Value encoding strategy

[dy]

What would be the right strategy to map characters?

0. Direct mapping to char codes with offset

• Super-simple uints.map(v => String.fromCharCode(v + offset))
• Easy to remember/understand
• Extendable as baseline shift (I guess? GPOS type 9 via comb characters?)
• Friendlier to variable font for align axis
• Good offsets for 255 characters: U+A000, U+1500, U+1E00, U+A500, U+AC00 (~0x1fff!), U+E000 (private, but not selectable)

1. Hex pairs (doubles) as a0a1a2

• Trivial encoding: uints.map(v => v.toString(16)).join('')
• Deleting first character rerenders full font, non-pair copying doesn't make sense
• Ligatures don't seem to support LigatureCaret and ligature width is equally split
• Doesn't seem to be possible to collapse one of the characters if they're normal ~ Only using combining accents make full-length ligatures

2. a-z as main values, 0-9 as fraction (0-width), or any other separation on base and fract

• No ligatures problem, selection is resolved: second char is simply 0-width like combining mark ~ Deletion still problematic, but not as much as the first option
  • Encoding is not as trivial: uints.map(v => Math.floor(v) + '\u0300'.repeat(v%1)) etc.
• Manually understandable: a0, b3, g9

3. 0-9 as main values, ↑ and ↓ as combining modifiers

~ similar to previous conceptually

• Selection is good
• Deletion is as good as combining characters ~ Manually not simple to type, but readable
• Precision control via eg. number of combining characters, eg. 0̂9̂̂̂5̂̂̂̂̂̂̂ → 0.1, 9.3, 5.7 etc.
• Encoding is zoom-dependent, from as simple as uints.map(v => Math.floor(v)) to zoom control. ~ 0̂̂̂̂̂̂̂̂̂̂ === 1, latin arithmetic-like system

4. Baseless modifiers only .̂̂̂̂.̬̬̬.̬̬.̬̬̬̬̬̬̬.̬.̬̬̬̬̬̬̬̬̬̬̬̬̬.̂̂̂.̂̂̂̂̂̂̂.̂̂̂̂̂̂̂̂̂̂.̂̂̂.̂̂̂̂.̂̂̂̂̂.̂̂.̂

~ Printer-like tech. ~ With vertical space combining mark enables xy-plot as a single string... ~ Would require sort of i,d,v,x composing characters to combine into bar

5. First value is chunk identifier (latin), rest is spaces with combining marks

• Dbl-click selection at least in browsers, selects word and all space after it.
• "Space" has literal meaning of "move caret right without print"
• Character more semantically acts as identifier in case of font fallback, which is a good mathematical/data practice (legend etc)
• Data is therefore put as combining marks, and since their number is limited - as positional notation, mb similar to roman numerals i, v, x https://unicode-table.com/en/0365/
• Encoding requires heuristic: marking a chunk, then defining spaces
• Dblclicking space after id doesn't select an id

[dy]

Probably mixed approach is the way. For values - 0 method. For vertical shift - 1 cursive feature see next issue.

[dy]

Maybe non-latin range has drawback of bad escaping. Latin range allows:

• mapping as toString(36)
• copy/paste to URL
• relatively gradual increase of values

[dy]

Simple mapping to 0x100+ is a good balance.