fix: sort rules without explicit indices

fixes: #299

Co-authored-by: Eric Joanis <eric.joanis@nrc-cnrc.gc.ca>

g2p/mappings/__init__.py

@@ -37,6 +37,7 @@
 )
 
 GEN_DIR = os.path.join(os.path.dirname(LANGS_FILE), "generated")
+EXPLICIT_INDEX_PATTERN = re.compile(r"{\d+}")
 
 
 class Mapping(_MappingModelDefinition):
@@ -153,7 +154,9 @@ def process_model_specs(self):  # noqa: C901
             self.rules = sorted(
                 # Temporarily normalize to NFD for heuristic sorting of NFC-defined rules
                 self.rules,
-                key=lambda x: len(normalize(x.rule_input, "NFD"))
+                key=lambda x: len(
+                    normalize(re.sub(EXPLICIT_INDEX_PATTERN, "", x.rule_input), "NFD")
+                )
                 if isinstance(x, Rule)
                 else len(normalize(x["in"], "NFD")),
                 reverse=True,
@@ -241,7 +244,7 @@ def rule_to_regex(self, rule: Union[Rule, dict]) -> Union[Pattern, None]:
                 "This is disallowed. Please check your mapping file for rules with null inputs."
             )
             return None
-        input_match = re.sub(re.compile(r"{\d+}"), "", rule.rule_input)
+        input_match = re.sub(EXPLICIT_INDEX_PATTERN, "", rule.rule_input)
         try:
             inp = create_fixed_width_lookbehind(rule.context_before) + input_match
             if rule.context_after:

g2p/tests/test_mappings.py

@@ -164,6 +164,13 @@ def test_rule_ordering(self):
         transducer_default = Transducer(Mapping(rules=rules))
         self.assertEqual(transducer_default("aa").output_string, "bb")
 
+    def test_rule_ordering_with_indices(self):
+        """a{1}b{3} should be shorter than abc"""
+        rules = [{"in": "a{1}b{2}", "out": "x{1}x{2}"}, {"in": "abc", "out": "y"}]
+        mapping = Mapping(rules=rules, rule_ordering="apply-longest-first")
+        transducer = Transducer(mapping)
+        self.assertEqual(transducer("abc").output_string, "y")
+
     def test_rule_ordering_given_invalid_value(self):
         """
         It should log an error messages if given an invalid value for