Skip to content

layout_override covers few declarations: side only for frames and datums, lane for 2 of 41 dimensions #2211

Description

@pzfreo

I probed Sheet.validate_layout_override on rc9 for every declaration in two Specify Sheet scripts (emitted by generate_sheet_script(..., pmi='annotate'); parts in https://gist.github.com/pzfreo/7c6a0be95a567e2774924207133b2d40):

  • side=…: supported only for the ControlFrame and DatumRef declarations (spindle 3 of 16, plate 6 of 38). Hole callouts, steps, pockets, slots, envelope dimensions and authored dimensions all return unsupported_declaration ("does not expose a supported side control").
  • parameter=…, lane=…: supported for 2 of 41 declared dimensions across both parts. The rest return:
    • "dimension 'step.diameter' on step does not expose a lane control" (also od.diameter, pocket_width.length, height.length, …)
    • "dimension 'location' is declared more than once; remove the conflicting intent before overriding its…"

The location refusal applies whenever a feature has several location declarations (one per axis or member), which is the emitter's own normal output.

A prototype click-to-edit UI over rc9 works for what is supported: moving datum B to the left side rebuilds with 0 errors. But most things a user would want to nudge can't be moved.

Asked for:

  • Side control for feature callouts (hole notes, diameter callouts) and for authored dimensions.
  • Lane control for the common dimension kinds (step and od diameters, envelope lengths, pocket and slot sizes).
  • location overrides addressable per axis or member, so that multiple declarations aren't a conflict.

All of these stay inside the bounded, coordinate-free vocabulary.

Activity

  1. pzfreo commented on Oct 6, 2026

    @pzfreo
    OwnerAuthor

    PR #2216 is the first end-to-end editor-control slice: exact declared hole/pattern callouts and envelope height can now choose a supported side. It checks rendered movement, emitted-script replay and report/schema evidence. The issue remains open: step/OD/pocket/envelope lane actuators and per-axis/member location selection are not implemented; exposing those as options before the renderers honor them would be misleading.

  2. pzfreo commented on Oct 6, 2026

    @pzfreo
    OwnerAuthor

    PR #2218 is stacked on #2216 and adds editor side control for source-authored measured dimensions with an unset current side (normal emitted-script case). It validates the dimension-kind/projection vocabulary and checks rendered movement, replay, and reporting. I am not auto-merging #2218 into its base branch while #2216 CI runs; after #2216 lands I will retarget #2218 to main and gate its merge on its own green checks. Lane actuators and per-axis/member location selection remain open.

  3. pzfreo commented on Oct 6, 2026

    @pzfreo
    OwnerAuthor

    Progress on bounded, coordinate-free editor controls:

    The issue remains open: step/OD diameter lanes, pocket size lanes, feature-callout sides not already covered, and per-axis/member addressing for repeated location declarations still need implementation. The stacked PRs are not being merged into their base branch while #2216 CI is running; they will be retargeted to main after it merges.

  4. pzfreo commented on Oct 7, 2026

    @pzfreo
    OwnerAuthor

    Added #2220, stacked on #2219: explicit lanes for rectangular-pocket width/length now draw the selected size as a measured linear dimension and remove that measurement from the combined pocket leader, so the editor control has a real effect without duplicate values. The existing no-lane pocket leader remains unchanged. The issue stays open for step/OD diameter controls, remaining feature callout sides, and per-axis/member location addressing.

  5. pzfreo commented on Oct 7, 2026

    @pzfreo
    OwnerAuthor

    Exact Z-normal hole/pattern X/Y location lanes are now in stacked PR #2221. The override requires an axis and member, retains independent compiler/replay/report identity, and uses the whole-ink solver with explicit drop evidence. Focused local regressions pass; CI is pending. Earlier bounded side/envelope/pocket work remains in #2216, #2218–#2220. This issue stays open until the stack lands and remaining supported families are adjudicated.

  6. pzfreo commented on Oct 7, 2026

    @pzfreo
    OwnerAuthor

    Added stacked PR #2222 for the overall height lane, on top of #2219. A real drawing shows lanes 1 and 2 move the full height dimension independently, and a missing lane solve reports the exact requested rank/blocker. Focused preflight and 100% changed-line coverage pass; CI is pending. The issue remains open for step/OD diameter lanes and remaining feature-callout sides, plus merge verification of the current stack.

