⍰ Unicode Tofu Review ⍰
Something nobody asked for but we'll do it anyway.
Tofu (named because it is usually white boxes) is what you see when there is no font support for the current glyph in your font stack, which can vary by family (serif/sans/mono) and application.
Something nobody asked for but we'll do it anyway.
Tofu (named because it is usually white boxes) is what you see when there is no font support for the current glyph in your font stack, which can vary by family (serif/sans/mono) and application.
The technical term is (in most cases) .notdef, in PostScript, TrueType/OpenType fonts, etc.
docs.microsoft.com/en-us/typograp…
Most of the time, what you see comes from the .notdef in one of your fontstack fonts. But, sometimes it is custom to the application.
docs.microsoft.com/en-us/typograp…
Most of the time, what you see comes from the .notdef in one of your fontstack fonts. But, sometimes it is custom to the application.
⍰
The most common multi-platform tofu lately seems to be the ubiquitous question block.
Cute, but not very informative. Surprised Nintendo hasn't sent DMCA's to every font using it.
[image via @emojipedia]
The most common multi-platform tofu lately seems to be the ubiquitous question block.
Cute, but not very informative. Surprised Nintendo hasn't sent DMCA's to every font using it.
[image via @emojipedia]
⛝
The dreaded X-box. Common on older Android phones lacking font/emoji support.
Ugly, not informative. Sense of impending dread. Surprised MicroSoft hasn't sent DMCA's to every font using it.
The dreaded X-box. Common on older Android phones lacking font/emoji support.
Ugly, not informative. Sense of impending dread. Surprised MicroSoft hasn't sent DMCA's to every font using it.
⎡FE⎤
⎣FF⎦
Hexagana. The hex code in a box. Usually not part of a font, but done procedurally in the application. Firefox text fields, for example.
There is a font to do this, though scripts.sil.org/cms/scripts/pa… for example.
A+. Very informative. If you have to tofu me, do this.
⎣FF⎦
Hexagana. The hex code in a box. Usually not part of a font, but done procedurally in the application. Firefox text fields, for example.
There is a font to do this, though scripts.sil.org/cms/scripts/pa… for example.
A+. Very informative. If you have to tofu me, do this.
⌷
Empty box. Probably the most common tofu on desktops.
Really boring. Who cares.
Empty box. Probably the most common tofu on desktops.
Really boring. Who cares.
䷀
Hexagram. Shows on iOS in some situations. Common during the "i+[FE0F]" autocomplete bug.
Kinda cool, but nothing to take home to meet the parents.
Hexagram. Shows on iOS in some situations. Common during the "i+[FE0F]" autocomplete bug.
Kinda cool, but nothing to take home to meet the parents.
👽
Aliens! Also iOS. Apple has a shitload of tofu variants. This one may have specifically been for missing emoji, but the glyph had another provenance...
Cute. Would let it abduct me.
Aliens! Also iOS. Apple has a shitload of tofu variants. This one may have specifically been for missing emoji, but the glyph had another provenance...
Cute. Would let it abduct me.
🄰
Last Resort. Originating on macOS as the bottom of the font stack, it would show a glyph representing each block.
Available here: unicode.org/policies/lastr… (by Michael Everson).
Not bad. Slightly more informative than all but Hexagana.
Last Resort. Originating on macOS as the bottom of the font stack, it would show a glyph representing each block.
Available here: unicode.org/policies/lastr… (by Michael Everson).
Not bad. Slightly more informative than all but Hexagana.
"⠀"
Nothing. Seen on some older Android phones.
Unsupported glyphs would just be... blank.
0/10. Fuck off with that.
Nothing. Seen on some older Android phones.
Unsupported glyphs would just be... blank.
0/10. Fuck off with that.
?
Bare question mark. Common on very old systems. Rarer now. Pre-Uncode Windows used to use it as a combination .notdef and Replacement Character.
Boring. Confusing. Hate.
Bare question mark. Common on very old systems. Rarer now. Pre-Uncode Windows used to use it as a combination .notdef and Replacement Character.
Boring. Confusing. Hate.
Finally this isn't tofu:
� FFFD REPLACEMENT CHARACTER
The source bytes did not reference Unicode codepoints.
Latin-1 parsed as UTF-8, well-formed references outside UTF-8's scope (Overbyte encoding/>10FFFF/split surrogates). etc
It supplants/replaces the illegal bytes. Lossy.
� FFFD REPLACEMENT CHARACTER
The source bytes did not reference Unicode codepoints.
Latin-1 parsed as UTF-8, well-formed references outside UTF-8's scope (Overbyte encoding/>10FFFF/split surrogates). etc
It supplants/replaces the illegal bytes. Lossy.
Bonus: This also isn't tofu or a replacement sign, but is so darned cute it gets honorable mention:
⯑ 2BD1 UNCERTAINTY SIGN
⯑ 2BD1 UNCERTAINTY SIGN
There is no Unicode codepoint for tofu. But there are non-characters [unicode.org/faq/private_us…] that are guaranteed never to be defined. And unlike Private Use Area ranges, they should never be assigned a glyph. Like U+10FFFF:
The most important one is U+FFFE (inverse BOM).
The most important one is U+FFFE (inverse BOM).
However, even non-characters can be misused. Asana-Math.otf for example incorrectly stored extra glyphs at the tail end of the Uncode range, dumping them in SPUA-B, and two of them hit U+10FFFE and U+10FFFF (non-characters).

There is another type of anti-tofu. When a codepoint has no glyph by default, it can appear as a dotted outline in the charts. This (mostly) occurs with Whitespace, Control Codes, and stuff with the property "Default_Ignorable".
unicode.org/Public/12.0.0/…
unicode.org/faq/unsup_char…
There is another type of anti-tofu. When a codepoint has no glyph by default, it can appear as a dotted outline in the charts. This (mostly) occurs with Whitespace, Control Codes, and stuff with the property "Default_Ignorable".
unicode.org/Public/12.0.0/…
unicode.org/faq/unsup_char…
