We should support multi-selection of 3D shapes.

This will then allow the following scenario:

• Select shape A
• Select shape B as well
• Press Union/Cut operator -> The form is already filled with shape A and shape B as parameters

I think https://3dviewer.net has a good design for the panels. Putting both the tree view and properties view in one panel makes things become very small. We should move the latter to the right panel.

Their tree view looks good too.

Now the opencascade package is available, we should be able to use the bot to update screenshots.

We should provide to the user some view options:

• orthographic and perspective projections
• wireframe view
• clipping

When clicking on shapes in the tree we close other objects, we should not change the tree state when clicking on shapes

Currently the WASM part of JupyterCad fills up user memory and can lead to issues on low memory laptops (my browser or other intensive apps like vscode simply crashes from time to time when using JupyterCad).

We should make a custom open-cascade build to reduce the amount of data shipped to the user.

See https://ocjs.org/docs/getting-started/file-size
Also https://github.com/donalffons/opencascade.js/blob/master/website/ocjs-editor-theme/src/customBuild/customBuild.yml

All the objects are wrongly placed at the same level in the tree view.

It looks like one edge out of two is visible on round meshes:

This was probably introduced by #9, indexing might be off.

It's a cool feature of opencascade web. It would be great if we could add it to jupytercad

We should show other collaborators cameras in the 3D view, we can use the CameraHelper for this: https://threejs.org/docs/#api/en/helpers/CameraHelper.

Wishlist for this feature:

• show camera with the right user color
• when clicking on someone's camera, switch to following that view

Currently the follow mode will make the user camera follow the other collaborator camera. We could probably redesign it so that when following someone else's camera we don't change the user camera.

We could have some kind of model/view design where:

• when following someone, we don't change the user camera model, only the view
• that means that if user1 follows user2 and user2 follows user3, user1 will see a static camera because user2's camera is actually not moving

Deleting an object actually deletes another object:

We should pre-fill the dialog when creating a new operation with the current selection

We should probably have an undo/redo feature with Ctrl + Z binding

Add a click behavior in the left panel so that when clicking the annotation it expands/blinks in the 3D view and goes into focus

It looks like resizing the 3-D view breaks the color picking

When applying an operator on two primitives we should hide the primitives.

JupyterCad can only open existing jcad or FCStd files. We should allow users to create new empty files and then save them in both supported formats.

As in https://github.com/QuantStack/jupytercad/actions/runs/4893347948

It seems to fail at downloading artifacts

1s
Run dawidd6/action-download-artifact@v2
==> Repository: QuantStack/jupytercad
==> Artifact name: extension-artifacts
==> Local path: ./
==> Workflow name: build.yml
==> Workflow conclusion: 
==> PR: [1](https://github.com/QuantStack/jupytercad/actions/runs/4893347948/jobs/8736202593#step:5:1)29
==> Commit: 3690a3c12af53fb17695c4ae12f89[14](https://github.com/QuantStack/jupytercad/actions/runs/4893347948/jobs/8736202593#step:5:15)0e10f4059
==> (found) Run ID: 4893335739
==> (found) Run date: [20](https://github.com/QuantStack/jupytercad/actions/runs/4893347948/jobs/8736202593#step:5:21)[23](https://github.com/QuantStack/jupytercad/actions/runs/4893347948/jobs/8736202593#step:5:24)-05-05T12:24:42Z
==> (not found) Artifact: extension-artifacts
==> Found the following artifacts instead:
Error: no artifacts found

While switching cameras between users, the new position is set immediately. We can make the camera move smoothly to the new place.
For example, the case of changing the up vectors in 3dviewer.net

Currently, when hiding an object we change the camera position, which can be a bit odd for users

After some operations, the view is quickly not centered anymore on the object. We should have a button that re-center the view to the visible objects on-click.

We currently have annotations (comments) implementation into JupyterCAD https://github.com/QuantStack/jupytercad/tree/main/src/annotation. This should be moved into another package so that it can be reused in other extensions (Notebook, GlueLab etc). A potential place to put this could be https://github.com/jupyterlab/jupyter_collaboration

We should show a nice error message to the user if they try to read an FCSTD file wihout freecad installed

When creating e.g. a box with the "New Box" command, the "name" entry should be filled up automatically with "Box" or "Box N" if all [1, N-1] names are already taken.

FreeCAD allows creating a Tube primitive. It would be nice to do the same.

Unfortunately there is no BRepPrimAPI_MakeTube* function in OpenCascade so we might need to implement the shape in another way if we want to provide this primitive.

browser_check is broken with the changes in jupyter_server 2.0.2 (jupyter-server/jupyter_server#1143). It is disabled in our CI and should be re-enabled once it's fixed upstream.

Related PR #77

For now, JupyterCad can only be opened from the main interface of JupyterLab, I think it's possible to extend the scope of JupyterCad to make it become:

• A notebook widget where users can visualize and interact with jcad or FCStd files using python APIs.
• A lab-based app, which can be opened independently of JupyterLab by calling jupyter cad

It would be nice to have a simple top bar menu and a simple launcher.

Currently, you delete an annotation by pressing the small cross handle on the top right corner.

To me, this button looked more like a "close" button than a "delete" button. We should probably have both a close button and a delete button, and make it clear what delete does.

@deathbeds/jupyterlab-rjsf creates forms from schema files with a JupyterLab look and feel. It is compatible only with JupyterLab 3, we need to pin some dependencies to make it work in JLab4 but it is not a long-term solution.
We need to update this library to make it compatible with both JLab 3 and 4.

deathbeds/jupyterlab-starters#97

We should be able to read IFC files with ifcopenshell, available on conda-forge https://github.com/conda-forge/ifcopenshell-feedstock

Add a Delete key binding for removing shapes

When filling a form e.g. for creating a box, pressing "Enter" does not validate the form, it discards it. We should have "Esc" discard the form and "Enter" validate it.

We should improve the toolbar and use some nice icons for primitives and operations:

Our opencascade build system needs docker installed but it is not available in the binder machine.

#8 introduced 3-D selection, we should bind this selection to the tree so that:

• when clicking in the tree view the clicked piece gets highlighted in the 3-D view
• when clicking in the 3-D view the clicked piece gets highlighted in the tree view

jupyterlab/maintainer-tools/.github/actions/update-snapshots@main

Currently, we are not able to display the robot arm from the FreeCAD example. We should see if that can be fixed.

We should probably not commit the wasm file as it's 4.5MB

Reference:

There might be a way to switch from a light to a dark theme already, similar to what we did for the tree view.

But we can also also extends react-jsonschema-form so that it uses Lumino widgets? https://react-jsonschema-form.readthedocs.io/en/latest/advanced-customization/custom-themes that way it would integrate better in the Lab UI. This should be a parallel project/repo though.

I am not able to select from the tree view anymore:

• it does not update the selection in the 3D view
• it does not show the current selection in the tree view