8000 Shell Mode by dbierdemand · Pull Request #215 · tracefinity/tracefinity · GitHub
[go: up one dir, main page]

Skip to content

Shell Mode - #215

Draft
dbierdemand wants to merge 12 commits into
tracefinity:mainfrom
dbierdemand:main
Draft

dbierdemand wants to merge 12 commits into
tracefinity:mainfrom
dbierdemand:main

Conversation

@dbierdemand
Copy link
Copy Markdown

No description provided.

Backend:
- New `shelled`, `shell_exterior_standard`, `shell_exterior_wall` options on
  BinParams (schemas.py) with validation; rebuild the bin as a
  constant-thickness shell via `_build_shelled_bin` in stl_generator_manifold.py:
  walls trace the outer perimeter and every tool outline, the top floor face
  is removed, and a 0.75mm trench floor plate (SHELL_TRENCH_PLATE_T) sits on
  the standard tapered base cells, leaving feet/grooves open
- Optional outer wall band (`shell_exterior_wall`): widens to the
  gridfinity-standard 2.6mm lip wall with a stacking lip
  (`shell_exterior_standard`); off, it runs the shell thickness up through
  the lip (non-standard stacking). Falls back to the solid bin when too
  short or too small for the wall thickness
- Cutout depth: `_max_pocket_depth` no longer deducts the 3.8mm lip notch
  for shelled bins (the lip collar is perimeter-only there); per-cutout
  depth overrides and insert height still apply
- Partial-bin cutters compose after the shell body

Frontend:
- Shell toggle, wall thickness slider (1–3mm), exterior wall and
  gridfinity-standard-exterior sub-options in BinConfigurator; new config
  fields in types/index.ts and binDefaults.ts
- calcMaxCutoutDepth drops the lip deduction for shelled bins
- DepthInput clamping max-depth propagation in BinEditorToolbar;
  maxCutoutDepth threaded through the bin editor page; 3D preview updates

Tests: new test_shell_mode.py; cutout-depth tests extended for shelled bins
Docs: shell mode in features.md, stl-generation.md, usage docs; screenshots
Export inserts in bin-space coordinates resting on their pocket floor
(wall_top_z - resolved pocket depth, which already includes insert_height)
instead of at z=0, so they drop into the pocket exactly where the tool
rests. Clip insert rings to the bin interior before the fit-clearance
shrink, matching the pocket cutter so inserts never overhang walls.

Subtract embossed text-label silhouettes from the insert's underside via
new _make_insert_text_cutters, so raised text inside a tool trace stays
visible through the insert; each cutter is offset outward by the insert
fit clearance (round join) to prevent binding on the letters.

In the 3D preview, share the main bin's bounding-box transform with the
insert STL instead of centering it on its own bounds, preserving its
true position inside the tool pocket.

Also round the Toggle knob/pill and fix label overflow/wrapping in
BinConfigurator, and remove unused docs screenshots.

Tests: pocket-floor z-range, per-polygon depth override, tool-position
x/y alignment, interior clipping of overhanging polygons, and text
cutter coverage.
trimesh's mesh.section polygons_full and ray-casting helpers, used by
the geometry test suite, require the optional rtree package. Install
it explicitly so the backend tests run on a fresh environment.

Also drop an unused manifold3d import and fix import ordering in
test_shell_mode.py (incidental ruff cleanup, no behaviour change).
CI's lint check runs `next build`, which rewrites next-env.d.ts to
reference ./.next/types/ instead of ./.next/dev/types/, causing the
git diff --exit-code check to fail. Commit the build-generated form.
Replace the hard-coded 0.75mm trench floor plate with a user-facing
Shell Depth slider under the Shell option. The slider expresses how
deep the open shell interior extends below the wall top (higher =
thinner floor plate = less filament) and uses the same range rule as
Cutout Depth: 1.5mm minimum, max = 1.5 + 7 x (height - 1), snapping
in 0.5mm steps.

- schemas: add shell_floor_plate (0.4-20mm) to BinParams; reset to
  0.75 when Shell is off
- generator: plate thickness from config via _shell_floor_plate_t;
  extend the too-short-for-shell guard to clear plate + 1mm; text
  label surface follows the plate
- routes: pass shell_floor_plate through to GenerateRequest (it was
  silently dropped, leaving the slider without effect)
- frontend: Shell Depth slider (inverted plate complement), Height
  and Shell toggle clamping, rename "Shell (less filament)" to
  "Shell"
- tests: plate thickness geometry, schema validation, route
  pass-through regression
Flat Bottom replaces the gridfinity feet with a flat, 45°-chamfered
underside so a bin can rest directly on a surface instead of a baseplate.
The grid footprint is unchanged (same cells, same 42mm pitch) — only the
underside differs.

