SharpMoji

Emoji data for .NET: lookup, localized search, skin tones and canonical ordering — in 28 languages, with no network access and no file I/O.

Status: not yet released. The API is complete and tested. Hebrew support and the 1.0 release are outstanding. See the specification.

SharpMoji is the data layer for Sourire (accessible emoji input) and a general-purpose emoji package for the .NET ecosystem.

Why another emoji library

Everything is in the package. There is no HTTP client, no cache directory, no storage abstraction and no first-run download — all 28 languages are embedded in about 1.4 MB. Nothing can fail at run time, nothing phones home, and it works offline, in a sandbox, and in an air-gapped build.

It is also correct about the parts that are easy to get wrong; see Details that bite.

Install

dotnet add package Oire.SharpMoji

Targets .NET 10. Trimming- and NativeAOT-compatible.

Usage

using Oire.SharpMoji;

var emoji = EmojiCatalog.English;

emoji.Find("👍")?.Label;              // "thumbs up" — with or without the variation selector
emoji.FindByHexcode("1F600")?.Label;  // "grinning face"
emoji.FindByShortcode(":+1:")?.Label; // "thumbs up"
emoji.FindByEmoticon(":D")?.Label;    // "grinning face with smiling eyes"

Catalogs are immutable, cached per language, and safe to share across threads. Every member is synchronous: the data is embedded, so there is no I/O to await.

Browsing

foreach (var group in emoji.Groups.Where(g => !g.IsComponent)) {
    Console.WriteLine($"{group.Name}: {emoji.GetByGroup(group).Length} emoji");
}

All is in canonical display order and excludes components — the bare skin-tone and hair modifiers, which are building blocks rather than anything a person picks.

Search

foreach (var match in emoji.Search("thum", limit: 5)) {
    Console.WriteLine($"{match.Emoji.Sequence}  {match.Emoji.Label}  ({match.Kind})");
}
// 👍  thumbs up     (LabelPrefix)
// 👎  thumbs down   (LabelPrefix)

Searches labels, tags, shortcodes and emoticons. Case and diacritics are ignored, so cafe finds café, and :-) finds :). Each result reports why it matched, so a UI can group or highlight accordingly.

Skin tones

emoji.GetSkinToneSlots("👍");   // 1
emoji.GetSkinToneSlots("🤝");   // 2 — two people, toned independently
emoji.GetSkinToneSlots("😀");   // 0

emoji.TryGetSkin("👍", SkinTone.Medium, out var skin);
skin.Sequence;   // "👍🏽"
skin.Label;      // "thumbs up: medium skin tone"

// Nineteen emoji take two modifiers. Pass both; matching tones resolve correctly.
emoji.TryGetSkin("🤝", SkinTone.Light, SkinTone.MediumLight, out var pair);

Other languages

var uk = EmojiCatalog.Load("uk");

uk.FindByHexcode("1F44D")?.Label;   // "великі пальці вгору"
uk.Groups[1].Name;                  // "люди"
uk.Search("серце", 5);              // works in Ukrainian
uk.FindByShortcode(":+1:");         // shortcodes work in every language

EmojiLocale.All;                    // all 28, with English and native names

Details that bite

These are the things a naive emoji library gets wrong. Each is measured rather than assumed — the numbers are in the specification.

• Variation selectors. 517 of 1949 records store a sequence ending in U+FE0F, while a pasted emoji usually does not. Find normalizes both sides, so two spellings that look identical resolve to one record.
• Two-tone emoji. Nineteen emoji depict two people who take independent skin tones, giving 25 variants each. Matching tones are stored with a single modifier rather than a repeated pair, so a lookup searching for [Light, Light] would silently lose five variants of every one.
• Localized categories. Group names come from per-language data, so a Ukrainian picker gets Ukrainian headings rather than English ones above Ukrainian emoji names.
• Stable keys. Hexcode is stable; Order renumbers whenever Unicode inserts emoji. Persist hexcodes for favorites and recents, never positions.

Documentation

• Technical specification — data model, API, measurements and design decisions
• Changelog

Data source

Emoji data comes from Emojibase (pinned to 17.0.0), which derives from the Unicode CLDR. It is regenerated by a build-time script and committed, so building from source needs neither a network connection nor the upstream corpus.

License

Code is licensed under the Apache License 2.0.

Embedded emoji data is not covered by that license: it is MIT (Emojibase) and Unicode-3.0 (CLDR). See NOTICE for the required attributions, which must accompany any redistribution.

---

Copyright © 2026 André Polykanine, Oire Software