Non-spherical particles in the structure viewer

[rosecers]

Putting this in here to bug my student @arthur-lin1027, along with a transcript between @Luthaf and myself.

[11:19 AM]
guillaume With 3Dmol, you could use https://3dmol.csb.pitt.edu/doc/GLShape.html#addCustom, manually constructing the ellipsoid shape
[11:20 AM] jsmol can do this natively, you could resurect the old renderer based on it and dynamically switch between 3Dmol and jsmol if the structure has ellipsoids to render
[11:20 AM] If you want to add this to chemiscope, could you open an issue to discuss changes to the data format to add required ellispoid information to the input?

@Luthaf I have done the thing

For the record, the data format changes are minimal, we can save the necessary information into the ase frame and add a function that changes that data (e.g. ellipsoid semi-axes and orientations) into isosurface data within the chemiscope backend. I would suggest a particle-type flag in the frame info/arrays that triggers an isosurface construction. Thoughts?

[Luthaf]

For the record, the data format changes are minimal, we can save the necessary information into the ase frame and add a function that changes that data (e.g. ellipsoid semi-axes and orientations) into isosurface data within the chemiscope backend. I would suggest a particle-type flag in the frame info/arrays that triggers an isosurface construction. Thoughts?

How does this look like in the chemiscope JSON format?

[rosecers]

I think we could put in some particle-level list that defines {"semiaxes": {"target": "atom", "values": Nx3 matrix}, "orientations" {"target": "atom", "values": Nx4 quaternion list}}, and then write a chemiscope backend that converts this into the isosurface data type for 3dmol.

For the record, the data format changes are minimal, we can save the necessary information into the ase frame and add a function that changes that data (e.g. ellipsoid semi-axes and orientations) into isosurface data within the chemiscope backend. I would suggest a particle-type flag in the frame info/arrays that triggers an isosurface construction. Thoughts?

How does this look like in the chemiscope JSON format?

[Luthaf]

I would put the data inside the structures object instead of the properties. Something like

{"structures": [
    {
        "size": 42,
        "names": ["H", "O", "C", "C", ...],
        "x": [0, 1.5, 5.2, ...],
        "y": [5.7, 7, -2.4, ...],
        "z": [8.1, 2.9, -1.3, ...],
        "cell": [10, 0, 0, 0, 10, 0, 0, 0, 10],
        // new data
        "orientation": [[1, 2, 3, 2], [1, 1, 1, 0], [-2, 1, 0, 1], ...],
        "semiaxes": [[1, 2, 3], [1, 1, 1], [-2, 1, 0], ...]
    },
]}

This data also feels quite specific to ellipses, right? For other kinds of shapes, would we need the same data?

Something more explicit like below could also be nice

{"structures": [
    {
        "size": 42,
        "names": ["H", "O", "C", "C", ...],
        "x": [0, 1.5, 5.2, ...],
        "y": [5.7, 7, -2.4, ...],
        "z": [8.1, 2.9, -1.3, ...],
        "cell": [10, 0, 0, 0, 10, 0, 0, 0, 10],
        // new data
        "shapes": [
            {"kind": "ellipsoid", "orientation": [1, 2, 3, 2], "semiaxes": [1, 2, 3]},
            {"kind": "sphere"}, // this would be the default if not provided
            {"kind": "ellipsoid", "orientation": [1, 2, -1, 2], "semiaxes": [1, 2, 3]},
            {"kind": "ellipsoid", "orientation": [1, 0, 0, 2], "semiaxes": [1, 0, -1]},
            ...
        ]
    },
]}

this would enable other kinds of shapes, and multiple types of shapes in the same dataset.

[ceriottm]

Very much like the idea - we could even hijack it to visualize arrows

…

On Fri, 14 Apr 2023 at 12:22, Guillaume Fraux ***@***.***> wrote: I would put the data inside the structures object instead of the properties. Something like {"structures": [ { "size": 42, "names": ["H", "O", "C", "C", ...], "x": [0, 1.5, 5.2, ...], "y": [5.7, 7, -2.4, ...], "z": [8.1, 2.9, -1.3, ...], "cell": [10, 0, 0, 0, 10, 0, 0, 0, 10], // new data "orientation": [[1, 2, 3, 2], [1, 1, 1, 0], [-2, 1, 0, 1], ...], "semiaxes": [[1, 2, 3], [1, 1, 1], [-2, 1, 0], ...] }, ]} This data also feels quite specific to ellipses, right? For other kinds of shapes, would we need the same data? Something more explicit like below could also be nice {"structures": [ { "size": 42, "names": ["H", "O", "C", "C", ...], "x": [0, 1.5, 5.2, ...], "y": [5.7, 7, -2.4, ...], "z": [8.1, 2.9, -1.3, ...], "cell": [10, 0, 0, 0, 10, 0, 0, 0, 10], // new data "shapes": [ {"kind": "ellipsoid", "orientation": [1, 2, 3, 2], "semiaxes": [1, 2, 3]}, {"kind": "sphere"}, // this would be the default if not provided {"kind": "ellipsoid", "orientation": [1, 2, -1, 2], "semiaxes": [1, 2, 3]}, {"kind": "ellipsoid", "orientation": [1, 0, 0, 2], "semiaxes": [1, 0, -1]}, ... ] }, ]} this would enable other kinds of shapes, and multiple types of shapes in the same dataset. — Reply to this email directly, view it on GitHub <#275 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAIREZ6QMBS664IFTG4BTALXBEQMXANCNFSM6AAAAAAW4KJ5AI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[rosecers]

Actively working on this, will let you know when we make progress...

[ceriottm]

I think we can now mark this as mission accomplished!