Update plan_download_op.py

Signed-off-by: Robert Nishihara <robertnishihara@gmail.com>

Author: robertnishihara

python/ray/data/_internal/planner/plan_download_op.py

@@ -190,13 +190,16 @@ def load_uri_bytes(uri_path_iterator):
             """Function that takes an iterator of URI paths and yields downloaded bytes for each."""
             for uri_path in uri_path_iterator:
                 last_error = None
-                # Handle both file and stream for uri download
+                # Handle both file and stream uris. The file-based approach handles the vast majority of cases.
+                # However, some data sources are not seekable and require the stream approach.
                 for uri_open_func in [fs.open_input_file, fs.open_input_stream]:
                     try:
                         with uri_open_func(uri_path) as f:
                             yield f.read()
                         break
                     except ValueError as e:
+                        # This handles the case where the file is not seekable and so we need to fall back to
+                        # using fs.open_input_stream.
                         last_error = e
                         continue
                     except OSError as e: