Skip to content

fix(egosuite): validate label report envelope #364

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
kstonekuan merged 3 commits into from
Sep 4, 2026
Merged

fix(egosuite): validate label report envelope #364

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
0b943e7
fix(egosuite): validate label report envelope
WilliamK112 Sep 2, 2026
752ce42
Merge main and update label report fixtures
WilliamK112 Sep 4, 2026
de57a4d
test(egosuite): pin the bool guard on the label report schema version
kstonekuan Sep 4, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 18 additions & 1 deletion examples/egosuite_evaluation/evaluate.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,6 +65,7 @@
)

SCHEMA_VERSION = 1
PROJECTED_HAND_LABEL_TYPE = "projected-hand-joints"
EXPECTED_HAND_JOINT_COUNT = 21
DEFAULT_RUNS_DIRECTORY = Path("data/egosuite-evaluation/runs")
DEFAULT_LABELS_DIRECTORY = Path("data/egosuite-evaluation/labels")
Expand Down Expand Up @@ -158,6 +159,22 @@ def load_projected_hand_label_report(
) from error
if not isinstance(report_payload, dict):
raise ValueError(f"projected-hand label report {report_path} must contain a JSON object")
schema_version = report_payload.get("schema_version")
if (
not isinstance(schema_version, int)
or isinstance(schema_version, bool)
or schema_version != SCHEMA_VERSION

@chatgpt-codex-connector chatgpt-codex-connector Bot Sep 2, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Preserve unknown label-report envelope versions

When a label report comes from a newer producer with an unknown schema version, this exact-version check raises during pipeline.py module initialization, preventing the pipeline from starting. The repository's forward-compatibility guidance requires unknown values to retain their raw representation in an explicit Unknown* variant, emit a warning, and be ignored gracefully rather than raising, so handle future versions through that path while still rejecting malformed known values.

AGENTS.md reference: AGENTS.md:L1-L1

Useful? React with 👍 / 👎.

):
raise ValueError(
f"projected-hand label report {report_path} field 'schema_version' has value "
f"{schema_version!r}; supported value is {SCHEMA_VERSION}"
)
label_type = report_payload.get("label_type")
if not isinstance(label_type, str) or label_type != PROJECTED_HAND_LABEL_TYPE:
raise ValueError(
f"projected-hand label report {report_path} field 'label_type' has value "
f"{label_type!r}; supported value is {PROJECTED_HAND_LABEL_TYPE!r}"
)
raw_frame_records = report_payload.get("frames")
if not isinstance(raw_frame_records, list):
raise ValueError(f"projected-hand label report {report_path} must contain a 'frames' array")
Expand Down Expand Up @@ -938,7 +955,7 @@ def write_label_report(
source_uri_by_path = _source_uri_by_path(tuple(labels_by_source))
report = {
"schema_version": SCHEMA_VERSION,
"label_type": "projected-hand-joints",
"label_type": PROJECTED_HAND_LABEL_TYPE,
"camera_view": camera_view.value,
"frame_stride": frame_stride,
"limit_per_episode": limit_per_episode,
Expand Down
141 changes: 119 additions & 22 deletions examples/egosuite_evaluation/tests/test_evaluation.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -269,31 +269,39 @@ def test_pipeline_registers_projected_hand_visibility_as_an_hflow_check() -> Non
def test_saved_label_report_selects_exact_frames_for_a_canonical_episode(tmp_path: Path) -> None:
source_path = tmp_path / "episode-123.mcap"
report_path = tmp_path / "labels.json"
report_path.write_text(
json.dumps(
{
"frames": [
{
"source_path": str(source_path),
"source_episode": "episode-123",
"camera_view": "head-left",
"frame_index": frame_index,
"left_in_frame_joint_count": 21,
"right_in_frame_joint_count": 21 if frame_index == 8 else 0,
"expected_hand_count": 2 if frame_index == 8 else 1,
"left_hand_issue_reasons": [],
"right_hand_issue_reasons": ["occlusion"] if frame_index == 8 else [],
}
for frame_index in (8, 3)
]
}
labels = [
ProjectedHandFrameLabel(
source_path=source_path,
source_episode="episode-123",
camera_view=CameraView.HEAD_LEFT,
frame_index=frame_index,
left_in_frame_joint_count=21,
right_in_frame_joint_count=21 if frame_index == 8 else 0,
expected_hand_count=2 if frame_index == 8 else 1,
left_hand_issue_reasons=(),
right_hand_issue_reasons=("occlusion",) if frame_index == 8 else (),
)
for frame_index in (8, 3)
]
write_label_report(
{source_path: labels},
camera_view=CameraView.HEAD_LEFT,
frame_stride=30,
limit_per_episode=None,
episode_count=None,
samples_per_episode=None,
samples_per_hand_count=None,
sample_seed=42,
output_path=report_path,
)

report_payload = json.loads(report_path.read_text())
recorded_source_uri = report_payload["frames"][0]["source_uri"]
assert isinstance(recorded_source_uri, str)
label_report = load_projected_hand_label_report(report_path)
selected_labels = labels_for_pipeline_episode(
tmp_path / "episode-123.canonical.mcap",
{"source_uri": "run-a/episode-123.mcap"},
{"source_uri": recorded_source_uri},
label_report,
)

