What is Unicodia?

It is a simple Unicode encyclopedia and the most comprehensive character map ever. Right now Windows only.

Lifecycle phase: 5 (production/stable). Minor troubles with sustainability, but generally survived five Unicode releases, 14.0 to 17.0. But this status is on the verge, just because…

I’m a Ukrainian officer now. Everything may happen: maybe I’ll start the work again. Maybe I’ll just get killed.

No vibecoding, at least while I’m alive.

Note about the Egyptian font

It has been moved to a separate repo. Visit https://github.com/Mercury13/unicodia-sesh

How to get portable

I was asked several times, but by this time, it had already been portable.

Open Unicodia.xml, it’s documented.

Privacy policy

Unicodia does not collect data at all, but uses GitHub API for updates.

How to translate?

• Ask the programmer to add localized buttons if needed. One button is international for now, A-Z, and it already has Cyrillic, Katakana and Chinese versions. The rest are unchangeable for now… until needed.
• Download Lang-src/en.uorig from this repo.
  • If you are able to use Git, better use it. We’ll be able to work together on one translation.
• Put Unicodia in a writeable location.
• Create a language directory, edit locale.xml for that language.
• Download UTranslator. New → Translation of *.uorig.
  • If you don’t know English, use another *.utran file as a reference translation.
  • Keys Ctrl+↕ are not documented, but greatly reduce mouse use.
• After saving, UTranslator created build-xx\lang.ini. Put it in the language directory. Or use a symlinking tool to link these files forever and avoid handwork. Or write <debug ini="c:\path\to\build-xx\lang.ini"> in locale.xml.
• Press F12 in Unicodia to reload translation without reopening the entire program.
  • Warning: it reloads strings only; all locales are loaded on startup.
• nspk template parameters: 1=language name (or script name, non-localizable), 2=pre-comment (e.g. synonym, localizable).
  • If there’s no {{nspk}} in languages and there’s language data, the default {{nspk}} is added automatically. So: {{nspk}} at the end → delete, it’ll be added! Need e.g. synonym → add {{nspk||=Klingon}}. Synonym is the SECOND parameter. See Script.Mroo in English/Russian.
• To test alphabetic sorting, especially in troublesome languages like Japanese: press Ct+Sh+W and look into the Blocks drop-down list (does not work in Sort by tech name). There’s only one telltale, [1] when the 1st character does not belong to the sorting alphabet. These [1]’s are often mistakes and always signs of attention.
• When a new original has arrived: open the translation, press File → Update data (Ctrl+F5). The interface will guide you. The command Go → Find warnings → All (F6) can also help.
  • Do not forget to reset the red eye/warning icon when the translation finally reflects the new original! Either double-click, or Ctrl+Enter.

Language policy

Common. No war jargon. Describe the 2022 war as neutrally as possible. Every lingua franca (English, Russian, French) is in its international form. Make examples as patriotic as possible for the language we’re writing in: the same letter is Russian and Ukrainian in respective L10n’s. And English if the same phenomenon exists in the English language. Apostrophe is U+2019.

Is Old in the front or in the back? It depends. 1) In Scripts — as convenient (they are not gathered to a large list). In Blocks… 2) Old is the main word (Ancient symbols) → better front. 3) Auxiliary block (Old Sogdian, Ancient Greek) → no matter, we’ll find it anyway by looking around Sogdian/Greek. 4) Old is an adjective to something more important (Italic old, Mongolian old, Permic old) → better back. It’s just the ease of finding a block in the long list of 300 blocks.

AI as a translator. Allowed, but at least check it somehow.

English. The dialect called “International English” or “English as a lingua franca”: use the best word for a non-native audience. Examples: truck > lorry, petrol > gas, -ize > -ise. Prefer the British form if both are good. Punctuation around quotes is British/international: it’s inside the quotes if it’s part of the “phrase being quoted”.

Though the grammar must be close to the British/American origin, with articles and tenses. Unless you are native/proficient, every new sufficiently large text must be grammar-checked with Grammarly or a similar AI tool.

Russian. Ё is mandatory. No grammatical concessions to Ukrainian.

(May apply to new languages as well.) Adjectives like Georgian may agree with script (письменность, female in Russian), or with language (язык, male). The rules are…

• BLOCKS: strongly connected to language → to language (грузинский=Georgian [language]). Otherwise, to script (батакская=Batak [script]).
• SCRIPTS: of course agree to script (грузинская=Georgian [script]).

