_registry

Lazy lexer registry for Rosettes.

All lexers are hand-written state machines with O(n) guaranteed performance and zero ReDoS vulnerability. Lexers are loaded on-demand using functools.cache for thread-safe memoization.

Design Philosophy:

The registry uses lazy loading with caching to balance startup time and runtime performance:

1. Zero startup cost: No lexers imported at module load time
2. O(1) lookup: Pre-computed alias table for instant name resolution
3. Single instance: functools.cache ensures one lexer per language
4. Thread-safe: cache is thread-safe; lexers are stateless

Architecture:

• _LEXER_SPECS: Static registry mapping names to (module, class) specs
• _ALIAS_TO_NAME: Pre-computed case-insensitive alias lookup table
• _get_lexer_by_canonical: Cached lexer instantiation (one per language)

Performance Notes:

• First call: ~1ms (module import + class instantiation)
• Subsequent calls: ~100ns (dict lookup + cache hit)
• Memory: ~500 bytes per loaded lexer

Common Mistakes:

# ❌ WRONG: Caching lexer instances yourself
lexer_cache = {}
if lang not in lexer_cache:
    lexer_cache[lang] = get_lexer(lang)

# ✅ CORRECT: Just call get_lexer() — it's already cached
lexer = get_lexer(lang)

# ❌ WRONG: Checking support by catching exceptions
try:
    lexer = get_lexer(lang)
except LookupError:
    lexer = None

# ✅ CORRECT: Use supports_language() for checks
if supports_language(lang):
    lexer = get_lexer(lang)

Adding New Languages:

To add a new language, create a lexer inrosettes/lexers/and add an entry to_LEXER_SPECSbelow. Seerosettes/lexers/_state_machine.pyfor the base class and helper functions.

See Also:

• rosettes.lexers._state_machine: Base class for lexer implementations
• rosettes._protocol.Lexer: Protocol that all lexers must satisfy
• rosettes._formatter_registry: Similar pattern for formatters