bugFree UCS Outline Fonts - Bugs: bug #52213, Devanagari -- Marathi support

 
 

bug #52213: Devanagari -- Marathi support

Submitter:  Steve White <Stevan_White>
Submitted:  Thu 12 Oct 2017 09:48:51 AM UTC
   
 
Category:  character range Severity:  3 - Normal
Item Group:  character substitution issue Status:  Fix posted
Privacy:  Public Assigned to:  Stevan_White
Open/Closed:  Open Release:  * development
* Mandatory Fields

Add a New Comment Rich Markup
   

Jump to the original submission

Wed 18 Oct 2017 02:24:52 PM UTC, comment #15: 

You are not wrong. I think that when all other examples contradict Velthuis, it makes sense to not follow it so closely. I’ll try to find some examples from old printed books in Marathi.

David Corbett <dscorbett>
Wed 18 Oct 2017 01:56:32 PM UTC, comment #14: 

Hi David,

Velthuis shows two forms of the letter "la".  That can be chosen by the user in FreeFont.  I think the default style, with the straight side, is also the default in Velthuis.  Am I wrong about that?

As to the forms of the conjunts h_la and m_la -- the font seems to be consistent with Velthuis.  Am I wrong about that?

I'm afraid you're suggesting that I just follow what some other special-purpose third-party fonts do.  You know how I'll feel about that.  And you know that in my opinion, Velthuis is a pretty good font, time-tested and heavily used by many people for many different things.

Steve White <Stevan_White>
Group administrator
Tue 17 Oct 2017 07:20:45 PM UTC, comment #13: 

I base that stylistic judgment not only on the TDIL script grammar, but also on the fonts Chandas, Nakula, and Siddhanta, where the la has a curved right side in all contexts. In other Devanagari fonts, the la has a straight right side in all contexts. FreeSerif’s style for this letter is inconsistent with other fonts and with itself.

David Corbett <dscorbett>
Mon 16 Oct 2017 10:03:34 PM UTC, comment #12: 

Hi David,

OK, but this is very definitely a stylistic judgment. 

I never took the TDIL documents to be dictating the exact form of glyphs, even where they're basically right.

Furthermore, as I've said, I have come to suspect that much of what is in the TDIL documents is idiosyncratic of the fonts used to produce them -- not reflective of general typographic conventions. 

And the Marathi Grammar is one of the ones for which I only have an initial copy -- it never underwent a cycle of improvement, as near as I can tell.

So... I would really need further outside evidence.



Steve White <Stevan_White>
Group administrator
Mon 16 Oct 2017 05:16:09 PM UTC, comment #11: 

The Marathi la (dev_bby_la) has a curved right side whereas the default la (ladeva) has a straight right side. However, the la part of hla and mla has a straight right side in FreeSerif even in Marathi, contrary to the TDIL script grammar for Marathi.

David Corbett <dscorbett>
Mon 16 Oct 2017 03:15:37 PM UTC, comment #10: 

Hi David,

I found the typo that caused the ḍya instead of ṅya.  Fixed.

I'm not sure what you're referring to with "right side of the la in hla and mla"
... in which document?
In the velthuis PFB files, there are ligatures h_l and m_l.  These should be about identical to the ones in FreeSerif.  Are they different?  Or should they be?  Why?

Steve White <Stevan_White>
Group administrator
Fri 13 Oct 2017 09:30:10 PM UTC, comment #9: 

I am having my doubts about many of the ligature glyphs in the TDIL Grammar for Marathi.  At least this set

ट्ख ठ्क ड्क ड्ख ड्ज
छ्व ट्व ठ्व ड्व ढ्व
फ्ल ब्ल ळ्ळ थ्न

They all show fancy special ligature glyphs, yet I have been unable to locate any native Marathi words that contain them.  Some I can't find any Marathi text at all that contains them.

प्लूटो (pluto)
प्लॅटिनम    (platinum)
प्लेयर्स    (players)

इन्फ्लुएन्झा    (influenza)

ब्लॉग   (blog)
ब्लू    (blue, as in blue movie)

But many don't seem to be in any text I can find on the web, or on mr.wikipedia.org, or anyplace.

Some of them aren't even listed in the lists of Sanskrit conjuncts I have.

What does this mean?  Is it standard practice in Marathi to represent foreign sounds with a fancy ligature?

Or is it that the glyphs represented in that document are idiosyncratic leftovers in the font?

Steve White <Stevan_White>
Group administrator
Fri 13 Oct 2017 06:43:03 PM UTC, comment #8: 

The right side of the la in hla and mla should be curved in Marathi.

The TDIL script grammar does not list a ligature for ṅya, but as long the font gives it one, it should not use the ḍya glyph.

David Corbett <dscorbett>
Fri 13 Oct 2017 12:56:29 PM UTC, comment #7: 

I implemented that 'blws' idea.  Not too bad to do -- works OK.

A contextual chain 'blws' decomposes these specific half-forms when  followed by la. IThis is followed by another 'blws' that replaces  virama-la to a below-form la.
Since the 'blws' is consulted only after almost all other lookups, it's fairly safe that any remaining virama+la was produced by the preceding decomposition.

