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

Install the released package from PyPI:

pip install serbian-translit

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.

Documentation

The API reference is generated from the public Python API and deployed by GitHub Actions.

Lines of Code