- Backend: new `flat_bottom` on `BinParams`, plumbed through the STL
  request. `_base_top_z` returns 0 for a flat bottom, so walls run the full
  height, per-cell feet and shell base cells are skipped, and the shell
  trench floor plate becomes the floor (dropping the 0.01mm overlap that
  would otherwise open the underside). The solid path builds the chamf
8000
er in
  `_build_flat_body`; the shelled path subtracts the same wedge via
  `_make_flat_chamfer_cutter`. Partial-bin cutters and stability plates
  anchor to the new base top.
- The 4.75mm the feet occupied becomes usable pocket/shell depth, bounded by
  a 2mm solid floor (`MIN_FLAT_FLOOR_DEPTH`) in both `_max_pocket_depth` and
  the frontend's `calcMaxCutoutDepth`, so 1u bins gain real depth too.
- Magnet holes are forced off while flat bottom is on (there is no baseplate
  to lock to); half-grid base still applies, since it describes the grid
  layout rather than the feet.
- Frontend: Flat Bottom toggle in the Base section, wired through the bin
  page, types and factory defaults. The configurator is now six collapsible
  sections (size, cutouts, base, features, partial, print) with collapse
  state in localStorage, keeping features/partial/print folded by default;
  `SectionHeader` gains a `dense` variant for the tight sidebar.
- Tests: new `backend/tests/test_flat_bottom.py` covering the geometry and
  depth limits, plus `calcMaxCutoutDepth` flat-bottom cases; the numeric
  input e2e spec opens the folded Cutouts section first.
- Docs: new "Flat bottom" section in `docs/stl-generation.md` and updated
  height/depth/toggle references in `docs/usage/bin-configuration.md`.
Flat Bottom replaces the gridfinity feet with a flat, 45°-chamfered
underside so a bin can rest directly on a surface instead of a baseplate.
The grid footprint is unchanged (same cells, same 42mm pitch) — only the
underside differs.

- Backend: new `flat_bottom` on `BinParams`, plumbed through the STL
  request. `_base_top_z` returns 0 for a flat bottom, so walls run the full
  height, per-cell feet and shell base cells are skipped, and the shell
  trench floor plate becomes the floor (dropping the 0.01mm overlap that
  would otherwise open the underside). The solid path builds the chamfer in
  `_build_flat_body`; the shelled path subtracts the same wedge via
  `_make_flat_chamfer_cutter`. Partial-bin cutters and stability plates
  anchor to the new base top.
- The 4.75mm the feet occupied becomes usable pocket/shell depth, bounded by
  a 2mm solid floor (`MIN_FLAT_FLOOR_DEPTH`) in both `_max_pocket_depth` and
  the frontend's `calcMaxCutoutDepth`, so 1u bins gain real depth too.
- Magnet holes are forced off while flat bottom is on (there is no baseplate
  to lock to); half-grid base still applies, since it describes the grid
  layout rather than the feet.
- Frontend: Flat Bottom toggle in the Base section, wired through the bin
  page, types and factory defaults. The configurator is now six collapsible
  sections (size, cutouts, base, features, partial, print) with collapse
  state in localStorage, keeping features/partial/print folded by default;
  `SectionHeader` gains a `dense` variant for the tight sidebar.
- Tests: new `backend/tests/test_flat_bottom.py` covering the geometry and
  depth limits, plus `calcMaxCutoutDepth` flat-bottom cases; the numeric
  input e2e spec opens the folded Cutouts section first.
- Docs: new "Flat bottom" section in `docs/stl-generation.md` and updated
  height/depth/toggle references in `docs/usage/bin-configuration.md`.
The merge that brought in the flat-bottom commit resolved
next-env.d.ts in favor of the dev variant (./.next/dev/types/...),
which is what next dev writes. CI regenerates the file with
next typegen, which emits the non-dev paths, so the checked-in file
must match that or the git diff --exit-code step fails.
All four tests in the spec failed in expandCutouts() before reaching any
of their own assertions. The helper clicked the Cutouts header
unconditionally and then expected aria-expanded="true".

The section is not folded by default: DEFAULT_COLLAPSED in
BinConfigurator.tsx sets cutouts: false, so a new bin opens with Cutouts
already expanded. The click therefore collapsed it and flipped
aria-expanded to "false", which the assertion could never observe. The
comment claiming the section "starts folded" was the source of the
misdiagnosis.

This is not a timing problem. toHaveAttribute already polls until the
expect timeout expires, so adding an explicit timeout or a
MutationObserver wait would not have changed the observed value.

Make the helper idempotent: only click when the section is actually
folded, then assert the expanded state. This also covers the case where
a previous test in the file left a different collapse state, since
collapsedSections is persisted in localStorage and shared across the
tests in this run.
next typegen emits .next/types/*.d.ts paths in CI (production
typegen), while the checked-in file referenced .next/dev/types/*,
which the linter job detects via git diff --exit-code.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

triage/needs-issue PR needs a linked issue triage/stale No activity for 30+ days

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants

0