Feature Request: Add support for numderline (easier number parsing)

[kazuoteramoto]

Description of the new feature/enhancement (with images if possible)

Visually parsing a large sequence of numbers like 1570068719 can be tricky. Tristan Hume introduced a solution called numderline. The basic idea is to use variants based on the context. I'm using it with cascadia code with a good success.

To illustrate, this is currently my setup (this repo have the needed code to patch the font, I'm using the following parameters --no-underline --shift-amount 90 --squish 1)

I think that something subtle is enough for daily use.

Proposed technical implementation details (optional)

The two linked pages have more details, its uses font shaping, but I'm not sure of the details. I'm hoping this can be toggled by a font feature if implemented upstream.

I'm bringing this issue here because I think this can be better integrated in the original font, two reasons I see:

• One problem with the patcher is that it don't support ttf variable so we miss the great hinting;
• Better shapes. I think that some hand adjustment of the shape of the digits can make it looks better.

[aaronbell]

Interesting.

Thinking through it, I don't think it would be particularly difficult to implement. Perhaps narrow the numbers by 10% or 20%, then create a variant of each where it is off to the right, centered, and off to the left.

Then it'd just be a question of the OT code to implement.

[PaulDempsey-ms]

Note that digit grouping conventions are culture-specific (not all cultures group into uniform 3s)