Merge pull request #1099 from python-babel/issue-1098

Support list format fallbacks

Author: akx

babel/lists.py

@@ -26,9 +26,11 @@
 DEFAULT_LOCALE = default_locale()
 
 
-def format_list(lst: Sequence[str],
-                style: Literal['standard', 'standard-short', 'or', 'or-short', 'unit', 'unit-short', 'unit-narrow'] = 'standard',
-                locale: Locale | str | None = DEFAULT_LOCALE) -> str:
+def format_list(
+    lst: Sequence[str],
+    style: Literal['standard', 'standard-short', 'or', 'or-short', 'unit', 'unit-short', 'unit-narrow'] = 'standard',
+    locale: Locale | str | None = DEFAULT_LOCALE,
+) -> str:
     """
     Format the items in `lst` as a list.
 
@@ -39,7 +41,11 @@ def format_list(lst: Sequence[str],
     >>> format_list(['omena', 'peruna', 'aplari'], style='or', locale='fi')
     u'omena, peruna tai aplari'
 
-    These styles are defined, but not all are necessarily available in all locales.
+    Not all styles are necessarily available in all locales.
+    The function will attempt to fall back to replacement styles according to the rules
+    set forth in the CLDR root XML file, and raise a ValueError if no suitable replacement
+    can be found.
+
     The following text is verbatim from the Unicode TR35-49 spec [1].
 
     * standard:
@@ -76,14 +82,9 @@ def format_list(lst: Sequence[str],
     if len(lst) == 1:
         return lst[0]
 
-    if style not in locale.list_patterns:
-        raise ValueError(
-            f'Locale {locale} does not support list formatting style {style!r} '
-            f'(supported are {sorted(locale.list_patterns)})',
-        )
-    patterns = locale.list_patterns[style]
+    patterns = _resolve_list_style(locale, style)
 
-    if len(lst) == 2:
+    if len(lst) == 2 and '2' in patterns:
         return patterns['2'].format(*lst)
 
     result = patterns['start'].format(lst[0], lst[1])
@@ -92,3 +93,31 @@ def format_list(lst: Sequence[str],
     result = patterns['end'].format(result, lst[-1])
 
     return result
+
+
+# Based on CLDR 45's root.xml file's `<alias>`es.
+# The root file defines both `standard` and `or`,
+# so they're always available.
+# TODO: It would likely be better to use the
+#       babel.localedata.Alias mechanism for this,
+#       but I'm not quite sure how it's supposed to
+#       work with inheritance and data in the root.
+_style_fallbacks = {
+    "or-narrow": ["or-short", "or"],
+    "or-short": ["or"],
+    "standard-narrow": ["standard-short", "standard"],
+    "standard-short": ["standard"],
+    "unit": ["unit-short", "standard"],
+    "unit-narrow": ["unit-short", "unit", "standard"],
+    "unit-short": ["standard"],
+}
+
+
+def _resolve_list_style(locale: Locale, style: str):
+    for style in (style, *(_style_fallbacks.get(style, []))):  # noqa: B020
+        if style in locale.list_patterns:
+            return locale.list_patterns[style]
+    raise ValueError(
+        f"Locale {locale} does not support list formatting style {style!r} "
+        f"(supported are {sorted(locale.list_patterns)})",
+    )

scripts/import_cldr.py

@@ -530,10 +530,11 @@ def parse_locale_display_names(data, tree):
 
 def parse_list_patterns(data, tree):
     list_patterns = data.setdefault('list_patterns', {})
-    for listType in tree.findall('.//listPatterns/listPattern'):
-        by_type = list_patterns.setdefault(listType.attrib.get('type', 'standard'), {})
-        for listPattern in listType.findall('listPatternPart'):
-            by_type[listPattern.attrib['type']] = _text(listPattern)
+    for list_pattern_el in tree.findall('.//listPatterns/listPattern'):
+        pattern_type = list_pattern_el.attrib.get('type', 'standard')
+        for pattern_part_el in list_pattern_el.findall('listPatternPart'):
+            pattern_part_type = pattern_part_el.attrib['type']
+            list_patterns.setdefault(pattern_type, {})[pattern_part_type] = _text(pattern_part_el)
 
 
 def parse_dates(data, tree, sup, regions, territory):

tests/test_lists.py

@@ -1,20 +1,32 @@
 import pytest
 
-from babel import lists
+from babel import lists, units
 
 
-def test_format_list():
-    for list, locale, expected in [
-        ([], 'en', ''),
-        (['string'], 'en', 'string'),
-        (['string1', 'string2'], 'en', 'string1 and string2'),
-        (['string1', 'string2', 'string3'], 'en', 'string1, string2, and string3'),
-        (['string1', 'string2', 'string3'], 'zh', 'string1、string2和string3'),
-        (['string1', 'string2', 'string3', 'string4'], 'ne', 'string1,string2, string3 र string4'),
-    ]:
-        assert lists.format_list(list, locale=locale) == expected
+@pytest.mark.parametrize(('list', 'locale', 'expected'), [
+    ([], 'en', ''),
+    (['string'], 'en', 'string'),
+    (['string1', 'string2'], 'en', 'string1 and string2'),
+    (['string1', 'string2', 'string3'], 'en', 'string1, string2, and string3'),
+    (['string1', 'string2', 'string3'], 'zh', 'string1、string2和string3'),
+    (['string1', 'string2', 'string3', 'string4'], 'ne', 'string1,string2, string3 र string4'),
+])
+def test_format_list(list, locale, expected):
+    assert lists.format_list(list, locale=locale) == expected
 
 
 def test_format_list_error():
     with pytest.raises(ValueError):
         lists.format_list(['a', 'b', 'c'], style='orange', locale='en')
+
+
+def test_issue_1098():
+    one_foot = units.format_unit(1, "length-foot", length="short", locale="zh_CN")
+    five_inches = units.format_unit(5, "length-inch", length="short", locale="zh_CN")
+    # zh-CN does not specify the "unit" style, so we fall back to "unit-short" style.
+    assert (
+        lists.format_list([one_foot, five_inches], style="unit", locale="zh_CN") ==
+        lists.format_list([one_foot, five_inches], style="unit-short", locale="zh_CN") ==
+        # Translation verified using Google Translate. It would add more spacing, but the glyphs are correct.
+        "1英尺5英寸"
+    )