Indic cleanup #3648
Indic cleanup #3648
Conversation
This was left off of the commit moving Indic categories to the generator. It didn't fail any tests, but adding it back because it has implications possibly.
Not sure how this was passing tests still.
|
I've tested with harfbuzz-testing-wikipedia corpus that the Myanmar shaping against NotoSansMyanmar didn't change. I'll test Khmer and Devanagari and report as well. |
|
I confirmed that Khmer and Devanagari corpus also render the same before/after this PR. |
|
What still can be done is to export categories from Indic machine like the other machines, and use those in the Indic shaper instead of the |
Done. |
|
@dscorbett This is ready for review. Thanks. |
|
Thanks @dscorbett for all the feedback. This PR is a refactoring of the code without any semantic changes. May I merge this and then address your comments in subsequent PRs? |
|
Okay I'm starting to apply your requested changes here. |
These three characters have Indic_Syllabic_Category=Consonant_Placeholder. The original evidence that prompted these overrides says they can take tone marks. They are not subjoined: Khamti Shan apparently does not use subjoined characters at all. Therefore, PLACEHOLDER is good enough and these need not be overridden to C. https://www.unicode.org/L2/L2008/08276-khamti-proposal.pdf #3648 (comment)
|
Okay. I think I've addressed all. |
Other than the removing the duplication completely. |
|
Let's merge this and continue separately. |
|
Other than the Myanmar medials’ comments, this is ready to merge. |
|
Thanks. |
This supersedes #3635 and moves all exception handling of indic/khmer/myanmar shapers to the indic table generator.