Stroke in 娩 is attributed as 'element' rather than 'type' ; Stroke in 禺 and its descendants declared as wrong type

[vipena-astro]

I've downloaded the database kanjivg-20160426-main.zip and dived deep into the XML hierarchy of grouped SVG paths that represent radicals, sub-radicals, etc.

In almost all kanji, the bottom XML level is a single stroke and contains the attribute 'type' but not 'element', as described in the documentation , whereas middle-level groups contain the attribute 'element' but not 'type'. This is not true for 娩 (U+05A29), where the second-to-last stroke is reported as the element '丿' (CJK Unified ideograph U+4E3F) rather than the type '㇒' (U+31D2, as it is done in the entry for kanji 免 (U+514D)), or '㇓' (the not included CJK stroke U+31D3). AFAIK, this is the only kanji in kanjivg-20160426-main.zip where this happens.

Also, the 'type' attribute of the eighth stroke of 禺 (U+79BA) is set to '６' (U+FF16). I'm no expert in Japanese calligraphy, but I would say it looks similar to the eight stroke of 球 (U+7403), i.e. '㇀' (U+31C0). Moreover, both type '６' and '㇀' are different from type '㇐' (U+31D0) that appears in the radical 114, viz. 禸 (U+79B8). This problem is passed down to some kanji 禺 acts as a group in, viz. 勵 (U+52F5), 寓 (U+5BD3), 嵎 (U+5D4E), 癘 (U+7658), 礪 (U+792A), 糲 (U+7CF2), 萬 (U+842C), 藕 (U+85D5), 蠣 (U+8823), and 邁 (U+9801).

[benkasminbullock]

On Mon, 28 Feb 2022 at 10:21, Valentín Peña-Donaire < ***@***.***> wrote: I've downloaded the database kanjivg-20160426-main.zip <https://github.com/KanjiVG/kanjivg/releases/download/r20160426/kanjivg-20160426-main.zip> and dived deep into the XML hierarchy of grouped SVG paths that represent radicals, sub-radicals, etc. In almost all kanji, the bottom XML level is a single stroke and contains the attribute 'type' but not 'element', as described in the documentation <https://kanjivg.tagaini.net/format.html> , whereas middle-level groups contain the attribute 'element' but not 'type'. This is not true for 娩 (U+05A29), where the second-to-last stroke is reported as the element '丿' (CJK Unified ideograph U+4E3F) rather than the type '㇒' (U+31D2, as it is done in the entry for kanji 免 (U+514D)), or '㇓' (the not included CJK stroke U+31D3). AFAIK, this is the only kanji in kanjivg-20160426-main.zip where this happens.

According to IDS (https://github.com/cjkvi/cjkvi-ids) the correct decomposition of the right side is U+514D 免 ⿱𠂊⑤[GTK] ⿳𠂊𫩏儿[J] where the Japanese usually uses the "leg" part. This is in the github repo of kanjivg as 05a29-Hyougai.svg. But this seems to be an error in kanjivg where the leg element should have been used.

Also, the 'type' attribute of the eighth stroke of 禺 (U+79BA) is set to '６' (U+FF16). I'm no expert in Japanese calligraphy, but I would say it looks similar to the eight stroke of 球 (U+7403), i.e. '㇀' (U+31C0). Moreover, both type '６' and '㇀' are different from type '㇐' (U+31D0) that appears in the radical 114, viz. 禸 (U+79B8). This problem is passed down to some kanji 禺 acts as a group in, viz. 勵 (U+52F5), 寓 (U+5BD3), 嵎 (U+5D4E), 癘 (U+7658), 礪 (U+792A), 糲 (U+7CF2), 萬 (U+842C), 藕 (U+85D5), 蠣 (U+8823), and 邁 (U+9801).

I'm currently examining this and I will respond tomorrow. Message ID: ***@***.***>

…

[benkasminbullock]

Also, the 'type' attribute of the eighth stroke of 禺 (U+79BA) is set to '６' (U+FF16). I'm no expert in Japanese calligraphy, but I would say it looks similar to the eight stroke of 球 (U+7403), i.e. '㇀' (U+31C0). Moreover, both type '６' and '㇀' are different from type '㇐' (U+31D0) that appears in the radical 114, viz. 禸 (U+79B8). This problem is passed down to some kanji 禺 acts as a group in, viz. 勵 (U+52F5), 寓 (U+5BD3), 嵎 (U+5D4E), 癘 (U+7658), 礪 (U+792A), 糲 (U+7CF2), 萬 (U+842C), 藕 (U+85D5), 蠣 (U+8823), and 邁 (U+9801).

From grepping the source files, kvg:type="６" only seems to be used in 禺 and characters derived from it, so presumably it's a "special case". The nature and origin of this special case is not documented in any repository file, but since it's only used here, it doesn't seem to be a particular problem.

Incidentally I could not find other numbers used in this way.

[benkasminbullock]

This is not true for 娩 (U+05A29), where the second-to-last stroke is reported as the element '丿' (CJK Unified ideograph U+4E3F) rather than the type '㇒' (U+31D2, as it is done in the entry for kanji 免 (U+514D)), or '㇓' (the not included CJK stroke U+31D3). AFAIK, this is the only kanji in kanjivg-20160426-main.zip where this happens.

This particular issue seems to have been fixed in

ecf9eb0.

which was a response to

#155

To save this bug from being reported again, another release of the data files should be made. I don't know how to do that yet but I'll look into it.

[benkasminbullock]

I've made a new release of the files, so I think this can be closed now. Please open it again if not satisfied. I've also added a note about the ６ into the documentation pages, but hopefully that can be replaced with a better stroke type soon.