Expand All @@ -302,11 +310,96 @@ def test_saved_label_report_selects_exact_frames_for_a_canonical_episode(tmp_pat
assert selected_labels[1].right_hand_issue_reasons == ("occlusion",)


@pytest.mark.parametrize(
("envelope", "field_name", "found_value", "supported_value"),
[
pytest.param(
{"label_type": "projected-hand-joints"},
"schema_version",
"None",
"1",
id="missing-schema-version",
),
pytest.param(
{"schema_version": "1", "label_type": "projected-hand-joints"},
"schema_version",
"'1'",
"1",
id="non-integer-schema-version",
),
pytest.param(
{"schema_version": 0, "label_type": "projected-hand-joints"},
"schema_version",
"0",
"1",
id="older-schema-version",
),
# bool subclasses int, so True passes an isinstance(_, int) check and
# compares equal to 1. Only the explicit bool guard rejects it.
pytest.param(
{"schema_version": True, "label_type": "projected-hand-joints"},
"schema_version",
"True",
"1",
id="bool-schema-version",
),
pytest.param(
{"schema_version": 2, "label_type": "projected-hand-joints"},
"schema_version",
"2",
"1",
id="future-schema-version",
),
pytest.param(
{"schema_version": 1},
"label_type",
"None",
"'projected-hand-joints'",
id="missing-label-type",
),
pytest.param(
{"schema_version": 1, "label_type": 1},
"label_type",
"1",
"'projected-hand-joints'",
id="non-string-label-type",
),
pytest.param(
{"schema_version": 1, "label_type": "bounding-boxes"},
"label_type",
"'bounding-boxes'",
"'projected-hand-joints'",
id="unsupported-label-type",
),
],
)
def test_saved_label_report_rejects_unsupported_envelope_before_frames(
tmp_path: Path,
envelope: dict[str, object],
field_name: str,
found_value: str,
supported_value: str,
) -> None:
report_path = tmp_path / "labels.json"
report_path.write_text(json.dumps({**envelope, "frames": [None]}))

with pytest.raises(ValueError) as error:
load_projected_hand_label_report(report_path)

message = str(error.value)
assert str(report_path) in message
assert repr(field_name) in message
assert f"value {found_value}" in message
assert f"supported value is {supported_value}" in message


def test_legacy_label_report_rejects_one_basename_for_multiple_sources(tmp_path: Path) -> None:
report_path = tmp_path / "labels.json"
report_path.write_text(
json.dumps(
{
"schema_version": 1,
"label_type": "projected-hand-joints",
"frames": [
{
"source_path": str(tmp_path / run_name / "episode.mcap"),
Expand All @@ -320,7 +413,7 @@ def test_legacy_label_report_rejects_one_basename_for_multiple_sources(tmp_path:
"right_hand_issue_reasons": [],
}
for run_name, frame_index in (("run-a", 3), ("run-b", 8))
]
],
}
)
)
Expand All @@ -338,6 +431,8 @@ def test_saved_label_report_matches_same_named_sources_by_canonical_provenance(
report_path.write_text(
json.dumps(
{
"schema_version": 1,
"label_type": "projected-hand-joints",
"frames": [
{
"source_path": str(source_path),
Expand All @@ -355,7 +450,7 @@ def test_saved_label_report_matches_same_named_sources_by_canonical_provenance(
(first_source_path, "run-a/episode.mcap", 1),
(second_source_path, "run-b/episode.mcap", 2),
)
]
],
}
)
)
Expand Down Expand Up @@ -387,6 +482,8 @@ def test_saved_label_report_rejects_missing_or_unrelated_canonical_provenance(
report_path.write_text(
json.dumps(
{
"schema_version": 1,
"label_type": "projected-hand-joints",
"frames": [
{
"source_path": str(source_path),
Expand All @@ -400,7 +497,7 @@ def test_saved_label_report_rejects_missing_or_unrelated_canonical_provenance(
"left_hand_issue_reasons": [],
"right_hand_issue_reasons": [],
}
]
],
}
)
)
Expand Down