Merge pull request #928 from arnoldwender/fix/i18n-lang-case-insensitive

fix(i18n): resolve language codes case-insensitively (#927)

mempalace/i18n/__init__.py

@@ -15,6 +15,7 @@ and the README section "Adding a new language" for the schema.
 
 import json
 from pathlib import Path
+from typing import Optional
 
 _LANG_DIR = Path(__file__).parent
 _strings: dict = {}
@@ -24,6 +25,23 @@ _current_lang: str = "en"
 _entity_cache: dict = {}
 
 
+def _canonical_lang(lang: str) -> Optional[str]:
+    """Resolve a language code to its on-disk canonical filename stem.
+
+    BCP 47 tags are case-insensitive (RFC 5646 §2.1.1), and the locale
+    files mix conventions (``pt-br.json`` vs ``zh-CN.json``). Match on
+    lowercase so callers can pass ``PT-BR``, ``zh-cn``, ``Pt-Br``, etc.
+    Returns ``None`` if no file matches.
+    """
+    if not lang:
+        return None
+    target = lang.strip().lower()
+    for path in _LANG_DIR.glob("*.json"):
+        if path.stem.lower() == target:
+            return path.stem
+    return None
+
+
 def available_languages() -> list[str]:
     """Return list of available language codes."""
     return sorted(p.stem for p in _LANG_DIR.glob("*.json"))
@@ -32,12 +50,12 @@ def available_languages() -> list[str]:
 def load_lang(lang: str = "en") -> dict:
     """Load a language dictionary. Falls back to English if not found."""
     global _strings, _current_lang
-    lang_file = _LANG_DIR / f"{lang}.json"
-    if not lang_file.exists():
-        lang_file = _LANG_DIR / "en.json"
-        lang = "en"
+    canonical = _canonical_lang(lang)
+    if canonical is None:
+        canonical = "en"
+    lang_file = _LANG_DIR / f"{canonical}.json"
     _strings = json.loads(lang_file.read_text(encoding="utf-8"))
-    _current_lang = lang
+    _current_lang = canonical
     return _strings
 
 
@@ -81,9 +99,10 @@ def get_regex() -> dict:
 
 def _load_entity_section(lang: str) -> dict:
     """Load the raw entity section for one language. Returns {} if missing."""
-    lang_file = _LANG_DIR / f"{lang}.json"
-    if not lang_file.exists():
+    canonical = _canonical_lang(lang)
+    if canonical is None:
         return {}
+    lang_file = _LANG_DIR / f"{canonical}.json"
     try:
         data = json.loads(lang_file.read_text(encoding="utf-8"))
     except (json.JSONDecodeError, OSError):
@@ -205,7 +224,12 @@ def get_entity_patterns(languages=("en",)) -> dict:
     """
     if not languages:
         languages = ("en",)
-    key = tuple(languages)
+    # Normalize via canonical filename so callers using different casing
+    # (e.g. "PT-BR" vs "pt-br") share the same cache entry and load the
+    # same locale file. Unknown codes are kept as-is so the merge loop's
+    # "found_any" branch fires the English fallback exactly once.
+    languages = tuple(_canonical_lang(lang) or lang for lang in languages)
+    key = languages
     if key in _entity_cache:
         return _entity_cache[key]
 

tests/test_i18n_lang_case.py

@@ -0,0 +1,86 @@
+"""Regression tests for issue #927 — language code lookup must be case-insensitive.
+
+The locale files use mixed case for the region subtag (``pt-br.json`` vs
+``zh-CN.json``). BCP 47 tags are case-insensitive (RFC 5646 §2.1.1), so
+``--lang PT-BR``, ``--lang zh-cn``, and ``--lang ZH-TW`` must all resolve
+to the canonical file rather than silently falling back to English.
+"""
+
+import pytest
+
+from mempalace import i18n
+from mempalace.i18n import (
+    _canonical_lang,
+    _load_entity_section,
+    available_languages,
+    get_entity_patterns,
+    load_lang,
+)
+
+
+@pytest.fixture(autouse=True)
+def _reset_state():
+    """Reset the module-level entity cache between tests."""
+    i18n._entity_cache.clear()
+    yield
+    i18n._entity_cache.clear()
+
+
+def test_canonical_lang_lowercase_passthrough():
+    assert _canonical_lang("en") == "en"
+    assert _canonical_lang("pt-br") == "pt-br"
+
+
+def test_canonical_lang_uppercase_resolves():
+    assert _canonical_lang("PT-BR") == "pt-br"
+    assert _canonical_lang("ZH-CN") == "zh-CN"
+    assert _canonical_lang("zh-cn") == "zh-CN"
+    assert _canonical_lang("Pt-Br") == "pt-br"
+
+
+def test_canonical_lang_unknown_returns_none():
+    assert _canonical_lang("xx") is None
+    assert _canonical_lang("") is None
+
+
+def test_load_lang_case_insensitive():
+    """`load_lang('PT-BR')` must load the pt-br dictionary, not English."""
+    en = load_lang("en")
+    pt_lower = load_lang("pt-br")
+    pt_upper = load_lang("PT-BR")
+    assert pt_lower == pt_upper, "case should not change the loaded dict"
+    # If load_lang silently fell back to English, both would equal `en`.
+    if "pt-br" in available_languages() and pt_lower != en:
+        assert i18n.current_lang() == "pt-br"
+
+
+def test_entity_section_loads_for_uppercase_input():
+    """`_load_entity_section('PT-BR')` must read pt-br.json, not return {}."""
+    pt_lower = _load_entity_section("pt-br")
+    pt_upper = _load_entity_section("PT-BR")
+    assert pt_lower == pt_upper
+
+
+def test_get_entity_patterns_case_insensitive():
+    """Entity patterns must be identical regardless of input case."""
+    lower = get_entity_patterns(("pt-br",))
+    upper = get_entity_patterns(("PT-BR",))
+    assert lower == upper
+
+
+def test_get_entity_patterns_shares_cache_across_cases():
+    """Different casing must hit the same cache entry — not duplicate work."""
+    get_entity_patterns(("zh-CN",))
+    cache_keys = list(i18n._entity_cache.keys())
+    get_entity_patterns(("ZH-CN",))
+    get_entity_patterns(("zh-cn",))
+    assert len(i18n._entity_cache) == len(
+        cache_keys
+    ), "different casings of the same language must not create new cache entries"
+
+
+def test_unknown_language_still_falls_back_to_english():
+    """A code with no matching file must fall through to English (existing contract)."""
+    patterns = get_entity_patterns(("xx-yy",))
+    en = get_entity_patterns(("en",))
+    assert patterns["candidate_patterns"] == en["candidate_patterns"]