Differentiate synonyms by priority

Author: cmutel

flowmapper/match.py

@@ -9,6 +9,10 @@
     rm_parentheses_roman_numerals,
     rm_roman_numerals_ionic_state,
 )
+from flowmapper.preferred_synonyms import (
+    match_identical_names_in_preferred_synonyms,
+    match_identical_names_in_synonyms,
+)
 
 logger = logging.getLogger(__name__)
 
@@ -26,17 +30,6 @@ def match_identical_identifier(s: Flow, t: Flow, comment: str = "Identical ident
         return {"comment": comment}
 
 
-def match_identical_names_in_synonyms(
-    s: Flow, t: Flow, comment: str = "Identical synonyms"
-):
-    if (
-        (t.synonyms and s.name in t.synonyms and s.context == t.context)
-        or (s.synonyms and t.name in s.synonyms and s.context == t.context)
-        # and not math.isnan(s.unit.conversion_factor(t.unit)):
-    ):
-        return {"comment": comment}
-
-
 def match_identical_cas_numbers(
     s: Flow, t: Flow, comment: str = "Identical CAS numbers"
 ):
@@ -247,5 +240,6 @@ def match_rules():
         match_identical_cas_numbers,
         match_non_ionic_state,
         match_biogenic_to_non_fossil,
+        match_identical_names_in_preferred_synonyms,
         match_identical_names_in_synonyms,
     ]

flowmapper/preferred_synonyms.py

@@ -0,0 +1,60 @@
+import re
+
+from flowmapper.flow import Flow
+
+ROMAN_NUMERAL_PATTERN = re.compile(r"\b\(?[ivx]+[\+-]?\)?\s*$", flags=re.IGNORECASE)
+PARENTHESES_PATTERN = re.compile(r"\([1-9]+[\+-]?\)\s*$")
+
+
+def has_roman_numeral_at_end(text: str) -> bool:
+    """
+    Check if a string ends with a roman numeral.
+
+    Args:
+        text (str): The string to check
+
+    Returns:
+        bool: True if the string ends with a roman numeral, False otherwise
+
+    """
+    return bool(ROMAN_NUMERAL_PATTERN.search(text))
+
+
+def has_number_pattern_at_end(text: str) -> bool:
+    """
+    Check if a string ends with a pattern like "(2+)".
+
+    Args:
+        text (str): The string to check
+
+    Returns:
+        bool: True if the string ends with the number pattern, False otherwise
+
+    """
+    return bool(PARENTHESES_PATTERN.search(text))
+
+
+def match_identical_names_in_preferred_synonyms(
+    s: Flow, t: Flow, comment: str = "Identical preferred synonyms"
+):
+    if t.synonyms and s.name in t.synonyms and s.context == t.context:
+        if s.name.normalized in t.name.normalized and (
+            has_roman_numeral_at_end(t.name.normalized)
+            or has_number_pattern_at_end(t.name.normalized)
+        ):
+            return {"comment": comment}
+    elif s.synonyms and t.name in s.synonyms and s.context == t.context:
+        if t.name.normalized in s.name.normalized and (
+            has_roman_numeral_at_end(s.name.normalized)
+            or has_number_pattern_at_end(s.name.normalized)
+        ):
+            return {"comment": comment}
+
+
+def match_identical_names_in_synonyms(
+    s: Flow, t: Flow, comment: str = "Identical synonyms"
+):
+    if (t.synonyms and s.name in t.synonyms and s.context == t.context) or (
+        s.synonyms and t.name in s.synonyms and s.context == t.context
+    ):
+        return {"comment": comment}