def test_get_execution_timestamps(self, window_size, schedule_stride,
schedule_start, start, end,
expected_timestamp):
schedule_data_spec = ScheduleDataSpec(
input=ScheduleInputSpec(),
output=ScheduleOutputSpec(),
stride=schedule_stride,
window_size=window_size,
start=schedule_start,
)
assert expected_timestamp == schedule_data_spec.get_execution_timestamps(
start, end)
def test_empty_specs_default_fields_to_empty_dict():
ds = ScheduleDataSpec(input=ScheduleInputSpec(),
output=ScheduleOutputSpec(),
stride=1,
window_size=1,
start=1)
assert {} == ds.input.time_series
assert {} == ds.output.time_series
def test_schedule_data_spec_schema_slack_missing_default_to_zero():
schedule_data_spec = ScheduleDataSpec.load({
"input": {},
"output": {},
"stride": 1,
"windowSize": 1,
"start": 1
})
assert 0 == schedule_data_spec.slack
def test_now_without_aggregate(self, now_cache):
now_cache.get_time_now.return_value = 3 * 60 * 60 * 1000 + 3000
schedule_data_spec = ScheduleDataSpec(
input=ScheduleInputSpec(
time_series={"ts1": ScheduleInputTimeSeriesSpec(id=3)}),
output=ScheduleOutputSpec(),
stride="1h",
window_size="10h",
start="now",
)
assert 3 * 60 * 60 * 1000 + 3000 == schedule_data_spec.start
def test_get_instances(ids, ext_ids):
schedule_ts_spec = {
"ts1": ScheduleInputTimeSeriesSpec(id=ids[0], external_id=ext_ids[0]),
"ts2": ScheduleInputTimeSeriesSpec(id=ids[1], external_id=ext_ids[1]),
}
schedule_data_spec = ScheduleDataSpec(
input=ScheduleInputSpec(time_series=schedule_ts_spec),
output=ScheduleOutputSpec(),
stride="1m",
window_size="1m",
start=60000,
)
data_specs = schedule_data_spec.get_instances(start=60000, end=6 * 60000)
expected_data_specs = []
for i in range(0, 5 * 60000, 60000):
time_series_spec1 = TimeSeriesSpec(id=ids[0],
external_id=ext_ids[0],
start=i,
end=i + 60000)
time_series_spec2 = TimeSeriesSpec(id=ids[1],
external_id=ext_ids[1],
start=i,
end=i + 60000)
data_spec = DataSpec(
time_series={
"ts1": time_series_spec1,
"ts2": time_series_spec2
},
metadata=DataSpecMetadata(
ScheduleSettings(start=i,
end=i + 60000,
window_size=60000,
stride=60000)),
)
expected_data_specs.append(data_spec)
assert len(expected_data_specs) == len(data_specs)
for expected, actual in zip(expected_data_specs, data_specs):
assert expected == actual
class TestSpecConstructor:
@pytest.fixture(autouse=True, scope="class")
def mock_time(self):
with mock.patch(
"cognite.model_hosting._cognite_model_hosting_common.utils.time.time"
) as m:
m.return_value = 10**6
yield
TestCase = namedtuple("TestCase", ["name", "constructor", "primitive"])
InvalidTestCase = namedtuple(
"InvalidTestCase", ["name", "constructor", "exception", "error_match"])
test_cases = [
TestCase(
name="time_series_time_ago",
constructor=lambda: TimeSeriesSpec(
id=123, start="2m-ago", end="now"),
primitive={
"id": 123,
"start": 10**9 - 2 * 60 * 1000,
"end": 10**9
},
),
TestCase(
name="time_series_datetime",
constructor=lambda: TimeSeriesSpec(
id=123, start=datetime(2018, 1, 1), end=datetime(2018, 1, 2)),
primitive={
"id": 123,
"start": 1514764800000,
"end": 1514851200000
},
),
TestCase(
