Surfaces are too dark

[ChristopherChudzicki]

Left: v0.0.5 / Right: v2+

@sritchie I dunno if this is a consequence of changes we've made in Mathbox, or a consequence of the ThreeJS upgrades. Do you have any idea? I think you've looked at more versions of ThreeJS than I have. Otherwise... I'll try to learn something about ThreeJS :)

[sritchie]

Interesting, I only remember the second one through all my testing. I will look!

[sritchie]

It looks like it was dark at 98c37401f337cc61460e9b9d9e6f32bf0fdf78b7, my first big merge (check out and run npm install && gulp build). So I must have introduced some problem during the migration from Coffeescript...

[sritchie]

If it helps to debug, it also happens in:

file:///Users/sritchie/code/js/mathbox/examples/test/vertexcolor.html file:///Users/sritchie/code/js/mathbox/examples/test/transition.html file:///Users/sritchie/code/js/mathbox/examples/test/surface.html file:///Users/sritchie/code/js/mathbox/examples/test/subdivide.html file:///Users/sritchie/code/js/mathbox/examples/test/repeat.html file:///Users/sritchie/code/js/mathbox/examples/math/procedural.html file:///Users/sritchie/code/js/mathbox/examples/math/hopf.html file:///Users/sritchie/code/js/mathbox/examples/demo/shapes.html file:///Users/sritchie/code/js/mathbox/examples/demo/cylindrical-stream.html

It does NOT seem to happen in file:///Users/sritchie/code/js/mathbox/examples/test/fragmentcolor.html file:///Users/sritchie/code/js/mathbox/examples/demo/cylindrical.html

which seemed surprising.

[ChristopherChudzicki]

• Left: Steven's original at eaeb8e15ef2d0252740a74505a12d7a1051a61b6 using ThreeJS v71
• Middle: Znah https://github.com/znah/mathbox/commit/593bdab1d4a482fe7d496e72ff1ccec691c597ff using ThreeJS v84
• Right: This current repo, with ThreeJS v137

Whatever happened, I'm fairly certain it is not related to Shadergraph. I dropped an old version of Shadergraph into the current repo and stuff was still dark. I also plugged in an old v0.05 version of shaders.js (recall it used to get built to build/shaders.js) and things were still dark.

Znah's is not as dark as modern, but it does have dark lines. Steven's original looks best :/ Hope we can get back to that.

EDIT: I think there are two separate issues: Shading and line color. If you turn off the lines, znah and 0.0.5 look the same.

[sritchie]

@ChristopherChudzicki I agree. here are some clues:

• change "shaded" to false on current master and you'll get the nice luminous colors again, BUT with dark lines and no shading:

• On 0.0.5, setting lineX or lineY to true vs false has no effect. the lines are missing in both cases. THAT feels like a bug to me in 0.0.5.

So that makes me suspect something in https://github.com/unconed/mathbox/blob/0.0.5/src/render/meshes/base.coffee#L120-L186... but they really read the same.

[ChristopherChudzicki]

On 0.0.5, setting lineX or lineY to true vs false has no effect. the lines are missing in both cases. THAT feels like a bug to me in 0.0.5

I'm pretty sure this does have an effect in 0.0.5, although the effect is not visible unless you're using a shaded surface. If not using shaded surface, the gridlines and surface are exact same color and you can't differentiate between the two. (I think the original intent was that the grid lines would be. adarker shade of the main surface color. I don't understand why the gridlines turned black, but I think I do understand why they were not behaving as intended in 0.0.5. Changing the x/y/z in this code to r/g/b fixes the lineX lineY issue in current master, and changing the corresponding lines in the old coffeescript code fix it in 0.0.5. https://github.com/unconed/mathbox/blob/4fd08d6cfdd41446685b53190be833055c10086f/src/primitives/types/draw/surface.js#L278-L288. It would be nice to understand why the turned balack, but I'll settle for understanding why they weren't working as intended originally...).

As for the why the shaded surface itself is darker... I spent a while investiating that this weekend. I still do not fully understand. The shading is done by the two GLSL shader snippts surface.position.normal and mesh.fragment.shaded. . I'm a GLSL noob, but my understanding is that the "inputs" to these shader snippets are the "uniforms", and the current master shaders vs 0.0.5 appear to be getting the same inputs.

As best I can tell, there's an issue with gl_FrontFacing, though I really don't understand why it was working in ThreeJS r71 but not ThreeJS r137. Sigh. See #26 for more.

I'll open a PR for the lineX, lineY thing soon, and hopefully get back to the mathbox-react bug you opened. (Thanks for opening that!)

[sritchie]

Fixed by #27.