Conversation
…ctions with transform expansion pass
…ured dimensions, highlights and entity metadata
…dded monospace font and richer agent metadata
…pector, layer ghost states and 3D stacked view
…ion labels and add JSON project reports
… JSON flags into the CLI
…d styled dimensions
…the taller example
…nonical spacing, comment preservation and a semantic safety check
…tity styles, and export dim offsets perpendicular to the measured axis
Feature/refine first version
…ews and derived plans wired into the live preview
…GitHub repo URLs kept until the repo is renamed)
…the same mesh builders as the 3D view
…ave recompiles and live-reloads)
…lowing light/dark theme and a built-in .cf editor
…(orbit, theme-aware, live-reload)
…e records compile_project registered every layer with ACI 7 (white) and compile_cf re-added it with the actual color; the dxf crate appends layer records, so the output contained two LAYER table entries per layer and external CAD readers picked the white one. Layers are now registered once, after loading the .cf files, using each layer's [layer].color. DxfWriter::add_layer is now an upsert: re-adding an existing layer updates its color instead of duplicating the table record. hex_to_aci previously matched only 9 exact hex strings and silently mapped everything else to white, so common colors like #FF4444 or #00CC44 lost all color in the exported DXF and on import roundtrip. It now maps any parseable color to the nearest entry of the standard ACI palette. The roundtrip integration test now uses a non-palette layer color and asserts the layer table has no duplicate records.
The build summary ("DXF generado", "entidades en capas") was the only
Spanish output left in an otherwise English CLI. Align it with the rest of
the command surface and the docs.
…rt --output The CLI reference and guides were missing flags that exist on the real surface: serve's background/daemon behavior (--foreground, --stop), preview --format all and --plano (drawing sheets declared as [[plano]] in project.toml), and import --output.
The [[plano]] project.toml feature (named views on paper-sized sheets with title blocks, including section cuts) had no user documentation beyond the new CLI reference row. Document the declaration keys, the available views, and the preview --plano workflow, all verified against the real CLI.
The build pipeline never emitted HATCH or MTEXT entities: filled regions compile to fan-triangulated SOLID entities, hatch patterns to clipped LINE segments (the dxf crate has no native HATCH), and annotations to TEXT. Update README and the building-and-export guide to describe the entities actually written.
…l boundaries A hatch or fill whose `boundary` id did not match a closed polyline or rect in the same layer file was skipped with no diagnostic, so a filled region could silently vanish from the DXF. Emit a stderr warning naming the entity, its layer, and the missing boundary id; the build still succeeds. Boundaries remain resolved per layer file.
The fl-closet-d1 fill referenced boundary rc-closet-d1, which is defined in muros.cf; since boundaries resolve per layer file the reference never matched and the fill was skipped (no SOLID in the DXF, no fill in the preview). Use inline points matching that rect so the closet renders as a filled region in both DXF and preview.
The importer skipped SOLID entities as unsupported, so a build → import round-trip lost every filled region. Import SOLIDs as fill polygons and fuse edge-adjacent ones per layer back into a single `[[fill]]`: the fan-triangulated triangles `cadspec build` emits are stitched into their original boundary by dropping shared interior edges and chaining the remaining boundary edges into one ring. Disjoint fills stay separate, and a ring that can't be stitched cleanly falls back to per-triangle fills. Recovers fill color/lineweight and keeps the entity count stable across round-trips.
A `[[hatch]]` expands into many clipped DXF LINE entities (the DXF format this tool targets has no native HATCH entity). A build→import round-trip used to bring every one of those back as a separate `[[line]]`, inflating one source hatch into dozens of stray lines. Stamp each generated pattern line with `CADSPEC_HATCH` XDATA carrying a group id plus the source region, pattern, angle and scale. On import, lines sharing a group id collapse back into a single `[[hatch]]` (using inline points), while untagged lines stay `[[line]]` — so a foreign DXF's ordinary lines are never mistaken for hatch pattern (no false positives). `[[hatch]]` now accepts inline `points` as an alternative to a `boundary` id, mirroring `[[fill]]`; the compiler and SVG renderer resolve a hatch region from either source. Verified stable across build→import→build on the vivienda example (2 hatches + 4 real lines round-trip unchanged).
…d svg Boundaries were already resolved per layer file in build (warns and skips since 209d345), but preview/serve rendering and the svg module resolved the same boundaries silently. Reuse the compiler's warning at the two render_layer sites (fills, hatches) so the same condition is visible regardless of which command triggered the render, and document that per-file resolution is the intended design (cross-layer references use inline `points`, not a project-wide boundary index).
D8 gap: fmt only had unit tests over inline TOML snippets, never an end-to-end pass over a real multi-file project. Copy examples/vivienda's sources into a tmp dir and drive format_project directly: verify the first pass, that the result still compiles to a valid DXF, that --check reports nothing outstanding, and that a second pass is a byte-for-byte no-op across every file.
D8 gap: planos.rs only had 2 unit tests exercising render_plano directly with a hand-built Plano struct; nothing drove the CLI-facing generate_plano, which looks a plano up by name in project.toml. Declare a [[plano]] on a copy of examples/vivienda and verify the happy path (preview.png/svg written, title block content present) and that an unknown name fails with a helpful error instead of panicking.
serve_project blocks forever in listener.incoming() and watch_project blocks forever in rx.recv(), so both are exercised by spawning them on a (deliberately leaked) thread and driving them from the outside: raw TCP/HTTP against an OS-assigned port for serve (GET /state, /preview.svg, 404, and a rebuild triggered by editing a .cf file), and a filesystem edit followed by polling for output.dxf for watch. Closes out D8.
hex_to_aci and aci_to_hex only knew about the 9 standard ACI colors, so any layer color outside that small set (or a DXF layer index above 9 on import) fell back to white. Both now use the full AutoCAD Color Index table (indices 1-255), giving a much closer nearest-color match on export and exact recovery of more colors on import. aci_to_hex now returns an owned String since it looks up entries in a 255-color const table instead of matching a handful of literals.
The writer stamped hatch pattern lines with XDATA under the CADSPEC_HATCH application name but never declared that name in the DXF's APPID table. cadspec's own importer tolerated this (it reads XDATA regardless of table registration), but strict DXF readers are expected to ignore XDATA from unregistered apps. DxfWriter::new now registers CADSPEC_HATCH via drawing.add_app_id, matching how AutoCAD's own reserved app names are already ensured present.
config_path() reads $HOME directly, so config_set/config_show had zero test coverage (testing them would have written into the real user's ~/.cadspec/config.toml). Split the path resolution out into config_set_at/config_show_at, taking an explicit path like the rest of the codebase's testable functions (fmt::format_project, scaffold::create_project, etc.), and added tests against scratch paths under /tmp.
Resolve the three D13 issues with the `serve` web viewer: - Branding (13a): add an inline SVG favicon served at `/favicon.svg` and linked from the page head, so the tab has a "cs" monogram instead of a blank icon. The "cadspec live" wordmark already lived in the header. - Themes (13b): drop the three-way theme toggle. The SVG background is hardcoded dark, so the light theme only ever recolored the chrome and read as a redundant second dark. Remove the toggle button, the pre-paint init script, the `[data-theme="light"]` and `prefers-color-scheme` CSS, and the `cadspec-theme` localStorage key, leaving a single dark theme. - Editor auto-reload (13c): the built-in .cf editor now resyncs from disk when a file changes externally (another editor, git checkout). On every SSE tick `refresh()` calls a new `__reloadEditorIfClean` hook that reloads the current file unless there are unsaved edits or the user typed within the last 2s while focused, so in-progress edits are never clobbered. Tests: unit tests assert the served HTML has the favicon/branding, no theme-toggle remnants, and the reload hook; the serve e2e test now checks `/favicon.svg` returns 200 with SVG content.
docs: initial documentation, glTF export, editor, and binary rename
chore: fix ci — track Cargo.lock + bump version
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Release: all develop changes merged to main.