Fix dedupe() crash when an item is emptied by the processor

test_thefuzz.py

@@ -1,5 +1,6 @@
 import unittest
 import re
+import warnings
 import pycodestyle
 
 from thefuzz import fuzz
@@ -505,6 +506,27 @@ def test_dedupe(self):
         result = process.dedupe(contains_dupes)
         self.assertEqual(result, deduped_list)
 
+    def test_dedupe_with_empty_processed_item(self):
+        """dedupe() should not crash when an item is reduced to an empty
+        string by the default processor. Such an item has no fuzzy
+        duplicates (every comparison scores 0), so it is kept as its own
+        cluster instead of raising ValueError from max() of an empty
+        sequence. See issue #94.
+        """
+        # '###' is reduced to '' by utils.full_process, so extractBests
+        # returns no matches for it.
+        self.assertEqual(utils.full_process('###'), '')
+
+        contains_dupes = ['###', 'apple', 'apple pie']
+        with warnings.catch_warnings():
+            warnings.simplefilter('ignore')
+            result = process.dedupe(contains_dupes)
+
+        # The empty-processed item is preserved rather than dropped or crashing.
+        self.assertIn('###', result)
+        # The genuine duplicates ('apple', 'apple pie') still collapse to one.
+        self.assertEqual(len(result), 2)
+
     def test_simplematch(self):
         basic_string = 'a, b'
         match_strings = ['a, b']

thefuzz/process.py

@@ -438,6 +438,14 @@ def dedupe(
     deduped = set()
     for item in contains_dupes:
         matches = extractBests(item, contains_dupes, scorer=scorer, score_cutoff=threshold, limit=None)
+        if not matches:
+            # The item's processed form is empty (e.g. it contains only
+            # characters the processor strips), so every comparison scores 0
+            # and nothing meets the threshold - not even the item itself.
+            # It has no fuzzy duplicates, so keep it as its own cluster
+            # instead of crashing on max() of an empty sequence.
+            deduped.add(item)
+            continue
         deduped.add(max(matches, key=lambda x: (len(x[0]), x[0]))[0])
 
     return list(deduped) if len(deduped) != len(contains_dupes) else contains_dupes