scriptconv 0.0.4a33__tar.gz → 0.0.4a35__tar.gz

{scriptconv-0.0.4a33/scriptconv.egg-info → scriptconv-0.0.4a35}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: scriptconv
-Version: 0.0.4a33
+Version: 0.0.4a35
 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-0.0.4a35}/scriptconv/phonemizers/africa.py

@@ -90,18 +90,41 @@ class AfricaG2PPhonemizer(BasePhonemizer):
     @classmethod
     def get_lang(cls, target_lang: str) -> str:
         """
-        Resolve *target_lang* to a supported africa-g2p ISO 639-3 code.
-
-        Only the primary subtag is matched (e.g. ``tw-GH`` -> ``twi`` is *not*
-        attempted — africa-g2p keys its rule files by exact ISO 639-3 code, not
-        BCP-47 macrolanguage/region mapping, so this is an exact-match lookup,
-        not the closest-match fuzzing :meth:`BasePhonemizer.match_lang` does).
+        Resolve *target_lang* to a supported africa-g2p rule-file name.
+
+        Two exact lookups, in order, and no fuzzing (africa-g2p keys its rule
+        files by name, not by BCP-47 macrolanguage/region mapping, so
+        :meth:`BasePhonemizer.match_lang`'s closest-match search does not
+        apply):
+
+        1. the whole tag, lowercased, first as written and then with ``_``
+           read as ``-``. 11 of the rule
+           files carry a region suffix rather than a bare ISO 639-3 code
+           (``hau-nigeria``, ``hau-niger``, ``dop-benin``, ``dop-benin2``,
+           ``ngb-zaire``, ``ngb-zaire2``, ``sag-congo``,
+           ``sag-central_african_republic``, ``sef-cote_d_ivoire``,
+           ``sef-cote_d_ivoire2``, ``snk-senegal``), and before this step none
+           of them could be reached at all: the primary subtag ``hau`` is not
+           a rule file, so Hausa raised.
+        2. the primary subtag (``twi-GH`` -> ``twi``).
+
+        A bare code whose only rule files carry region suffixes still raises.
+        Picking one region for ``hau`` would be a claim about which variety a
+        caller meant, and this wrapper does not make it.
 
         Raises:
             ValueError: If africa-g2p has no rule file for *target_lang*.
         """
+        langs = cls.supported_langs()
+        lowered = target_lang.lower()
+        # Two of the names keep an underscore of their own
+        # (sef-cote_d_ivoire, sag-central_african_republic), so the tag is
+        # tried as written before ``_`` is read as ``-``.
+        for whole in (lowered, lowered.replace("_", "-")):
+            if whole in langs:
+                return whole
         key = _primary_subtag(target_lang)
-        if key in cls.supported_langs():
+        if key in langs:
             return key
         raise ValueError(f"africa-g2p: unsupported language {target_lang!r}")
 

{scriptconv-0.0.4a33 → scriptconv-0.0.4a35}/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-0.0.4a35}/scriptconv/version.py

@@ -2,7 +2,7 @@
 VERSION_MAJOR = 0
 VERSION_MINOR = 0
 VERSION_BUILD = 4
-VERSION_ALPHA = 33
+VERSION_ALPHA = 35
 # END_VERSION_BLOCK
 
 VERSION_STR = f"{VERSION_MAJOR}.{VERSION_MINOR}.{VERSION_BUILD}"

{scriptconv-0.0.4a33 → scriptconv-0.0.4a35/scriptconv.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: scriptconv
-Version: 0.0.4a33
+Version: 0.0.4a35
 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-0.0.4a35}/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-0.0.4a35/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")

{scriptconv-0.0.4a33 → scriptconv-0.0.4a35}/tests/test_phonemizers_africa.py

@@ -47,6 +47,30 @@ class TestAfricaG2PPhonemizer(unittest.TestCase):
         self.assertIn("twi", AfricaG2PPhonemizer.supported_langs())
         self.assertGreater(len(AfricaG2PPhonemizer.supported_langs()), 300)
 
+    def test_region_suffixed_rule_files_resolve_by_their_own_name(self):
+        # 11 rule files carry a region suffix instead of a bare ISO 639-3
+        # code; each must be selectable by that name (T-2701).
+        for code in ("hau-nigeria", "hau-niger", "dop-benin", "ngb-zaire",
+                     "sag-congo", "sef-cote_d_ivoire", "snk-senegal"):
+            self.assertEqual(AfricaG2PPhonemizer.get_lang(code), code)
+            self.assertIn(code, AfricaG2PPhonemizer.supported_langs())
+
+    def test_a_region_suffixed_code_phonemizes(self):
+        p = AfricaG2PPhonemizer(alphabet=Alphabet.IPA)
+        self.assertEqual(p.phonemize_string("ƙasa", "hau-nigeria"),
+                         "kʼ a s a")
+
+    def test_an_underscore_is_read_as_a_hyphen(self):
+        self.assertEqual(AfricaG2PPhonemizer.get_lang("HAU_NIGERIA"),
+                         "hau-nigeria")
+
+    def test_a_bare_code_with_only_region_files_still_raises(self):
+        # africa-g2p ships no bare 'hau'. Choosing a region for the caller
+        # would be a claim this wrapper does not make.
+        self.assertNotIn("hau", AfricaG2PPhonemizer.supported_langs())
+        with self.assertRaises(ValueError):
+            AfricaG2PPhonemizer.get_lang("hau")
+
     def test_get_lang_resolves_exact_iso_639_3_code(self):
         self.assertEqual(AfricaG2PPhonemizer.get_lang("twi"), "twi")
         # region subtag is stripped to the primary subtag before lookup