Fix: dictionary changed size during iteration in GCSObjectMetadataClient

[ikyasam18]

The issue was in the _adjust_gcs_metadata_limit_size method where labels
dictionary was being modified while iterating through it. Fixed by:

1. Creating a list of keys to remove first
2. Removing the keys after completing the iteration

This prevents the "dictionary changed size during iteration" error that
occurred when large metadata needed to be adjusted to fit GCS size limits.

[Copilot]

Pull Request Overview

This PR fixes a crash in _adjust_gcs_metadata_limit_size by avoiding in-place dict modification during iteration and adds a test to verify metadata trimming.

• Accumulates keys to remove before deleting them to prevent “dictionary changed size during iteration” errors
• Adds a unit test to confirm total metadata size remains within GCS limits

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
gokart/gcs_obj_metadata_client.py	Change loop to collect and then remove keys
test/test_gcs_obj_metadata_client.py	New test ensuring large metadata is truncated

Comments suppressed due to low confidence (3)

test/test_gcs_obj_metadata_client.py:143

• [nitpick] The test name suggests an expected runtime error but actually checks size truncation. Rename it to something like test_adjust_gcs_metadata_limit_size_truncates_labels for clarity.

def test_adjust_gcs_metadata_limit_size_runtime_error(self):

test/test_gcs_obj_metadata_client.py:151

• Consider adding assertions to verify that specific keys beyond the size limit were removed, not just the total size, for a more robust test.

self.assertLessEqual(total_size, 8 * 1024)

gokart/gcs_obj_metadata_client.py:159

• This method mutates the labels dict in-place and returns it. To avoid unexpected side effects, consider returning a new dict or documenting the in-place behavior in the docstring.

def _get_label_size(label_name: str, label_value: str) -> int:

[hiro-o918]

Can you define a constant value in gokart/gcs_obj_metadata_client.py

MAX_GCS_METADATA_SIZE: Final[int] = 8 * 1024

and refer here and L157 of gokart/gcs_obj_metadata_client.py

[ikyasam18]

Thank you for your review. I've fixed it!

[kitagry]

LGTM! Thanks

[kitagry]

Is encode('utf-8') needs?

[ikyasam18]

When working with non-ASCII characters (such as Japanese etc.), it is necessary because the number of characters and the number of bytes do not match.

[kitagry]

I think so. But, in this test case, is it need?

[ikyasam18]

I apologize. I misunderstood it as production code. Since this test only needs to verify that no error occurs, it doesn't seem necessary.

[hiro-o918]

LGTM!

[kitagry]

Thank you!