Oh I didn't know about this :(

scripts against. I really love the way https://godbolt.org/
does that for a huge list of compilers.

I don't think I follow, versions of openscad to run scripts against?

have some defined resolution limit to prevent trivial attacks
just setting $fn to huge values. It probably will still need
some time threshold.

So I'm planning on just going with timeout to begin with, I don't want to
start parsing the code if I can avoid it, but if there were some way of
limiting max $fn in the command line, that would be very useful.

working on using a newer CSG code path in CGAL which is much
faster in most cases. So maybe there's more chance generating
the actual mesh.

Oh awesome, I'll keep an eye on that.

also put up a related topic for Google Summer of Code which is
just starting this years instance.

Did you have much of a look. I'm happy to even jump on a call to give a bit
of an overview to anyone if that's useful. I'm using node to run the command
line, partly because it's what I'm best at and I want to keep the CadHub
repo all one language where possible.
Keep me in the loop with the summer of code stuff and if I can be of
assistance.

--
Sent from: http://forum.openscad.org/

You may not need to run it, just pick out the javascript pieces that manage customizer variable extraction. I believe all that was done directly in the browser anyway, although I may be wrong..

-Marius