name="schedule_data_spec_default_start_now",
constructor=lambda: ScheduleDataSpec(input=ScheduleInputSpec(),
output=ScheduleOutputSpec(),
stride=123,
window_size=234,
slack=345),
primitive={
"input": {},
"output": {},
"stride": 123,
"windowSize": 234,
"start": 10**9,
"slack": 345
},
),
TestCase(
name="schedule_data_spec_default_slack_zero",
constructor=lambda: ScheduleDataSpec(input=ScheduleInputSpec(),
output=ScheduleOutputSpec(),
stride=123,
window_size=234),
primitive={
"input": {},
"output": {},
"stride": 123,
"windowSize": 234,
"start": 10**9,
"slack": 0
},
),
TestCase(
name="schedule_data_spec_string_formats",
constructor=lambda: ScheduleDataSpec(
input=ScheduleInputSpec(),
output=ScheduleOutputSpec(),
stride="1m",
window_size="2m",
start="2m-ago",
slack="3m",
),
primitive={
"input": {},
"output": {},
"stride": 60000,
"windowSize": 120000,
"start": 10**9 - 2 * 60 * 1000,
"slack": 180000,
},
),
TestCase(
name="schedule_data_spec_datetime",
constructor=lambda: ScheduleDataSpec(
input=ScheduleInputSpec(),
output=ScheduleOutputSpec(),
stride=timedelta(minutes=1),
window_size=timedelta(minutes=2),
start=datetime(2018, 1, 1),
slack=timedelta(minutes=3),
),
primitive={
"input": {},
"output": {},
"stride": 60000,
"windowSize": 120000,
"start": 1514764800000,
"slack": 180000,
},
),
]
invalid_test_cases = [
InvalidTestCase(
name="time_series_start_none",
constructor=lambda: TimeSeriesSpec(id=123, start=None, end=2),
exception=TypeError,
error_match="type",
),
InvalidTestCase(
name="time_series_end_none",
constructor=lambda: TimeSeriesSpec(id=123, start=2, end=None),
exception=TypeError,
error_match="type",
),
]
@pytest.mark.parametrize("name, constructor, primitive", test_cases)
def test_valid(self, name, constructor, primitive):
spec = constructor()
assert spec.dump() == primitive
@pytest.mark.parametrize("name, constructor, exception, error_match",
invalid_test_cases)
def test_invalid(self, name, constructor, exception, error_match):
with pytest.raises(exception, match=error_match):
constructor()
class TestSpecValidation:
TestCase = namedtuple("TestCase", ["name", "obj", "primitive"])
InvalidTestCase = namedtuple(
"TestCase", ["name", "type", "constructor", "primitive", "errors"])
valid_test_cases = [
TestCase("minimal_file_spec", FileSpec(id=6), {"id": 6}),
TestCase("minimal_file_spec_external_id", FileSpec(external_id="abc"),
{"externalId": "abc"}),
TestCase("minimal_time_series_spec",
TimeSeriesSpec(id=6, start=123, end=234), {
"id": 6,
"start": 123,
"end": 234
}),
TestCase(
"time_series_include_outside_points",
TimeSeriesSpec(id=6,
start=123,
end=234,
include_outside_points=True),
{
"id": 6,
"start": 123,
"end": 234,
"includeOutsidePoints": True
},
),
TestCase(
"time_series_aggregate",
TimeSeriesSpec(id=6,
start=123,
end=234,
aggregate="average",
granularity="1m"),
{
"id": 6,
"start": 123,
"end": 234,
"aggregate": "average",
"granularity": "1m"
},
),
TestCase(
"time_series_external_id",
TimeSeriesSpec(external_id="abc",
start=123,
end=234,
aggregate="average",
granularity="1m"),
{
"externalId": "abc",
"start": 123,
