MAINT: register the vtk.js renderer as a 3D backend

Replaces the _activate monkeypatching with a real "jupyterlite" backend so it
works with the pytest renderer fixtures. Also fixes the glyph templates, which
drew spheres at twice their size and EEG cylinders off their own axis.

Author: natinew77-creator

doc/changes/dev/14144.other.rst

@@ -1 +1 @@
-Add a vtk.js drawing backend for MNE's 3D renderer in ``mne/viz/backends/_lite.py``, used by the JupyterLite documentation where VTK cannot load, by `Natneal B`_.
+Add a vtk.js drawing backend for MNE's 3D renderer, selected with ``mne.viz.set_3d_backend("jupyterlite")`` and used by the JupyterLite documentation where VTK cannot load, by `Natneal B`_.

doc/sphinxext/jupyterlite_lite_renderer.py

@@ -12,12 +12,12 @@
 # Copyright the MNE-Python contributors.
 
 LITE_RENDERER_CELL = """
-# Using pyvista-js (vtk.js) to draw MNE's 3D rendering in JupyterLite.
+# Draw MNE's 3D figures with pyvista-js (vtk.js): VTK has no WebAssembly build.
 # See mne/viz/backends/_lite.py for more details.
 try:
-    from mne.viz.backends._lite import _activate as _mne_activate_lite_renderer
+    import mne.viz
 
-    _mne_activate_lite_renderer()
+    mne.viz.set_3d_backend("jupyterlite")
 except Exception as _e:
-    print("[JupyterLite] could not install the pyvista-js renderer: " + repr(_e))
+    print("[JupyterLite] could not select the pyvista-js renderer: " + repr(_e))
 """

mne/conftest.py

@@ -755,6 +755,22 @@ def renderer_notebook(request, options_3d):
         yield renderer
 
 
+@pytest.fixture(params=[pytest.param("jupyterlite", marks=pytest.mark.pvtest)])
+def renderer_lite(request, options_3d):
+    """Yield the JupyterLite (vtk.js) renderer."""
+    from mne.viz.backends import renderer as renderer_module
+
+    # use_3d_backend only puts a backend back if one was already selected, and
+    # this one draws for a browser, so make sure it is never what a later test
+    # inherits
+    was = (renderer_module.MNE_3D_BACKEND, renderer_module.backend)
+    try:
+        with _use_backend(request.param, interactive=False) as renderer:
+            yield renderer
+    finally:
+        renderer_module.MNE_3D_BACKEND, renderer_module.backend = was
+
+
 @pytest.fixture(params=[pytest.param("pyvistaqt", marks=pytest.mark.pvtest)])
 def renderer_interactive_pyvistaqt(request, options_3d, qt_windows_closed):
     """Yield the interactive PyVista backend."""
@@ -798,6 +814,10 @@ def _use_backend(backend_name, interactive):
 def _check_skip_backend(name):
     from mne.viz.backends._utils import _notebook_vtk_works
 
+    if name == "jupyterlite":
+        # draws with vtk.js in a browser: no VTK, no Qt, no ffmpeg
+        pytest.importorskip("pyvista_js")
+        return
     pytest.importorskip("pyvista")
     pytest.importorskip("imageio_ffmpeg")
     if name == "pyvistaqt":

mne/viz/_3d.py

@@ -841,7 +841,7 @@ def plot_alignment(
 
     # initialize figure
     renderer = _get_renderer(
-        fig,
+        fig=fig,
         name=f"Sensor alignment: {subject}",
         bgcolor=(0.5, 0.5, 0.5),
         size=(800, 800),
@@ -3931,7 +3931,7 @@ def snapshot_brain_montage(fig, montage, hide_sensors=True):
         )
 
     # initialize figure
-    renderer = _get_renderer(fig, show=True)
+    renderer = _get_renderer(fig=fig, show=True)
 
     xyz = np.vstack(xyz)
     proj = renderer.project(xyz=xyz, ch_names=ch_names)