<kintel>
teepee Re. 3mf color import: If the 3mf file contains multiple meshes, it looks like we currently perform a CGAL union on import, wiping out all colors. Did you look into that as part of color import?
<teepee>
yes, added option to use manifold ops
<kintel>
cool, with the most recent commit, you can just look up render settings for which backend is selected :)
<teepee>
hmm, I'm in that 3mf export branch, so I need to update that, but it needs some cleanup anyway
<kintel>
Are you planning to merge your import work into guso78k's branch?
<teepee>
that might make sense as that would allow for export->import testing
<kintel>
_1
<kintel>
+1
<teepee>
it's quite a rabbit hole
<teepee>
as the build items can reference object that are meshes or objects that have components
<TylerTork>
good morning me fabbies. My question today is about lighting. Is there not a way I can get light on my model so I can view it from the angle I want and still have decent contrast to show the features, so I can take a snapshot? I've had to resort to adding a Customizer parameter to rotate the model so I can get it lit correctly.
ooxoo has quit [Remote host closed the connection]
ooxoo has joined #openscad
<J24k17>
iirc light is tied to the camera - so you have to rotate the model .. but can't change the angle to the lightsource
<TylerTork>
J24k17: it's very frustrating trying to get it right. Say I put in a rotation parameter, how do I know what rotation to enter to get the lighting I want? I can see how it changes as I alter the values, but then I have to try to orient the camera for the shot and that totally changes the lighting.
<TylerTork>
Ultimately I had to take the screenshot with the model upside-down on the screen, then rotate the image in post. It's fairly ridiculous.
<kintel>
I guess one confusing aspect of having two backend: If we switch backends, we won't blow the geometry caches, which makes already cached geometry calculated from one backend be accessed from the other backend
<J24k17>
Rendering cancelled on first warning. - That would be something the user can't really influence
TylerTork has joined #openscad
<TylerTork>
J24k17: Don't see how that solves it. I want a way to take quick screenshots -- and actually see surface features -- in the OpenSCAD client. What is this, a command line utility?
<J24k17>
to check something you can use modifier and simply change the view
<TylerTork>
as someone who also does a little photography, to take still shots, I want to orient the model to the camera first, then move the lights around until it looks right. Being able to make a pretty animation is nice, but there's no guarantee any frame of it will be what I want.
TylerTork has quit [Ping timeout: 256 seconds]
TylerTork has joined #openscad
<TylerTork>
J24k17: As someone who first heard of this TODAY, the relevance of PovRay to my issue is unclear. Rather the providing more links, can you write one sentence telling how I can apply this to getting a good still shot without writing scripts, exporting files, etcetera? I just want to get a decent view of a model in OpenSCAD preview without having to
<TylerTork>
jump through hoops.
<J24k17>
TylerTork if you want nice images/animations you don't use openSCAD - you dont't have materials, shadows, reflections etc - you use a raytracer like povray .. or blender
<J24k17>
for using only openSCAD - select the right colors to get contrast for your objects
<TylerTork>
Yes, I get that. And if I need NICE images I'll go whole hog. What I need more often is an image where I can SEE WHAT THE F*** I'M LOOKING AT.
<TylerTork>
Horrible, and no way to adjust the lighting while still getting the angle I wanr
<J24k17>
both look quite equal
<TylerTork>
J24k17: not to me. The contrast in the second image is markedly less. The color of the pegboard is nowhere close to the specified color. It's dark, it's boring. I want a crisp and attractive shot. It doesn't need to be beautiful, I just don't want to have to strain to make out features. This is also an accessibility issue since colorblind people
<TylerTork>
really need that high contrast.
<TylerTork>
Compare with this: file:///C:/Users/tyler/OneDrive/Documents/Lightshot/Screenshot_2.png
<JordanBrown>
The difference between the darker image and the lighter ones is that the darker image has the face pointing towards us and left, while the lighter ones are pointing right.
<J24k17>
TylerTork as i said - you want a nice image you can now use povray to get it .. or blender (which is probably more work)
<J24k17>
I agree that it would be nice to set the light source in openSCAD too .. you can open a feature request for this
<TylerTork>
J24k17: There is already a feature request which I added a "me to" to. Meanwhile, can you write one sentence telling how I can easily apply PovRay to getting a good still shot?
<TylerTork>
Do I -- what? Export a 3mf file then open it with...?
<JordanBrown>
Are those shots orthographic, or perspective?
<J24k17>
I think we had some shaders but they are not implemented afaik
<TylerTork>
Do I have to choose now or can I decide once I see how either of them looks?
<J24k17>
TylerTork the next dev snapshot version should have included the option to "export povray" which is then opened in povray and press render
<J24k17>
RUN it is
<JordanBrown>
I'm wrong about which way the image is pointing. But given the same camera angle, you should be getting the same colors.
<TylerTork>
So I need "POV-Ray for Windows" and that will integrate with OpenSCAD?
<JordanBrown>
With an experimental feature that's coming out in the next few days.
<J24k17>
another option is to export the geometry as 3mf / off / obj which should include the colors on the next version so you can use any 3D-software like 3D builder to get a nice image
<JordanBrown>
And where "integrate" means that OpenSCAD can export a file that POV-Ray can import.
<TylerTork>
Jordan, the camera angle is not the same. One is from below and one from above. I had to jump through some hoops to get the part aligned so I could get good lighting on it, then turn the picture to the right orientation.
<teepee>
you can even send the file directly now
<J24k17>
you click the file that is exported ↦ which should open PovRay ↦ click RUN
<JordanBrown>
But OpenSCAD is primarily a modeling tool; it doesn't try to make high-quality images. POV-Ray tries to make high quality images.
<JordanBrown>
Right, if the camera angles are different then how the light hits the part is different.
<TylerTork>
That would be close enough for my purposes. I'll be watching the skies for the next dev drop then
<teepee>
you can register on the github PR which should give an email once it's merged
<teepee>
yes, like I said to not just break all import (or change the output) we first decided to only allow increasing resolution as compared to the previously hard-coded values
<JordanBrown>
kintel thanks, just wanted to make sure somebody knew.
<JordanBrown>
One of the key expectations that should be set for $fa/$fs is that they do not produce precisely defined results, that they only control the approximation used.
<teepee>
1) treat the length of the curve as the diameter of a circle
<JordanBrown>
"the length of the curve" is remarkably difficult to determine.
<teepee>
calculate the length of the arc covering the $fa angle
<teepee>
use that as segment length if bigger than $fs
<J24k17>
swiggly lines ?
<teepee>
for that, no, ignore all curving, use bounding box
<teepee>
well bounding circle
<JordanBrown>
even using a constant segment length is hard.
<JordanBrown>
But maybe something like "take the straight-line distance between the two points, and divide by $fs to get the number of segments" would be better than nothing.
<JordanBrown>
Or the distance from the first point to the first control point to the second control point to the endpoint.
<teepee>
2 big sources for deep diving that. Pomax and Freya Holmér
<othx>
teepee linked to YouTube video "The Beauty of Bézier Curves" => 1 IRC mentions
<JordanBrown>
Sounds like you're ahead of me. My mystery is solved - you said that the $fX applied, and I couldn't see any effect, and that's now understood.
<JordanBrown>
While I'm here... kintel, teepee, I'm still waiting on what work would call architectural review on the PR#4478 / OEP8 object work.
<JordanBrown>
Not code review or testing, just specification review.
<nihil>
hrm, anyone had issues with svg exports to slic3r for DLP/SLA?
<kintel>
JordanBrown I say post-release. I don't have the mental surplus to start thinking about new language features right now
<JordanBrown>
OK
<teepee>
same here unfortunately, the only thing I can imagine is maybe starting to collect a number of points to validate against
<JordanBrown>
I'm happy to go back to work on it, but don't want to spend time fine-tuning when I haven't even gotten an OK for the general approach.
<teepee>
like namespaces, imports, ...
<teepee>
yeah, fine tuning seems not useful
<teepee>
(yet)
<teepee>
I tried merging with master but got stuck with sometimes no geometry
<teepee>
getting a version to build would be useful as right now we can't ask anyone to try
<teepee>
artificats time out
<teepee>
30 day IIRC
<JordanBrown>
Keeping the changes in sync is kind of like fine tuning... I don't mind doing it, if somebody will look at it. But mostly you don't need a build; the specs describe the syntax and user-facing data structures.
<teepee>
for bringing other people into the discussion it's pretty important
<JordanBrown>
But I can resync if that will increase the chances that somebody will look at it.
<teepee>
well, I tried and failed a while ago
<JordanBrown>
I just don't want to resync forever without getting feedback from the key people.
<teepee>
I'll also grab those versions so we can put them on the file server
<JordanBrown>
But if all you really want is *some* version that is built, I could just push a trivial change onto the existing PR and that would cause a new build.
<JordanBrown>
They're a year or so old, but since there hasn't been much language evolution since then they are valid for "play with it" purposes.
<kintel>
JordanBrown It's technically possible to hide it all behind an experimental feature if that helps. We reserve the right to completely change anything experimental, but if it's still up in the air, we could tag it as "really experimental" :)
<JordanBrown>
that's not the issue (yet)... I'm not claiming that it's ready for even experimental-in-nightly status yet. But I don't want to spend much more effort without having key people (and here I mean you and teepee) say "yeah, that new syntax looks more or less acceptable".
ooxoo has quit [Read error: Connection reset by peer]
ooxoo has joined #openscad
myosotis has joined #openscad
teepee_ has joined #openscad
teepee has quit [Ping timeout: 260 seconds]
teepee_ is now known as teepee
J24k72 has joined #openscad
J24k66 has quit [Ping timeout: 256 seconds]
<JordanBrown>
Anybody know what the "term" output format is? Or, a bit closer to what I'm trying to understand, what CSGTreeEvaluator is used for?
<JordanBrown>
ah, never mind, pilot error, I dropped debug statements in a different function than I thought and was mystified that they didn't come out.