def clear_cached_render_result_for_step(step: Step) -> None:
"""Delete a CachedRenderResult, if it exists.
This deletes the Parquet file from disk, _then_ empties relevant
database fields and saves them (and only them).
"""
delete_parquet_files_for_step(step.workflow_id, step.id)
step.cached_render_result_delta_id = None
step.cached_render_result_errors = []
step.cached_render_result_json = b"null"
step.cached_render_result_status = None
step.cached_render_result_columns = None
step.cached_render_result_nrows = None
step.save(update_fields=STEP_FIELDS)
def cache_render_result(
workflow: Workflow, step: Step, delta_id: int, result: LoadedRenderResult
) -> None:
"""Save `result` for later viewing.
Raise AssertionError if `delta_id` is not what we expect.
Since this alters data, call it within a lock:
with workflow.cooperative_lock():
step.refresh_from_db() # may change delta_id
cache_render_result(workflow, step, delta_id, result)
"""
assert delta_id == step.last_relevant_delta_id
assert result is not None
json_bytes = json_encode(result.json).encode("utf-8")
if not result.columns:
if result.errors:
status = "error"
else:
status = "unreachable"
else:
status = "ok"
step.cached_render_result_delta_id = delta_id
step.cached_render_result_errors = result.errors
step.cached_render_result_status = status
step.cached_render_result_json = json_bytes
step.cached_render_result_columns = result.columns
step.cached_render_result_nrows = result.table.num_rows
# Now we get to the part where things can end up inconsistent. Try to
# err on the side of not-caching when that happens.
delete_parquet_files_for_step(workflow.id, step.id) # makes old cache inconsistent
step.save(update_fields=STEP_FIELDS) # makes new cache inconsistent
if result.table.num_columns: # only write non-zero-column tables
with tempfile_context() as parquet_path:
cjwparquet.write(parquet_path, result.table)
s3.fput_file(
BUCKET, parquet_key(workflow.id, step.id, delta_id), parquet_path
) # makes new cache consistent