Skip to content

M284-human-review.md

Latest commit

 

History

History
144 lines (92 loc) · 6.37 KB

M284-human-review.md

File metadata and controls

144 lines (92 loc) · 6.37 KB
Raw

M292 requalified owner review

M292 requalified this portable review pack after the owner's M284 findings and the Datoviz binding safety repairs. All fourteen captures were regenerated from exact committed wheels outside both source trees and are 800×600. Camera raster geometry now differs by at most 1.2% in either dimension. Both backends preserve the same semantic scene and canvas, but are not expected to match pixels.

Static captures

The captures use direct repository-relative image references. Each preview also has a plain full-size link for Markdown viewers that do not display inline images.

Priority 2D

Matplotlib (open full size)

Matplotlib priority 2D

Datoviz (open full size)

Datoviz priority 2D

Perspective 3D

Matplotlib (open full size)

Matplotlib perspective 3D

Datoviz (open full size)

Datoviz perspective 3D

Orthographic 3D

Matplotlib (open full size)

Matplotlib orthographic 3D

Datoviz (open full size)

Datoviz orthographic 3D

Camera fit

Matplotlib (open full size)

Matplotlib camera fit

Datoviz (open full size)

Datoviz camera fit

Camera orbit

Matplotlib (open full size)

Matplotlib camera orbit

Datoviz (open full size)

Datoviz camera orbit

Camera pan

Matplotlib (open full size)

Matplotlib camera pan

Datoviz (open full size)

Datoviz camera pan

Camera zoom

Matplotlib (open full size)

Matplotlib camera zoom

Datoviz (open full size)

Datoviz camera zoom

Static acceptance

  • 2D composition is legible and complete on both backends.
  • 3D perspective and orthographic composition is legible and complete on both backends.
  • Fit, orbit, pan, and zoom states are distinct, coherent, and unclipped.
  • Pixel visuals have the intended positions and relative logical sizes.
  • Sphere visuals are distinct and correctly placed.
  • Vector visuals preserve direction and relative magnitude.
  • Gallery 3's uniform primitive preserves topology without implying interpolation parity.
  • Gallery 3's pixel anchors are visibly distinct from primitive vertices.
  • Text is legible, separated, and correctly associated with the scene.
  • Mesh geometry and depth are credible on both backends.

Known adaptations: both backends receive a pixel-exact 800×600 canvas. Matplotlib uses adapted painter/projection paths and suppresses its unintended native View3D frame. Matplotlib preserves semantic panel titles. The qualified Datoviz binding has no public PanelTextGuide renderer, so its missing title is explicitly diagnosed as unsupported. Fonts, metrics, antialiasing, vector heads, and billboard placement may differ. Datoviz raycast spheres use native shading and analytic surface depth, while Matplotlib spheres are flat projected circles with adapted depth ordering; this is not a sphere material contract. Matplotlib 3D vectors/pixels/text are adapted overlays, and neither backend claims strict pixel parity or billboard occlusion parity.

Live Datoviz review

The exact-wheel run is already recorded in the automated qualification evidence. For the portable owner interaction check, copy this repository-relative command from the Mission Control GSP_API checkout. It changes into the sibling VisPy2 checkout itself:

cd ../vispy2 &&
GSP_DATOVIZ_SOURCE=../datoviz \
GSP_DATOVIZ_ENABLE_EXPERIMENTAL_VIEW3D_NAV=1 \
PYTHONPATH=src:../datoviz \
../gsp/.venv/bin/python examples/gallery_05_datoviz_navigation.py

Controls: left-drag orbits, right-drag pans, the wheel zooms, and double-click resets the camera. Close the native window to end the blocking loop and release the context-managed session. If the window cannot be closed, focus the terminal and use Ctrl-C; verify that the process exits.

  • Live orbit, pan, zoom, and reset controls respond naturally.
  • Closing the window cleans up the process and native resources.

Query and capability review

The installed-wheel checks produced a point HIT with caller identity preserved and a structured UNSUPPORTED result for the deliberately unsupported sphere/3D request. Review the exact-head capability matrix for the bounded contracts and adaptations.

  • Point HIT behavior is sufficient for the first experimental release.
  • Structured UNSUPPORTED behavior is clear and honest for unsupported queries.

Owner decision:

  • Accept S065 experimental feature coverage.
  • Request bounded corrections before accepting S065.

Accepted by the owner on 2026-07-28. The live camera controls and cleanup work as intended. The uniform unlit Gallery 5 mesh was accepted as functionally correct but visually unattractive. The requested presentation follow-up subsequently landed: Gallery 5 now uses the bounded cross-backend flat-Lambert path without expanding the material model.