The script which checks the errors in the strokes is very slow & also
requires clustering of results. It also has to be run over and over to
find these errors, since once too many wrong things get into the
cluster of results it can't discriminate any more. This represents
about three runs.
Many of the Kaisho variants consist of nothing other than a down-line
in a tree shape with a hook at the end, but this difference is not
represented in the stroke's kvg:type. This uses a graphical search
algorithm to find instances where a line is labelled ㇑, but appears
to actually be ㇚ graphically, and automatically changes the files.
There are about 150 changes. I've done spot checks on about ten of the
files to make sure that the correct things are being done.
Two cases were missed in the previous commit, the case where the ㇕v
was on a child path of the top-level group, and the case where ㇕v
appeared more than once in the same kanji.
This replaces kvg:type="㇕v" with kvg:type="㇁" in all cases.
This edit was done automatically. I did not check every edit but I
checked that an appropriate element was present in most of the files.
The position of the Nelson radical on stroke 4 was extremely unlikely
to be the correct one, since Nelson's reworking of the radical system
is mostly about insisting on using the upper left corner as the
radical, rather than using a recognisable shape. Unfortunately that
also means that the radical on 053db-Hyougai is the wrong shape. I
think this is yet another case of the Nelson radical being quite
ill-informed. Anyway I've moved both of them even though this means
that the radical on the Hyougai variant looks more like radical 3 than
radical 4.
These type and group mistakes were found during the processing of an
automated script for detecting stroke shapes. The processing so far
only covers 3000 of the 10,000 files.
I decided to use CJK composition characters rather than "bucket" or
similar, since those are clearer. I used A and B to indicate the outer
and inner parts of the composition characters.
Add missing stroke, labels to 07726-KaishoHzFst.svg
This commit consists of multiple additions of groups on empty elements
from the list 敫𦥯㳟𦔮𡗗耑电翏夌. Two characters, 09675.svg, with an
incorrect group element, and 07ac5.svg, where the white element was
covering more strokes than it should, were edited in a text editor to
correct the mistakes. The rest of the edits were automated.
Ben Bullock