-
Notifications
You must be signed in to change notification settings - Fork 682
Description
The sequence Ra + Virama + ZWJ is used to create the reph form. Virama + ZWJ + Ra is for the below base form of Ra and Virama + ZWJ + Ya the post base Ya. This creates an ambiguity for Ra + Virama + ZWJ + (Ra or Ya).
I am not sure whether Ya-Reph or Ra-Yapost is preferable. It seems Ya-Reph is OK. Ra-RaBelow doesn't happen because shapers seem to prefer the reph form. But the Noto fonts are set up to prevent Ra-Reph. For Windows this works with a "pres" lookup that splits the Reph into Virama + Ra, when it follows Ra. Proofing combinations with reph look like this in Edge on Windows 10:
And like this in Chrome on Mac:
Harfbuzz orders the Reph after Virama and vowel signs, rather than after the consonants. Also note the difference in the first word where there is consonant + Virama + ZWJ + Ra + Virama + ZWJ + Ra. Harfbuzz inserts two below base Ra's and Windows must given precedence to making a Reph instead (which is split up).
When I add a lookup to the akhn feature that removes the ZWJ between two Ra's then Edge gives me this:
The first word "ස්ථ්ර්රවම" now uses a subscript for the first of the two Ra's. Not sure if that is an improvement or not. And the triple Ra no longer has a Reph floating in front, which is better.
For Chrome I get this:
This is an improvement from the point of view of preventing overlaps. But when there are split-vowels, something must be between the Ra's when the akhn lookup is applied. And then the fix doesn't work.
When there is a Reph and a post base Ya the Reph is attached to the base consonant, unless the base is the Ya itself. Then it is attached to the post form. Windows seems to put the Reph after the Yapost, and it is reordered by a lookup. (Which doesn't work when there is a below base Ra too, but I am not worried about that.)