So that took 3 lookups.  Might implement it in serif too.

Steve White <Stevan_White>
Group administrator
Fri 13 Oct 2017 11:13:15 AM UTC, comment #6: 

The case of the Marathi subscript -la.

TDIL shows it on k g p ph b n.  Proper ligatures with -la are formed with l and h, also, the half form of s takes a different shape before -la.

One question: do other combinations with -la ever occur in Marathi?  If so, the forms with a full -la could be viewed as a fallback rendering, or even a font peculiarity.

The obvious thing would be separate ligature glyphs for the six cases -- no catastrophe, but still...could it be done with tables?  To implement this, I tried a 'blwf' table, but these days it seems really to act only on virama+ra combinations (?) maybe there's no context for the chain.  Another option would be a 'blws' contextual chain which for these six decomposes the half-form of half+la combos, followed by a second chain that forms the subscript la.  Each chain requires 2 lookups: 4 lookups in all, to save 6 ligature glyphs.  Is it worth it?

Steve White <Stevan_White>
Group administrator
Thu 12 Oct 2017 05:56:19 PM UTC, comment #5: 

The conjucts are pretty much finished in serif, and I think they're usable in sans. 
Only that strange द्न्य is completely missing, near as I can tell.

I had mistaken some conjuncts: j-ja ज्ज should have been j-nya.  The strange form is that "bby" style of the ligature.
Attached are corrected test files.

(file #42128, file #42129)

Steve White <Stevan_White>
Group administrator
Thu 12 Oct 2017 02:31:22 PM UTC, comment #4: 

Note: not one of the two-consonant conjuncts still missing from Sans appears in the UDHR for Marathi.

Steve White <Stevan_White>
Group administrator
Thu 12 Oct 2017 02:01:57 PM UTC, comment #3: 

Committed more changes to support some three-consonant ligatures listed in the TDIL grammar.  Remaining:

Still no glyph: स्त्र, द्न्य (but the latter is very odd, and of special use)
Forming in different order: त्त्र 
Can't see well the alternative form: ह्म्य

Steve White <Stevan_White>
Group administrator
Thu 12 Oct 2017 11:51:17 AM UTC, comment #2: 

Last commit to serif basically finish all the two-consonant conjuncts.  The sh-la was already in the font, and doesn't look like the TDIL, but I like it better.  The j-ja in the TDIL looks very strange to me -- I would never think it was a j-ja.  I looked in other Devanagari fonts that support the ligature, nothing looks remotely like it -- they are all distinct ligatures of two ja's (either vertical or horizontal).  So I made my own vertical ligature.

There is now a chain lookup in Sans for the subscript -la.  It might be worth implementing in Serif too -- it would save a few ligatures.

Steve White <Stevan_White>
Group administrator
Thu 12 Oct 2017 10:07:17 AM UTC, comment #1: 

Recent commits to serif improve that a lot.  Using existing glyphs in the font, just making new entries to the Marathi tables, gets most of the two-consonant conjuncts.  Still missing:

छ्व ज्ज ड्ख श्ल

I think I'll just go ahead and make those.

The TDIL also lists several consonant triple ligatures.  Most of these are already present in the font, too.  Need a test file.

Sans is going to be a bigger problem.  Its set of existing ligatures isn't as big.  Mono...is a much bigger proposition, having only maybe half of the required vertical ligatures.

Steve White <Stevan_White>
Group administrator
Thu 12 Oct 2017 09:48:51 AM UTC, original submission:  

Marathi has been only partially supported.  Reason was: hadn't seen the TDIL Grammar for Marathi until after work stopped on Devanagari.

All the faces support the eyelash-ra to some degree.

FreeSerif had special lookups to form the vertical ligatures that are more common in Marathi than in Hindi.  These lookups are absent in Sans and Mono, and even in Serif, about 20% of the consonant pair ligatures are not forming.

Attached is a table of two-consonant ligatures that appear in the TDIL Grammar for Marathi.

Steve White <Stevan_White>
Group administrator

 

(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)

Attach Files:
   
   
Comment:
   

Attached Files
file #42128:  mr-tdil-conj.html added by Stevan_White (18KiB - text/html)

 

Depends on the following items: None found

Items that depend on this one: None found

 

Carbon-Copy List
  • -email is unavailable- added by Stevan_White (Submitted the item)
  •  

    There are 0 votes so far. Votes easily highlight which items people would like to see resolved in priority, independently of the priority of the item set by tracker managers.

    Only logged-in users can vote.

     

    Follow 7 latest changes.

    Date Changed by Updated Field Previous Value => Replaced by
    2017-10-12 Stevan_White Attached File- Added mr-tdil-conj.html, #42128
        Attached File- Added mr-tdil-conj3.html, #42129
        StatusProceeding Fix posted
    2017-10-12 Stevan_White Attached File#42123 Removed
    2017-10-12 Stevan_White Attached File#42126 Removed
    2017-10-12 Stevan_White Attached File- Added mr-tdil-conj3.html, #42126
    2017-10-12 Stevan_White Attached File- Added mr-tdil-conj.html, #42123

    Back to the top

    Powered by Savane 3.13-3230.
    Corresponding source code