Decide which diacritics to make alts of about shantell-sans

[arrowtype]

Finnish has words like ulkonäöltään and Estoniantöötajaskonnas that use double diacritics:

https://www.setuptype.com/x/cod/languages/fi

And I’ll test how much data it adds to include the most-common diacritics.

ÉéÓóÍíÁáÜüÇçÃãÖöÄäÑñ

[arrowtype]

Okay, so with 4 alts, plus those few most-common diacritics, the size only increases by a little. Below, core skips the diacritics for 451kb total, while select_alpha adds them for 464kb total:

To me, that says that it’s worth including alts for these selected diacritics. People can always slice the axes & subset the character set to reduce size, anyway.

[arrowtype]

Something nice: it’s really simple to remove all the .alt glyphs with pyftsubset, by removing the calt feature and keeping all unicodes.

pyftsubset fonts/shantell-sans-v13/shantell-wght_BNCE_IRGL--reverse_bounce-select_alpha-VF.ttf --layout-features-="calt" --unicodes="*"

Because the font still has positions randomized in the core glyphs, it still looks pretty decent. It’s not as good, so this isn’t the main version for releases, but it may be worth shipping as an alternative option for the web.

Top: no calt / no alts. 212kb TTF / 85kb WOFF2.
Middle: calt with 3 alts. Decent, but fails in a word like "experience" where letters repeat every 3 positions. 393kb TTF / 136kb WOFF2.
Bottom: calt with 4 alts & improved randomization. 464kb TTF / 143kb WOFF2. Not actually that bad. Components are helping a lot, I think.

[arrowtype]

For now, I’ll conclude on making alts for the following characters:

ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz

ÉéÓóÍíÁáÈèÜüÇçÃãÖöÄäÑñ # notably, excludes À, which could mess up line heights on macOS

0123456789!"#$%&'()*+,-./\:;<=>?@[]^_`{|}~“”‘’

This feels like a good trade-off between including a lot, but not everything. If we need to in the future (or if someone wants their own, custom version), this can be adjusted at scripts--build/prep-build.py.