- The per-character width estimate is 0.66 text heights, measured by
rendering the text of 106 saved dimension pictures with the bundled
fonts; the old 0.60 left every fit and hook estimate ten percent narrow.
With it, the "text plus DIMGAP must clear the arrowheads" rule reproduces
AutoCAD's arrow placement for 1047 of 1078 pictured linear dimensions
whose text sits between the extension lines.
- DIMTMOVE 1 stores where the leader's hook starts, not where it ends: on 39
such radial dimensions the stored point sits DIMASZ plus DIMGAP before the
text. The hook now runs from the stored point one arrow toward the text.
- The diameter prefix is U+2205, matching the %%c expansion and the pictures.
- A diameter whose text the user dragged keeps the DIMTAD lift from #1323;
the four AutoCAD-authored diameters with DIMTAD above in the sample all
store the lifted point directly, so this applies only to dragged text.
Saved dimension pictures across 1311 diameter dimensions carry U+2205 (806) or a
literal %%c (168); U+2300 never appears. The bundled romans and simplex fonts
have U+2205 and lack Latin O-stroke, so one character now serves the %%c
expansion and, in the dimension formatter, the diameter prefix.
The fit and break estimates used two halves at STACK_HALF_SCALE; the stack
is laid out from the lower half's baseline to the upper half's cap, which is
STACK_RAISE + STACK_HALF_SCALE of the run height. The taller estimate flips
arrows outside for a few more stacked texts that AutoCAD drew that way.
A negative DIMLFAC scales a paper-space dimension only when it measures
model space through a viewport. AutoCAD records the factor it used as
ACAD_DIMASSOC_CALC_DIMLFAC; a zero there marks a dimension of paper-space
geometry, which reads its sheet distance unscaled. Regenerating such a
dimension multiplied it by the factor anyway, turning 4.457 into 18.110.
Text placed past an extension line along the dimension's own axis pulls the
dimension line out to its near edge rather than stopping at the two-arrow
stub. A radial leader already within 15 degrees of horizontal runs straight to
the text instead of turning a hook, and a diameter with outside text draws its
single leader for every DIMTMOVE except 2, which keeps only the arrowhead.
Subset parity: 870 of 1076 pictured dimensions now regenerate as drawn.
Comparing the saved picture block of every dimension in a set of production
drawings with the geometry regenerated after a refresh showed half of them
differing. The differences trace to a few layout rules:
- The stored text middle point (group 11) is where the text was drawn, for
automatic and hand-placed text alike; a regenerated dimension keeps it and
only a cleared point is placed from the style.
- Arrows move outside when the text, where it is drawn, would run into an
arrowhead. The text is measured along the dimension line, so horizontal
text on a vertical dimension counts its height, a limits stack its two
lines and `\P` continuation lines their number.
- Text pushed outside starts past the two-arrow stub and a gap; arrow
length is DIMASZ, not the text height.
- Horizontal text outside a non-horizontal dimension gets a leg to its
height and a hook one arrow long that it sits against, both when drawn
and when auto-placed.
- A radius with outside text draws no line to the centre unless DIMTOFL,
a leader to a hook the text sits against, and an arrowhead whose body
points toward the text. A diameter with outside text and no DIMTOFL is a
single leader with one arrowhead, no line across the circle.
- DIMTMOVE 1 stores the hook end; the leader is drawn to it exactly.
- Width estimates measure the widest `\P` line and skip MText format codes.
On a 30-drawing subset with 1076 pictured dimensions the regenerated
drawing now matches the picture for 838 of them, up from 584.
Production drawings show AutoCAD stacking two equal limits for a DIMLIM
style with zero offsets; the plain-value shortcut matched one drawing from
another authoring tool and broke twenty-five that AutoCAD wrote.
AutoCAD prints 0.4255 as 0.426; Rust's formatter rounds the binary value,
which sits a hair below the decimal tie, and printed 0.425. Decimal
dimension text, tolerances and limits now settle the tie the way the
drawings do.
Alternate units multiplied the raw measurement instead of the displayed
DIMLFAC-scaled value, so a paper-space dimension's bracketed value was off
by the viewport factor, and alternate limits were written high/low with a
slash where AutoCAD stacks them like the primary limits.
A limits style whose upper and lower offsets are both zero prints the plain
value in AutoCAD, not a stack of two equal numbers. A three-point angular
dimension whose extension point sits on its vertex has no measurable angle;
AutoCAD keeps showing the angle it stored, and the text now does too instead
of reading 0 degrees. Both cases came out of comparing saved picture blocks
with regenerated text across a set of production drawings.
Dimensions with DIMLIM on were regenerated as the nominal value plus a
separate stacked pair, and the pair was built from the unscaled
measurement. AutoCAD prints only the upper limit over the lower one, with
the offsets applied to the DIMLFAC-scaled, DIMRND-rounded value, and keeps
both halves at the same number of decimals under DIMTZIN. Drawings with a
limits-only style looked right on load, from their saved picture blocks,
and gained an extra number as soon as an associative refresh dropped the
picture.
Fit and fill estimates now count visible cells, so a stack is as wide as
its wider half at DIMTFAC, and the text-to-line offset uses the stacked
height.
Grid snap locks to SNAPUNIT instead of magnetizing; snapped 3D points keep elevation instead of flattening to z=0. New SnapTypes Vertex/EdgeMidpoint/FaceCenter/Knot/FacePerpendicular/NearestFace with their own F4 master toggle and mode set, tiers, and markers, fully independent of the 2D master. Solid B-rep corners/edge-midpoints/face-centers from mesh edges and kernel faces, spline knot points, and live perpendicular/nearest-to-face on mesh triangles merged priority-aware in move/click paths. Drafting Settings 3D Object Snap tab is fully live (was dead placeholders); Nearest-to-face off by default. TDD: behavioral LINE-to-solid tests plus engine/scene unit tests.
- IMAGEEMBED: pack raster images into OLE2FRAME entities so a .dwg is
self-contained; 'Embed Image' entry in the XREF attach menu; control/MCP
op embed_image for external tools; dxf_name for OLE2FRAME queries.
- Scene: populate_images_from_document now rebuilds OLE2FRAME and PDF
underlay models too (they used to vanish after any XREF op).
- Printing: add PRINTER_ENUM_NETWORK pass when listing printers; GDI
spooler fallback (hayro raster at 300 DPI + StretchDIBits) for hosts
whose PDF handler registers no print verb; DEVMODE orientation with a
90-degree raster rotation fallback for v4 drivers; fix the fit-scale
precedence bug that printed blank pages.
- PDF export dialogs on Windows use the async file dialog so the native
overwrite confirmation no longer deadlocks the UI thread.
- Plot dialog: persist the chosen printer (saved on selection) and the
picked plot window across sessions; page setups with no device no
longer reset the output device.
- Community fonts: detect unresolvable .shx references on open and offer
a download from the repository's fonts/ folder (contribution guide in
fonts/README.md); custom font source URL setting for intranet font
servers; downloaded fonts resolve from the per-user fonts directory.
Refresh associations after complete SpaceMouse camera updates and recognize imported paper-space owners without Layout objects. Use kernel containment, intersection, perpendicular and spline tangent solving; pin cadkernel 7ce27f2 and add regressions.
DIMLINEAR / DIMALIGNED / DIMRADIUS / DIMDIAMETER / DIMANGULAR created on a
sheet now report the model measurement of the geometry seen through a
viewport, at any viewport scale, while definition points, the dimension line
and the text stay in paper coordinates and the annotation keeps its sheet
size (DIMSCALE is untouched).
A viewport frame is a similarity - uniform scale plus in-plane twist - so
projecting the model span onto the paper axis and compensating gives exactly
the raw paper measurement times 1/scale. No geometry is rewritten: the
compensation is persisted once as a negative DIMLFAC override on the
dimension and applied once by the linear formatter, so preview text,
dimension text, the properties measurement row and DXF/DWG export all agree.
User measurement multipliers fold into the same magnitude, and an explicit
text override still bypasses formatting entirely, so it is never re-scaled.
Angular dimensions are never compensated - a similarity preserves angles.
Mixed input (one point through viewport A, another through viewport B or on
the sheet) is refused rather than silently mis-measured: the command line
explains why and the dimension falls back to plain paper measurement.
A dimension measured through a viewport is left unassociated: its definition
points are projected paper coordinates, so inferring a paper-space
association would bind it to geometry that is not there. PR3 records the real
dimension -> viewport -> entity chain.
The result is one ordinary dimension entity carrying one style override, so
undo/redo works through the existing history with no new out-of-document
state.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
DXF gives DIMLFAC a sign convention: a positive value scales every
dimension, a negative one applies |DIMLFAC| only to dimensions that live in
a paper-space layout (where it compensates a viewport's scale). The linear
formatter multiplied by the raw signed value, so a drawing with negative
DIMLFAC (e.g. 115448.dxf) rendered negative measurements.
Resolve the sign once, in `resolved_dimension_style`, the single place an
effective style is derived, so dimension text, alternate units, exploded and
baked text and the properties panel all agree. `extended_data` keeps the raw
signed value, so DXF/DWG export is unchanged. The properties "Measurement"
row now reports the value the text shows rather than the raw distance, while
`actual_measurement` (group 42) stays raw as the format requires.
Also adds the PR2 seams the dimension work builds on:
* MeasurementScale::user_lfac_for_space / viewport_dimlfac_override —
the persisted representation of viewport compensation.
* Scene::viewport_frame and paper-point viewport resolution.
* Model-geometry hit testing through a content viewport, including
descent into block instances with the instance transform baked in.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
A linear or aligned dimension whose normal is not world Z was drawn and
exploded as if it lay in world XY. The builders read the offset and the
extension directions in XY only, so a dimension turned into a working plane
such as XZ lost its offset (now along Z): the dimension line fell onto the
measured points and the extension lines stuck out along world Y.
Move such a dimension into its own OCS before building it and map the result
back onto its plane, as arc length and large radial dimensions already do,
for both tessellation and EXPLODE.
Co-Authored-By: Claude Opus 5 <[email protected]>
Optimize wide and tapered arc fill mesh generation by eliminating general
ear-clipping triangulation and optimizing arc boundary evaluation:
1. Replaced triangulate_planar with triangulate_band_ring and direct
quad-strip vertex construction in wide_band_tris. Polyline band fills
possess strict quad-strip topology (straight segments: 4-vertex trapezoid;
arc segments: 2m-vertex annular sector). Generating triangles directly in
O(N) time avoids Newell normal calculation, planar projection, Delaunay
triangulation, and O(N^2) ear-clipping searches.
2. Exact buffer capacity pre-allocation in wide_band_tris based on ring
lengths, eliminating dynamic reallocation during triangle emission.
3. Replaced top and bottom cap triangulation in thick_band_tube with
triangulate_band_ring.
4. Unified inner and outer arc point calculation in polyline_segment_fill
into a single loop using ang.sin_cos(), halving trigonometric calls and
eliminating dynamic vector resizing.
Performance gains (wide_and_tapered_arc_tessellation benchmark across 7,500 entities):
- Latency: 27.34 ms -> 23.30 ms (-14.8%)
- Mean latency: 28.08 ms -> 24.15 ms (-14.0%)
- Throughput: 265.6k entities/s -> 321.9k entities/s (+21.2%)
- Correctness: 100% pass across all 45 polyline and 5 tessellation unit tests.
With a working plane other than world XY, ROTATE and MIRROR handed a
non-ordinate dimension to acadrust's Entity::apply_transform, which only
translates those dimension types by where the transform sends the origin.
The definition points, plane and angles stayed unturned, so the dimension
jumped back once its block was baked again.
Send both through the existing Affine branch, which ALIGN already uses and
which transforms points, normal and stored angles together. The world-Z
paths are unchanged.
Co-Authored-By: Claude Opus 5 <[email protected]>
Hakan Seven