scriptconv 0.0.4a33__tar.gz → 0.0.4a34__tar.gz

{scriptconv-0.0.4a33/scriptconv.egg-info → scriptconv-0.0.4a34}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: scriptconv
-Version: 0.0.4a33
+Version: 0.0.4a34
 Summary: Zero-dependency script & phoneme-notation core — ISO-15924 detection & metadata, IPA↔ARPABET/X-SAMPA/Lexique/Kirshenbaum/Cotovía/RFE, Buckwalter↔Arabic, Hangul→jamo, kana
 License: Apache-2.0
 Project-URL: Homepage, https://github.com/TigreGotico/scriptconv

{scriptconv-0.0.4a33 → scriptconv-0.0.4a34}/scriptconv/phonemizers/mul.py

@@ -336,13 +336,19 @@ class EspeakPhonemizer(BasePhonemizer):
             "zh-tw": "cmn", "zh-hant": "cmn", "zh-hant-tw": "cmn",
             "zh-hk": "yue", "zh-mo": "yue",
         }
-        if target_lang.lower() in ZH_ALIASES:
-            return ZH_ALIASES[target_lang.lower()]
-        if target_lang in cls.ESPEAK_LANGS:
-            return target_lang
-        if target_lang.lower().split("-")[0] in cls.ESPEAK_LANGS:
-            return target_lang.lower().split("-")[0]
-        return cls.match_lang(target_lang, cls.ESPEAK_LANGS)
+        # ESPEAK_LANGS is all lowercase, so the tag is lowercased once and
+        # matched as a whole BEFORE the primary subtag. A BCP-47 tag such as
+        # pt-BR or fr-BE used to miss the exact check on case and fall to the
+        # bare pt or fr, which are European Portuguese and metropolitan
+        # French: a Brazilian voice phonemized as European.
+        tag = target_lang.lower().replace("_", "-")
+        if tag in ZH_ALIASES:
+            return ZH_ALIASES[tag]
+        if tag in cls.ESPEAK_LANGS:
+            return tag
+        if tag.split("-")[0] in cls.ESPEAK_LANGS:
+            return tag.split("-")[0]
+        return cls.match_lang(tag, cls.ESPEAK_LANGS)
 
     @staticmethod
     def _run_espeak_command(args: List[str], input_text: str = None, check: bool = True) -> str:

{scriptconv-0.0.4a33 → scriptconv-0.0.4a34}/scriptconv/version.py

@@ -2,7 +2,7 @@
 VERSION_MAJOR = 0
 VERSION_MINOR = 0
 VERSION_BUILD = 4
-VERSION_ALPHA = 33
+VERSION_ALPHA = 34
 # END_VERSION_BLOCK
 
 VERSION_STR = f"{VERSION_MAJOR}.{VERSION_MINOR}.{VERSION_BUILD}"

{scriptconv-0.0.4a33 → scriptconv-0.0.4a34/scriptconv.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: scriptconv
-Version: 0.0.4a33
+Version: 0.0.4a34
 Summary: Zero-dependency script & phoneme-notation core — ISO-15924 detection & metadata, IPA↔ARPABET/X-SAMPA/Lexique/Kirshenbaum/Cotovía/RFE, Buckwalter↔Arabic, Hangul→jamo, kana
 License: Apache-2.0
 Project-URL: Homepage, https://github.com/TigreGotico/scriptconv

{scriptconv-0.0.4a33 → scriptconv-0.0.4a34}/scriptconv.egg-info/SOURCES.txt

@@ -503,6 +503,7 @@ tests/test_conventions.py
 tests/test_diacritics.py
 tests/test_diacritics_graph.py
 tests/test_errors_policy.py
+tests/test_espeak_get_lang_case.py
 tests/test_examples.py
 tests/test_graph.py
 tests/test_he_extra_declares_phonikud.py

scriptconv-0.0.4a34/tests/test_espeak_get_lang_case.py

@@ -0,0 +1,58 @@
+"""EspeakPhonemizer.get_lang reads a BCP-47 tag whatever its case (T-2783).
+
+``ESPEAK_LANGS`` is all lowercase. The exact-match check used the tag as
+written, so ``pt-BR`` missed it and fell to the bare primary subtag ``pt``,
+which is European Portuguese: a phoonnx Brazilian voice phonemized as
+European. ``fr-BE`` fell to metropolitan ``fr`` the same way.
+
+These tests need no espeak binary: ``get_lang`` is a classmethod over the
+static voice list.
+"""
+import pytest
+
+from scriptconv.phonemizers.mul import EspeakPhonemizer
+
+
+@pytest.mark.parametrize("tag, voice", [
+    ("pt-BR", "pt-br"),
+    ("PT-BR", "pt-br"),
+    ("pt_BR", "pt-br"),
+    ("fr-BE", "fr-be"),
+    ("fr-CH", "fr-ch"),
+    ("ES-419", "es-419"),
+    ("vi-VN-x-central", "vi-vn-x-central"),
+    ("vi-VN-x-south", "vi-vn-x-south"),
+])
+def test_a_region_tag_keeps_its_region_whatever_its_case(tag, voice):
+    assert EspeakPhonemizer.get_lang(tag) == voice
+
+
+@pytest.mark.parametrize("tag, voice", [
+    ("pt-br", "pt-br"),
+    ("fr-be", "fr-be"),
+    ("en-us", "en-us"),
+    ("pt", "pt"),
+    ("fr", "fr"),
+])
+def test_lowercase_tags_are_unchanged(tag, voice):
+    assert EspeakPhonemizer.get_lang(tag) == voice
+
+
+def test_the_bare_language_is_not_silently_substituted_for_a_region():
+    # the defect: pt-BR must never resolve to European pt while pt-br exists
+    assert EspeakPhonemizer.get_lang("pt-BR") != "pt"
+    assert EspeakPhonemizer.get_lang("fr-BE") != "fr"
+
+
+@pytest.mark.parametrize("tag, voice", [
+    ("EN-GB", "en-gb-x-rp"),
+    ("zh-CN", "cmn"),
+    ("ZH-HK", "yue"),
+])
+def test_the_existing_aliases_still_apply_in_any_case(tag, voice):
+    assert EspeakPhonemizer.get_lang(tag) == voice
+
+
+def test_an_unknown_tag_still_raises():
+    with pytest.raises(ValueError):
+        EspeakPhonemizer.get_lang("xx")