Chinese can be sorted by pinyin, stroke count, or radical

What it is

Chinese collation methods do not reduce to one alphabet for Han characters. The same names can be organized by Mandarin pinyin, total stroke count, or radical followed by residual strokes; products may also use Bopomofo, shape-based codes, confirmed readings for proper names, or an institutional directory convention. CLDR supplies several Chinese collation types and index markers for multiple orders. Its fallback commonly selects pinyin for a Simplified Chinese locale and stroke order for a Traditional Chinese locale. Those are CLDR defaults, not proof that one order is universally preferred by every Chinese reader, region, or task.

Why it happens

Pinyin collation maps Han characters to pronunciation weights and works well when readers know the Mandarin reading. Polyphonic characters, name-specific readings, and non-Mandarin pronunciations make automatic inference from glyphs ambiguous. Stroke order uses total strokes and further tie-breakers, offering a visual route when pronunciation is unknown, but it must adopt a consistent glyph tradition and counting convention. Radical–stroke order uses Unihan radical-stroke data, sorting by radical and then residual strokes; it supports dictionary-like lookup but assumes that readers can identify the selected radical. These methods produce different neighbors and index labels, so visible order, bucket boundaries, and navigation labels must come from the same tailoring. Search may accept Han text, pinyin, and other transliterations, while record identity still requires a stable ID; neither a pinyin string nor a radical-stroke code is a safe primary key.

Studying it

State the target population, region, glyph tradition, and lookup task before comparing methods. Ask participants to locate contacts, places, or terms under pinyin, stroke, and radical–stroke order; measure time to first correct selection, wrong buckets, search fallback, and whether the reader knew the pronunciation or radical. Analyze polyphonic characters, rare characters, simplified/traditional variants, and non-Chinese content separately. Engineering studies can version a full-order fixture and index markers against the chosen CLDR tailoring, with override tests for proper names whose reading was confirmed by a person. Results do not transfer automatically between dictionary-trained readers and people scanning a mobile contact list.

Where it stops holding

Pinyin, stroke, and radical–stroke are alternatives, not lossless transformations or a universal ranking of quality. Readings depend on word context, topolect, and naming convention; automatic transliteration without context or curated metadata can place a record in an unexpected bucket. Stroke and radical data depend on glyph tradition and data version, so simplified, traditional, and variant forms must not be guessed from code-point order. Mixed lists containing Latin text, digits, kana, or emoji also need an explicit script-order and overflow-bucket policy. A legal register or established printed index may impose a convention that should be documented and tested with its actual users.

Applying it

• Offer clearly named choices such as Pinyin, Stroke, and Radical–Stroke based on the content and lookup task. If only one is offered, document its locale, glyph tradition, and tailoring instead of calling it the Chinese default.
• Use a maintained Unicode/CLDR collation implementation rather than code-point order, UTF-8 bytes, a handmade initial table, or a partial Han mapping. Allow confirmed sort readings for proper names while preserving original text for display.
• Generate buckets from markers belonging to the active collation: A–Z boundaries for pinyin, stroke-count labels for stroke order, and radical labels for radical–stroke order. A mode switch must reorder records, rebuild buckets, and update navigation state together.
• Treat the original Han spelling and a stable record ID as authoritative. Keep readings, radicals, and stroke counts as rebuildable derived search data, and test polyphonic and rare characters, simplified/traditional variants, mixed scripts, and empty buckets.

Related

• Same group: S2.08.1 Alphabetical order varies by language · S2.08.3 Case and diacritics affect collation results · S2.08.4 Index navigation must switch with language
• Adjacent: S1.03.2 The same character can have different glyph forms across regions · S2.05.1 The order and number of name parts are not fixed
• Search terms: Chinese pinyin collation · stroke collation · Unihan radical-stroke