  7. pzfreo commented on Oct 7, 2026

    @pzfreo
    OwnerAuthor

    Rotational OD diameter lane control is now in #2223. It preserves the existing no-override mark, sends explicit lanes through the shared measured-ink solve, and reports an unsatisfiable requested lane rather than silently moving it. Local tests cover Z/X/Y orientations and refusal; CI is running. #2211 remains open for step diameters, remaining callout-side controls, and stack merge verification.

  8. pzfreo commented on Oct 7, 2026

    @pzfreo
    OwnerAuthor

    Step-diameter lane controls are now in stacked PR #2224 (on #2223). X/Z profile leaders and Y end-on leaders use exact bounded ranks in the shared leader solve; no-override paths remain unchanged. It also preserves Y diameter coverage and reports an infeasible requested lane with measurement evidence. Focused local tests, full static checks and strict docs pass; the broader related slice and CI are running. The issue remains open for remaining callout-side controls and stack merge verification.

  9. pzfreo commented on Oct 7, 2026

    @pzfreo
    OwnerAuthor

    Step-diameter side control is now in stacked PR #2225 (on #2224): X steps offer above/below, Z steps left/right, and Y end-on steps all four sides. The exact requested hemisphere is enforced by the shared leader solve, with measurement-linked refusal when infeasible; normal no-override routing is unchanged. Focused local behavior, routing, static, docs, and source-shape checks pass; CI is pending. #2211 stays open for the remaining callout-side families and stack merge verification.

  10. pzfreo commented on Oct 7, 2026

    @pzfreo
    OwnerAuthor

    Boss OD diameter callout sides are now in stacked PR #2226 (on #2225), using the same bounded side semantics as step diameters. Real X/Y/Z drawings and focused source-shape/static preflights pass locally; CI is pending. #2211 remains open until the supported callout families and whole stack are verified merged.

  11. pzfreo commented on Oct 7, 2026

    @pzfreo
    OwnerAuthor

    Further bounded editor controls are now in review: pocket depth-leader sides in #2227 and independent slot width/length sides in #2228. Both use the existing solver/corridors, preserve exact declaration identity, and locally pass focused placement and source-shape checks. They remain stacked behind the earlier #2211 slices; this issue stays open until the chain lands.

  12. pzfreo commented on Oct 7, 2026

    @pzfreo
    OwnerAuthor

    Progress review for #2211 (not closing yet): the bounded editor-control PRs are prepared as one linear stack, starting at #2223 and ending at #2221. Their rebased combined tip passes 85 focused/source-shape tests, 184 architecture/routing guards, mypy, Ruff, and strict docs build; the rewritten heads are now published as GitHub stack #2229 on merged #2217. Fresh PR CI is queued or running; those local results are not a substitute for it.

    Coverage is declaration/parameter-specific, not a blanket promise that every mark can move: rotational OD and step-diameter lanes; step/boss diameter sides; pocket depth-callout side; independent slot size sides; authored PMI-dimension sides; envelope, pocket-size, overall-height, and exact hole/pattern location lanes. A hole's compound note is addressed through its declared bore.diameter side (existing renderer path), rather than a generic feature-level side. The shared solve still decides feasibility and must report a rejected placement. Measurements without a renderer-backed control are intentionally not advertised. I will verify the final PR diffs and CI before adjudicating closure.

  13. pzfreo commented on Oct 7, 2026

    @pzfreo
    OwnerAuthor

    Final status: green atomic stack #2229 is merged on main (#2223–#2228, #2218–#2220, #2222, #2221), and main's tree matches the tested combined tip. The requested bounded controls now cover hole callout sides (through declared bore.diameter), other feature and authored-PMI sides, OD/step/envelope/pocket/height lanes, and exact X/Y location lanes by axis/member. Slot width/length lanes were already supported. Capability queries still do not promise geometric fit: an infeasible requested corridor reports a placement drop. No raw-coordinate edit path was added. #2208 remains open for separate per-view relayout work.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions