The parity line said a soft-edged gray stroke running out of black loses
"3 or 4 pixels" where it meets black. It now says 3 or 4 of the 16 pixel
centers along its middle nearest black, as docs/image-trace-adjustments.md
and the test count them.
In docs/image-trace-adjustments.md, gray near black on blurred edges now
traces gray at fewer of its pixel centers instead of "losing pixels", a
wider band loses some or most of its gray instead of part of it (bands 3
and 4 pixels wide at a blur of 1 pixel keep about 60 of 336), and the
paragraph is wrapped to the file's line width.
The band test now also sums the samples traced gray for bands 3 and 4
pixels wide blurred by 1 pixel (about 60 of 336 over the three grays;
16 without the sharp-edge stop) and for bands 4 pixels wide blurred by
3 pixels (about 39 of 192).
across_edge decides from the first step alone whether one side's edge is
sharp, so each side's path is walked once. On 856 code-drawn band and
stroke cases and 39 images the results are the same as before.
The docs count pixel centers instead of pixels, write "0.6 pixels"
throughout, give the parity line's band claim its bound of 1 in 16, and
state again that a light gray stroke 6 or 8 pixels from black keeps its
gray because its pixels do not reach black in 5 steps.
The Trim at Crossings Tool joins the Scissors group in both shipped
toolbars, after the Knife, and stays in Object › Plug-ins › Path
Precision Tools. The All Tools drawer lists it once, in the Scissors
group, and Edit › Keyboard Shortcuts › Tools lists it once. It has no
default shortcut. A toolbar the user already edited keeps its saved
layout; Reset Toolbar adds the tool.
The Knife's cursor is now drawn as a knife, its blade's point on the
hotspot, instead of borrowing the crosshair. The Knife tool and the Trim
at Crossings Tool both show it. Use Precise Cursors still shows the
crosshair.
Suggested by erolka and leanma.
With #8144, an edge pixel between black and white is read as a blend of
black and white when paths of up to 5 steps reach past a gray between
them both ways. On edges blurred by 0.6 pixel, the outer pixels of a gray
band 3 or 4 pixels wide along black also met that rule: their first step
outward is white, and the path toward black crosses the band's gray. A
3 px band at that blur kept 16 of its 48 samples gray, against 46 before
#8144.
When the first step of one path already stands for its color (the edge
on that side is sharp), the other path now stops at a pixel of the gray
it comes to before getting past it. Bands 2 to 4 pixels wide blurred by
0.6 pixel keep 306 of their 432 samples gray (207 with #8144, 324 before
it). Edges blurred by 1 to 3 pixels trace as with #8144, except that a
light gray (230) blurred by 1.5 pixels leaves at most 4 slivers of up to
3 pixels beside the disc.
The band test now also runs at blurs of 0.6, 1 and 3 pixels, the spoke
test at 2 pixels, and the stroke test's ranges are the measured ones.
The docs state the distance rule as the code computes it and limit each
measured claim to the blur it was measured at.