housenumber analyzer: avoid creating too many variants

Housenumber fields with lots of text are likely bad data. So is
data with many changes from letter to digit. Exclude them from adding
optional spaces.

diff --git a/nominatim/tokenizer/token_analysis/housenumbers.py b/nominatim/tokenizer/token_analysis/housenumbers.py
index 6a838e00d395e4c05f478328aa9b7558d3da84eb..96e86b28e79982825898757756cf03cff0981475 100644 (file)
--- a/nominatim/tokenizer/token_analysis/housenumbers.py
+++ b/nominatim/tokenizer/token_analysis/housenumbers.py
@@ -15,17 +15,18 @@ from nominatim.tokenizer.token_analysis.generic_mutation import MutationVariantG
 RE_NON_DIGIT = re.compile('[^0-9]')
 RE_DIGIT_ALPHA = re.compile(r'(\d)\s*([^\d\s␣])')
 RE_ALPHA_DIGIT = re.compile(r'([^\s\d␣])\s*(\d)')
 RE_NON_DIGIT = re.compile('[^0-9]')
 RE_DIGIT_ALPHA = re.compile(r'(\d)\s*([^\d\s␣])')
 RE_ALPHA_DIGIT = re.compile(r'([^\s\d␣])\s*(\d)')

+RE_NAMED_PART = re.compile(r'[a-z]{4}')

 
 ### Configuration section
 
 
 ### Configuration section
 

-def configure(rules, normalization_rules):
+def configure(rules, normalization_rules): # pylint: disable=W0613

     """ All behaviour is currently hard-coded.
     """
     return None
 
 ### Analysis section
 
     """ All behaviour is currently hard-coded.
     """
     return None
 
 ### Analysis section
 

-def create(normalizer, transliterator, config):
+def create(normalizer, transliterator, config): # pylint: disable=W0613

     """ Create a new token analysis instance for this module.
     """
     return HousenumberTokenAnalysis(normalizer, transliterator)
     """ Create a new token analysis instance for this module.
     """
     return HousenumberTokenAnalysis(normalizer, transliterator)


@@ -48,8 +49,14 @@ class HousenumberTokenAnalysis:
             return name
 
         norm = self.trans.transliterate(self.norm.transliterate(name))
             return name
 
         norm = self.trans.transliterate(self.norm.transliterate(name))

-        norm = RE_DIGIT_ALPHA.sub(r'\1␣\2', norm)
-        norm = RE_ALPHA_DIGIT.sub(r'\1␣\2', norm)
+        # If there is a significant non-numeric part, use as is.
+        if RE_NAMED_PART.search(norm) is None:
+            # Otherwise add optional spaces between digits and letters.
+            (norm_opt, cnt1) = RE_DIGIT_ALPHA.subn(r'\1␣\2', norm)
+            (norm_opt, cnt2) = RE_ALPHA_DIGIT.subn(r'\1␣\2', norm_opt)
+            # Avoid creating too many variants per number.
+            if cnt1 + cnt2 <= 4:
+                return norm_opt

 
         return norm
 
 
         return norm