Commit 4ea3737 changes shaping behavior

[dscorbett]

4ea3737d04c575cd9b6ffda1e4e0f2c9d2b60d9c changes shaping behavior, so it is not a valid optimization. For example, given Duployan-Regular.otf, this command:

hb-shape Duployan-Regular.otf -u '1BC09 1BC9E 1BC9E 1BC9E' -O json \
| jq -r 'map(.g | select(startswith("_") | not))[]'

used to produce:

u1BC09.v
u1BC9E_u1BC9E_u1BC9E_.double__double__double.30

but now produces:

u1BC09.v
u1BC9E_u1BC9E_.double__double.30
dupl_.Line.0.dotted
u1BC9E_.double.90.wl

[behdad]

Thanks. I'll investigate.

[behdad]

@dscorbett Is there a Duployan corpus (no matter how small) I can use for performance testing?

[dscorbett]

The Kaltash Wawa has some pages in Duployan which you can find by searching for “font”. The longest Duployan page is https://kaltashwawa.ca/2022/01/20/lovecraft-in-chinook-puspus-kopa-ulthar-tawn/.