Ukrainian. See Lang-src/Ukrainian.md.

New languages.

• As English uses lots of capital letters, translations into other languages may use small where English is capitalized. Refer to Russian/Ukrainian for letter case.
• See the Russian script/language rule.

About war jargon. Open-source software with a neutral license and without a special purpose (e.g. censorship circumvention) should be neutral. Period.

How to build?

• Slight C++20 and std::filesystem here → so need either MSYS or recent Qt with MinGW 11.
• Also need cURL (present in W10 18H2+), 7-zip, UTransCon, SvgCleaner.
• Run !rel.bat file.
  • If there are troubles with paths, make a local configuration by creating ~setup_local.bat. Write only the keys that are bad in the default xsetup.bat.

How to develop?

See develop.md.

Compatibility and policies

Platforms

Win7/10/11 x64 only. Rationale:

• WXP, WVista and W8 are completely abandoned by all imaginable software. Though I did some improvements specially for W8.
• No obstacles for x86, just untested because no one compiled Qt for x86.
• Though W11 is not the main OS, I did many W11-specific improvements.

Tofu/misrenderings

• W10/11 should support everything possible, W7 just runs somehow. At the time of testing, still no BMP tofu, per old policy.
• Previously, W7 supported the entire base plane and three important plane 1 scripts. I dropped that guarantee, though I did nothing against it, just did not test
• Small misrenderings in descriptions are tolerable, I’ll fix them only if samples are bad or if the font has other problems.

Update Unicode

Wartime: as soon as the base arrives and the release date is frozen, even on alpha review stage

Peacetime (probably): stable release + some big font covering a major set arrives. Han too, if the coverage is really high

Emergency releases of a few characters (e.g. currency, Japanese era): instantly, even if they are tofu

Fonts

Fonts are always updated to release versions. A font is updated to alpha/beta if it fixes a major misrender, and/or professionally implements a new character.

Naming: Noto if tables and existing glyphs are surely untouched; Uto otherwise.

These fonts are taken to Unicodia without the author’s consent:

• Craggy font with missing/trivial tables. Examples: Garay, Tolong Siki
• Font without a license, belonging to the author of Unicode request and released by him/her. Examples: Makasar (now replaced), Tangsa
• The author is surely SIL, even if found elsewhere. Example: Toto
• Incomplete fonts with only a few fixup characters, if found in requests in TTF format. Example: a few rare CJK chars.
• ASCII mapping is NOT a reason. Example: Ol Onal (wrote own)

I never rip fonts from Unicode charts, always use a TTF form. But the authors of fixup fonts can do that on completely unrelated characters. Examples: PlanGothic P2 (now probably OK).

CJK

The only person I could ask about ideographs has died. Let these rules be for history, maybe I’ll coin others.

• Serif style > correctness
• One country is enough
• Preference of countries
  • confirmed Chinese (G)
  • = modernized confirmed Chinese
  • > confirmed other (J, K, T, H, M, KP, V)
  • > hypothetical Chinese, country-independent (JV)
• It’s perfectly OK to take hypothetical Chinese if it’s wrong in…
  • SimSun makes the same decision
  • stroke types and stroke joins only (what is invisible or barely seen in sans style), even pointy vs dot
  • whether the strokes leave a small gap or are written together, even two crosses (T) vs a horizontal dagger (G)
  • whether a stroke is convex up or down
  • minor difference of stroke length e.g. in “three”
  • style of roof stroke e.g. in 2F34: straight S (31D1, older) vs backslash dot D (31D4, current)
  • the number of strokes is off by one e.g. in 2E3D9: one dot in Chinese, two in SAT

Anyway, Unicodia will never be a good ideograph guide. Everything I write about ideographs I suck from other sources.

Data

Data is as neutral as possible. Examples.

• Number of people speaking Russian. Its current status is Lingua franca, so the # of L2 speakers is always shaky, especially amid the current world war. So just # of L1 speakers
• Number of people speaking Ukrainian. Amid this war, people tend to conceal their native Russian, and Ukrainian’s status is Alive, so # of L1 speakers is ⪢ L2 → so the total number
• Disputable territory. Disputable, who currently controls, and maybe who is disputing

Future functionality

• Finish GlyphWiki loader.
• Better CJK reference.
• Plane map.