Improve camera init

[archmoj]

This PR would improve camera initialisation setup for gl3d plots.
First - it fixes bug #3576 by disabling undesirable rotation/pan when the click is done on the initial mouse location.

Second - it removes one extra expensive call to create camera object. This is done by improving gl-plot3d interface to also accept & reuse camera object created inside plotly.js scene. Also gl-plot3d's createCamera and createScene are revised and their mouse listeners are packed to be enabled by centralised functions.

Finally - all the previous gl3d jasmine tests with noCI flag could be run on CI.

@plotly/plotly_js

[antoinerg]

@archmoj I checked out your branch locally and I can't rotate any of the gl3d mocks!

This mock does not rotate: http://localhost:3000/devtools/test_dashboard/#gl3d_surface_after_heatmap
whereas it used to: https://rreusser.github.io/plotly-mock-viewer/#gl3d_surface_after_heatmap

Maybe there is something wrong on my side... Please let me know if rotation does work for the mock above. Also, if it doesn't work, we really should have a test to catch this 😮 !

[archmoj]

@antoinerg thanks for testing this. Interestingly it used to work without that so I thought it could be enabled by the gl-plot3d module. With that we may possibly enable it once the scene is complete.

[antoinerg]

@archmoj so you do confirm that you also can't rotate the gl3d traces? If that is the case, I am really worried about the fact that no tests failed. Why is that?

[archmoj]

@archmoj so you do confirm that you also can't rotate the gl3d traces?

@antoinerg Yes, I do confirm that.

[archmoj]

https://github.com/gl-vis/gl-plot3d/compare/accept-camera-object

[archmoj]

@archmoj so you do confirm that you also can't rotate the gl3d traces? If that is the case, I am really worried about the fact that no tests failed. Why is that?

I added this test in respect to that.

plotly.js/test/jasmine/tests/gl3d_plot_interact_test.js

Lines 131 to 145 in a200e78

	.then(_clickOtherplace)
	.then(delay(20))
	.then(function(){
	varcameraFinal=gd._fullLayout.scene._scene.glplot.camera;
	expect(cameraFinal.up[0]).toBeCloseTo(0,2,'cameraFinal.up[0]');
	expect(cameraFinal.up[1]).toBeCloseTo(0,2,'cameraFinal.up[1]');
	expect(cameraFinal.up[2]).toBeCloseTo(1,2,'cameraFinal.up[2]');
	expect(cameraFinal.center[0]).toBeCloseTo(0,2,'cameraFinal.center[0]');
	expect(cameraFinal.center[1]).toBeCloseTo(0,2,'cameraFinal.center[1]');
	expect(cameraFinal.center[2]).toBeCloseTo(0,2,'cameraFinal.center[2]');
	expect(cameraFinal.eye[0]).not.toBeCloseTo(1.2,2,'cameraFinal.eye[0]');
	expect(cameraFinal.eye[1]).not.toBeCloseTo(1.2,2,'cameraFinal.eye[1]');
	expect(cameraFinal.eye[2]).not.toBeCloseTo(1.2,2,'cameraFinal.eye[2]');
	expect(cameraFinal.mouseListener.enabled===true);
	})

[antoinerg]

I could reproduce the bug and confirm that this PR fixes it. Here's the Codepen with the fix: https://codepen.io/antoinerg/pen/rRMOox?editors=1010

@etpinard, I think it's ready to go but maybe we should wait before merging to master?

[etpinard]

This looks good! I'm curious: how expansive was that extra call to create camera object?

I'm a little scared of releasing this in a patch release. Maybe we should wait for 1.46.0 🤔

@archmoj have you manually tested zoom/pan/rotate on multiple gl3d mocks?

[archmoj]

This looks good! I'm curious: how expansive was that extra call to create camera object?

Console.time may not not show much difference in terms of milliseconds. But it was complicated to modify camera object or link other events after it is was created in plotly.js, mainly because it was recreated and re-linked by gl-plot3d.

I'm a little scared of releasing this in a patch release. Maybe we should wait for 1.46.0

Yes I totally agree.

@archmoj have you manually tested zoom/pan/rotate on multiple gl3d mocks?

Yes and they seem to be working.

[etpinard]

Yes I totally agree.

Great. Let's put this under 1.46.0

[archmoj]

@etpinard This PR may be a good candidate to merge now that #3601 merged. What do you think?

[etpinard]

This PR may be a good candidate to merge now that #3601 merged. What do you think?

It is! I'm testing it one last time now. 👌

[etpinard]

Not anything new, but I noticed this test

plotly.js/test/jasmine/tests/isosurface_test.js

Line 272 in e938c40

it('should clear *cauto* when restyle *cmin* and/or *cmax*',function(done){

is missing a @gl tag.

[etpinard]

💃 once you fix https://github.com/plotly/plotly.js/pull/3585/files/e938c407c5889b60d04ee9cbdc3835bfba111a02#diff-72ac990ccb2a9251e4dcf5e7090af0c4 and bump gl-plot3d

[etpinard]

💃 💃 let's get this in master 🍾