def test_ParquetExporter_date_with_string(event, dates_file_date_string_value):
(prefix, file) = dates_file_date_string_value()
event_data = event(prefix, chunksize=None, schema=schema_dates)
exporter = ParquetExporter(event_data)
result = exporter.export()
assert result["status"] == "CONVERSION_FAILED"
assert len(result["errors"]) == 1
def test_ParquetExporter_invalid_year_too_late(event,
dates_file_year_too_late):
(prefix, file) = dates_file_year_too_late()
event_data = event(prefix, chunksize=None, schema=schema_dates)
exporter = ParquetExporter(event_data)
result = exporter.export()
assert result["status"] == "CONVERSION_FAILED"
assert len(result["errors"]) == 1
def test_ParquetExporter_with_schema_wrong_number(event, schema_wrong):
(prefix, file) = schema_wrong()
event_data = event(prefix, chunksize=None, schema=SCHEMA)
exporter = ParquetExporter(event_data)
with pytest.raises(
ConversionError,
match=
".*cannot safely convert passed user dtype of bool for int64.*",
):
exporter.export()
def test_ParquetExporter_no_chunks(event, husholdninger_single):
(prefix, file) = husholdninger_single()
event_data = event(prefix, chunksize=None)
exporter = ParquetExporter(event_data)
exporter.export()
expected = pd.read_csv(file)
output_prefix = event_data["payload"]["output_dataset"][
"s3_prefix"].replace("%stage%", "intermediate")
result = pd.read_parquet(f"s3://{BUCKET}/" + output_prefix +
event_data["task"] + "/husholdninger.parquet.gz")
# result is a subset of expected
assert len(result.merge(expected)) == len(result)
def export_and_read_result(event_data, outputprefix):
exporter = ParquetExporter(event_data)
exporter.export()
output_prefix = event_data["payload"]["output_dataset"][
"s3_prefix"].replace("%stage%", "intermediate")
fs = s3fs.core.S3FileSystem()
result_path = f"{BUCKET}/{output_prefix}{event_data['task']}/{outputprefix}*"
source_paths = fs.glob(path=result_path)
result = pd.concat(
pd.read_parquet(f"s3://{parquet_file}")
for parquet_file in source_paths)
return result
def test_ParquetExporter_chunked(event, husholdninger_single):
prefix, file = husholdninger_single()
event_data = event(prefix, chunksize=2)
exporter = ParquetExporter(event_data)
output_prefix = event_data["payload"]["output_dataset"][
"s3_prefix"].replace("%stage%", "intermediate")
task = event_data["task"]
with patch.object(exporter, "_parallel_export") as mocked_parallel_export:
exporter.export()
mocked_parallel_export.assert_called_once_with(
"husholdninger",
ANY,
None,
f"s3://{BUCKET}/{output_prefix}{task}/husholdninger",
)
def test_ParquetExporter_with_schema(event, schema):
(prefix, file) = schema()
event_data = event(prefix, chunksize=None, schema=SCHEMA)
exporter = ParquetExporter(event_data)
exporter.export()
output_prefix = event_data["payload"]["output_dataset"][
"s3_prefix"].replace("%stage%", "intermediate")
result = pd.read_parquet(f"s3://{BUCKET}/" + output_prefix +
event_data["task"] + "/schema.parquet.gz")
assert list(result.dtypes)[1].name == "bool"
assert list(result.dtypes)[2].name == "bool"
assert list(result.dtypes)[3].name == "float64"
assert list(result.dtypes)[4].name == "datetime64[ns]"
assert list(result["date"])[0] == pd.Timestamp("2020-03-14")
assert list(result["date"])[1] == pd.Timestamp("2020-01-01")
assert pd.isnull(list(result["date"])[2])