Punjabi (Gurmukhi script):consonant+halant+consonant not handled correctly

[xiangyexiao]

When type consonant+halant+consonant, the rendering of "Noto Sans Gurmukhi" comes side by side instead of 2nd consonant below the 1st consonant.

For example for the consonant ਕ, the rendering is correct only for 4 types of conjuncts : ਕ +halant+ਰ = ਕ੍ਰ, ਕ +halant+ਵ = ਕ੍ਵ, ਕ +halant+ਹ = ਕ੍ਹ, ਕ +halant+ਯ = ਕ੍ਯ. This is the case with all the consonant conjuncts.

STEPS TO REPRODUCE: Type on ਕ + halant+ਗ

EXPECTED RESULTS: Left in 'example.PNG'

OBSERVED RESULTS: Right in 'example.PNG'

In the attached spreadsheet (punjabi conjunct.xlsx), all conjuncts marked in green are rendering correctly with Noto Sans Gurmukhi for all consonants ,while the conjuncts marked in white & yellow are rendering incorrectly, i.e., side by side.

Attached the following files Gurmukhi MN on Mac.PNG: rendering of Gurmukhi font Gurmukhi MN on Macbook Noto on Mac.PNG: rendering of Noto Sans Gurmukhi on Macbook Noto using hb-view: rendering of Noto Sans Gurmukhi using HarfBuzz (hb-view) punjabi conjunct.txt: plain text of punjabi conjunct trix punjabi conjunct.xlsx: MS Excel version of punjabi conjunct trix

punjabi conjunct.txt punjabi conjunct.xlsx

[xiangyexiao]

internal bug: http://b/24795868

[jungshik]

/cc @JelleBosmaMT , @BrianKraimerMT , @kmansourMT , @behdad

[tiroj]

Some background on why the fonts may be the way they are:

Only the three subscript consonants ਕ +halant+ਰ = ਕ੍ਰ ਕ +halant+ਵ = ਕ੍ਵ ਕ +halant+ਹ = ਕ੍ਹ and the postscript -Ya ਕ +halant+ਯ = ਕ੍ਯ are used in modern Panjabi, and grammars of the modern language typically show only these forms (modern, in this instance, means 18th Century onwards).

The other subscript consonants are found only in older texts and, most notably, the Sikh scriptures, the Adi Granth.

So it is possibly 'by design' that the Noto Sans Gurmukhi fonts support only the modern subscripts, but given the general purpose of the Noto fonts and the importance of the Adi Granth text, I agree that this issue should be addressed.

[Note that there should be no concern about the older subscripts occurring unwanted in modern texts: in Panjabi, unlike most other Indian scripts, the halant is not written and U+0A4D is used only when the subscripts are specifically desired.]

[xiangyexiao]

downgraded the priority based on tiroj's comment.

[JelleBosmaMT]

At the time of the design, my knowledge of these additional presentation forms did not go beyond "old inscriptions". Because digital fonts are not really intended to be used for new old inscriptions, it did not seem a good idea to delay the work trying to support these. "Old inscriptions" sounds to me like a very limited category, compared to "old manuscripts" or "old books". If would be vital to know how these are used in relation to each other and vowel marks, in order not to descend into an incredible effort to try and combine all possible combinations of marks.

[tiroj]

On the subject of subscript mark interaction, both Noto Sans Gurmukhi and the MS Nirmala UI font need to avoid double stacking of below marks because of restricted linespacing in target environments. In this screenshot, Murty Gurmukhi and the MS Raavi font show conventional arrangement of subscript and halant in a three-letter conjunct in the word ਓਲ੍ਹ੍ਹਗੀ. Noto Sans Gurmukhi and Nirmala UI each resolve this horizontally in a different way. I am wondering if native users have any feedback on which method is preferable?

[JelleBosmaMT]

To me the only logical solution is to have the subscript last. Subscript forms are used at the end of conjuncts, rather than in the middle. As non-native reader the visual appearance is l-h-ha in Noto and l-h ha in the others. The vertical line of the 2 disconnects the final full ha.

