Skip to content

test(dataset): assert the provenance sidecar payload whole #402

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 1 commit into from
Sep 4, 2026
Merged

test(dataset): assert the provenance sidecar payload whole #402

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
43 changes: 38 additions & 5 deletions tests/test_dataset.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,8 +8,14 @@

import hflow
from hflow.cli import main as cli_main
from hflow.dataset import create_dataset, dataset_slug, default_dataset_sql
from hflow.dataset import (
DATASET_MANIFEST_VERSION,
create_dataset,
dataset_slug,
default_dataset_sql,
)
from hflow.testing import SyntheticEpisodeSpec, synthesize_episode
from hflow.workspace import Workspace

PIPELINE_SOURCE = """
import hflow
Expand Down Expand Up @@ -145,7 +151,16 @@ def test_the_manifest_and_its_provenance_land_together(
_ingest(ingested_project, monkeypatch)
app = hflow.import_pipeline_application(str(ingested_project / "pipeline.py"))

dataset = create_dataset(app, "Clean Corpus!")
# Minted here because ingest does not mint one, and the assertion is
# only worth making against a real id: with none stored the field is
# None, which is also what dropping it from the payload would look like.
workspace_id = Workspace(app.storage_root).ensure_identity().workspace_id

# Passed rather than defaulted, so the assertion below pins that the
# sidecar records the caller's stamp and not a second now() taken inside
# _sidecar_payload.
stamped_at = "2026-01-02T03:04:05+00:00"
dataset = create_dataset(app, "Clean Corpus!", created_at=stamped_at)

manifest_file = Path(dataset.manifest_path)
sidecar_file = Path(dataset.sidecar_path)
Expand All @@ -160,13 +175,31 @@ def test_the_manifest_and_its_provenance_land_together(
assert len(episode_ids) == 1

provenance = json.loads(sidecar_file.read_text())
assert provenance["name"] == "Clean Corpus!"
assert provenance["sql"] == dataset.sql
assert provenance["row_count"] == 1
assert provenance["pipeline"]["pipeline_name"] == "dataset-demo"
assert provenance["pipeline"]["pipeline_version"] == app.pipeline_version
assert [check["name"] for check in provenance["pipeline"]["checks"]] == ["duration"]

# The rest of the payload as one dict, so a key going missing fails here
# as loudly as a key holding the wrong value. `pipeline` is compared
# above instead: it is the whole manifest, and pinning it whole would
# mean restating every check on every pipeline change.
assert provenance.pop("pipeline")
assert provenance == {
# Imported rather than written as a literal, so bumping the schema
# moves this test and a bump nobody meant fails it.
"dataset_manifest_version": DATASET_MANIFEST_VERSION,
"name": "Clean Corpus!",
"created_at": stamped_at,
"sql": dataset.sql,
"row_count": 1,
"total_episodes": 1,
# Real numbers, not merely a non-empty list: the auditability claim
# is that this says how many episodes each check actually ran on,
# and an empty list reads the same as a full one without them.
"coverage": [{"check_name": "duration", "episodes_ran": 1, "total_episodes": 1}],
"workspace_id": workspace_id,
}

def test_two_datasets_of_one_name_never_overwrite(
self, ingested_project: Path, monkeypatch: pytest.MonkeyPatch
) -> None:
Expand Down