serbian-translit

Deterministic Serbian and Montenegrin script conversion, Cyrillic ↔ Latin. Case preservation, digraph handling, quoted-region protection, Roman-numeral and non-native-word filtering.

Both official scripts of Serbian (and Montenegrin) map one-to-one at the letter level: љ↔lj, њ↔nj, џ↔dž, plus с́↔ś, з́↔ź for Montenegrin. The library plays the pairing from a YAML table; there is no per-language code path in the engine.

Installation

The library is distributed as git tags (not published to PyPI). Install from GitHub:

# latest release
pip install git+https://github.com/apakabarlabs/serbian-translit-python.git

# pin a specific version (see releases: https://github.com/apakabarlabs/serbian-translit-python/releases)
pip install git+https://github.com/apakabarlabs/serbian-translit-python.git@vX.Y.Z

Usage

from serbian_translit import srp, cnr

srp.to_cyr("Njujork")           # 'Њујорк'
srp.to_cyr("LJUBAV")            # 'ЉУБАВ'
srp.to_cyr("New York")          # 'New York' (word skipped, has non-native letters)
srp.to_cyr('grupa „AC/DC"')     # 'група „AC/DC"' (quoted region preserved)
srp.to_lat("Њујорк")            # 'Njujork'

cnr.to_cyr("śever")             # 'с́евер' (с + U+0301)
cnr.to_lat("с́евер")             # 'śever'

Behaviour

• Digraphs lj, nj, dž (Latin) ↔ љ, њ, џ (Cyrillic) with case preservation (Nj in title-case position, NJ inside all-caps).
• Montenegrin extras ś, ź ↔ с́, з́ (base letter + combining acute U+0301; no precomposed codepoints exist).
• Đ variants Đ (U+0110), đ (U+0111), Ð (U+00D0 Eth), ð (U+00F0 eth) all map to Ђ/ђ.
• Roman numerals (II, XIV, XX) stay in Latin regardless of direction.
• Words with non-native letters (Latin w, x, y, q) are skipped whole; treated as foreign inclusions.
• Quoted regions ("…", „…", “…”, «…») are preserved verbatim so brand names and foreign quotes survive round-trip.
• Non-alphabetic content (numbers, punctuation, whitespace) is left unchanged.

Rules and tests

Rules live in serbian_translit/data/rules.yaml; test cases in tests/tests.yaml. Both files are the source of truth shared with the Swift and (upcoming) Kotlin ports so behaviour stays identical across languages.

Lines of Code