An investigation into syllables with two Virama’s, it is interesting to see most of them make use of these additional subscripts that Raavi has and are missing in Nirmala and Noto. In most cases it includes a double t, such as in ਤ੍ਤ੍ਰ (ttra) And it shows the danger of adding these, because in Raavi none of these are legible….

[jungshik]

@tiroj : in the above screenshot, did you use "Noto Sans Gurmukhi" or "Noto Sans Gurmukhi UI"? I expected the former to behave more like Raavi (I haven't checked it out) while the latter (UI font) behaves roughly like 'Nirmala UI'

[JelleBosmaMT]

It makes very little difference. In the UI font the subscript ha is smaller. Raavi has a very, very large default line distance. They could have easily used two ha subscripts and still have room for a virama. I don’t think neither Raavi nor Nirmala rendering looks desirable.

On 20-okt.-15 10:23, "jungshik" notifications@github.com wrote:

@tiroj https://github.com/tiroj : in the above screenshot, did you use "Noto Sans Gurmukhi" or "Noto Sans Gurmukhi UI"? I expected the former to behave more like Raavi (I haven't checked it out) while the latter (UI font) behaves roughly like 'Nirmala UI' — Reply to this email directly or view it on GitHub https://github.com/googlei18n/noto-fonts/issues/529#issuecomment-14947487 6.

[marekjez86]

see https://github.com/googlefonts/get-noto/tree/master/unhinted/Gurmukhi

[nizarsq]

@marekjez86 Not sure what is the decision came out of this issue. Tested characters: ਕ੍ਗ ਕ੍ਚ ਕ੍ਟ ਕ੍ਠ ਕ੍ਤ ਕ੍ਥ ਕ੍ਦ ਕ੍ਨ ਖ੍ਗ ਖ੍ਚ ਖ੍ਟ ਖ੍ਠ ਖ੍ਤ ਖ੍ਥ ਖ੍ਦ ਖ੍ਨ ਗ੍ਗ ਗ੍ਚ ਗ੍ਟ ਗ੍ਠ ਗ੍ਤ ਗ੍ਥ ਗ੍ਦ ਗ੍ਨ ਘ੍ਗ ਘ੍ਚ ਘ੍ਟ ਘ੍ਠ ਘ੍ਤ ਘ੍ਥ ਘ੍ਦ ਚ੍ਗ ਚ੍ਚ ਚ੍ਟ ਚ੍ਠ ਚ੍ਤ ਚ੍ਥ ਚ੍ਦ ਛ੍ਗ ਛ੍ਚ ਛ੍ਟ ਛ੍ਠ ਛ੍ਤ ਛ੍ਥ ਛ੍ਦ ਜ੍ਗ ਜ੍ਚ ਜ੍ਟ ਜ੍ਠ ਜ੍ਤ ਜ੍ਥ ਜ੍ਦ I tested reported issue on the current NotoSansGurmukhi and the consonant+halant+consonant still rendered side by side. I also tested NotoSansGurmukhiUI and NotoSerifGurmukhi both fonts have same behavior as NotoSansGurmukhi. I tested same companiation with none Noto fonts (Times) and (Mukta Mahee) on MAC and both rendered halant under the consonant. Hopefully I'm not missing anything. See attached for the current behavior

[simoncozens]

OK, there are a variety of issues here; I'm going to split this issue up so it can be addressed.

• Some conjuncts are implemented as atomic glyphs (e.g. karaguru).
• For the conjuncts that do work but which are implemented through mark attachment, Noto Serif Gurmukhi is not attaching them correctly. (This is probably my fault, as I'm guessing it happened during the MTI->FEA conversion?)
• Having a la / ha-ha conjunct instead of a la-ha / ha conjunct does seem unusual, regardless of UI/non-UI considerations. (The ...UI fonts will eventually be going away when the Android team rewrites their code to support taller and deeper boxes.)
• I agree that the conjuncts found in the Adi Granth are significant enough to be supported, and should be supported. The conjuncts required will need to be identified by analysing the text, and this may need some new glyphs drawn - ideally in a way that uses combining attachment instead of atomic glyphs. This part is obviously a longer term project.

[simoncozens]

Closing this in favour of #4, #5 and #6.