[S3] Fix S3 list_recursive filtering with paths containing consecutive slashes

metaflow/plugins/datatools/s3/s3.py

@@ -53,6 +53,10 @@
 if TYPE_CHECKING:
     import metaflow
 
+# Compiled regex for normalizing consecutive slashes in S3 paths
+# Matches two or more consecutive slashes not preceded by a colon (to preserve s3://)
+_CONSECUTIVE_SLASHES_REGEX = re.compile(r"(?<!:)//+")
+
 
 def _check_and_init_s3_deps():
     try:
@@ -557,7 +561,9 @@ def __init__(
                 raise MetaflowS3URLException(
                     "s3root needs to be an S3 URL prefixed with s3://."
                 )
-            self._s3root = s3root.rstrip("/")
+            # Normalize the path by collapsing consecutive slashes
+            normalized = _CONSECUTIVE_SLASHES_REGEX.sub("/", s3root)
+            self._s3root = normalized.rstrip("/")
         else:
             # 3. use the client only with full URLs
             self._s3root = None

test/data/s3/test_s3.py

@@ -982,10 +982,13 @@ def test_list_recursive_sibling_prefix_filtering(inject_failure_rate):
     ) as s3:
         objects = s3.list_recursive()
 
+        # Use the normalized s3root from the S3 object for accurate path extraction
+        test_prefix_path = f"{s3_setup._s3root}/{test_prefix}/"
+
         found_relative_paths = []
         for obj in objects:
             # Get path relative to our test prefix
-            relative_path = obj.url.replace(f"{base_s3root}/{test_prefix}/", "")
+            relative_path = obj.url.replace(test_prefix_path, "")
             found_relative_paths.append(relative_path)
 
         expected_under_log = ["log/test.txt"]