"end": 234,
"aggregate": "average",
"granularity": "1m"
},
),
TestCase(
"schedule_input_time_series",
ScheduleInputTimeSeriesSpec(id=6,
aggregate="average",
granularity="1m"),
{
"id": 6,
"aggregate": "average",
"granularity": "1m"
},
),
TestCase(
"schedule_input_time_series_external_id",
ScheduleInputTimeSeriesSpec(external_id="abc",
aggregate="average",
granularity="1m"),
{
"externalId": "abc",
"aggregate": "average",
"granularity": "1m"
},
),
TestCase("empty_data_spec", DataSpec(), {}),
TestCase(
"full_data_spec",
DataSpec(
time_series={
"ts1": TimeSeriesSpec(id=6, start=123, end=234),
"ts2": TimeSeriesSpec(id=7, start=1234, end=2345),
},
files={
"f1": FileSpec(id=3),
"f2": FileSpec(id=4)
},
),
{
"timeSeries": {
"ts1": {
"id": 6,
"start": 123,
"end": 234
},
"ts2": {
"id": 7,
"start": 1234,
"end": 2345
},
},
"files": {
"f1": {
"id": 3
},
"f2": {
"id": 4
}
},
},
),
TestCase(
"data_spec_with_metadata",
DataSpec(metadata=DataSpecMetadata(
ScheduleSettings(stride=1, window_size=1, start=1, end=2))),
{
"metadata": {
"scheduleSettings": {
"stride": 1,
"windowSize": 1,
"start": 1,
"end": 2
}
}
},
),
TestCase(
"full_data_spec_with_metadata",
DataSpec(
time_series={
"ts1": TimeSeriesSpec(id=6, start=123, end=234),
"ts2": TimeSeriesSpec(id=7, start=1234, end=2345),
},
files={
"f1": FileSpec(id=3),
"f2": FileSpec(id=4)
},
metadata=DataSpecMetadata(
ScheduleSettings(stride=1, window_size=1, start=1, end=2)),
),
{
"timeSeries": {
"ts1": {
"id": 6,
"start": 123,
"end": 234
},
"ts2": {
"id": 7,
"start": 1234,
"end": 2345
},
},
"files": {
"f1": {
"id": 3
},
"f2": {
"id": 4
}
},
"metadata": {
"scheduleSettings": {
"stride": 1,
"windowSize": 1,
"start": 1,
"end": 2
}
},
},
),
TestCase("minimal_schedule_input_spec", ScheduleInputSpec(), {}),
TestCase(
"full_schedule_input_spec",
ScheduleInputSpec(
time_series={
"ts1": ScheduleInputTimeSeriesSpec(id=6),
"ts2": ScheduleInputTimeSeriesSpec(id=7)
}),
{"timeSeries": {
"ts1": {
"id": 6
},
"ts2": {
"id": 7
}
}},
),
TestCase(
"schedule_output_time_series_spec",
ScheduleOutputTimeSeriesSpec(id=123, offset=-5),
{
"id": 123,
"offset": -5
},
),
TestCase(
"schedule_output_time_series_spec_external_id",
ScheduleOutputTimeSeriesSpec(external_id="abc", offset=-5),
{
"externalId": "abc",
"offset": -5
},
),
TestCase("minimal_schedule_output_spec", ScheduleOutputSpec(), {}),
TestCase(
"full_schedule_output_spec",
ScheduleOutputSpec(
time_series={
"ts1": ScheduleOutputTimeSeriesSpec(id=123, offset=5),
"ts2": ScheduleOutputTimeSeriesSpec(id=234, offset=0),
}),
{
"timeSeries": {
"ts1": {
"id": 123,
"offset": 5
},
"ts2": {
"id": 234,
"offset": 0
}
}
},
),
TestCase(
"minimal_schedule_data_spec",
ScheduleDataSpec(input=ScheduleInputSpec(),
output=ScheduleOutputSpec(),
stride=1,
window_size=2,
start=3),
{
"input": {},
"output": {},
"stride": 1,
"windowSize": 2,
"start": 3,
"slack": 0
},
),
TestCase(
"full_schedule_data_spec",
ScheduleDataSpec(
input=ScheduleInputSpec(
time_series={"ts1": ScheduleInputTimeSeriesSpec(id=5)}),
output=ScheduleOutputSpec(
time_series={
"ts1": ScheduleOutputTimeSeriesSpec(id=123, offset=100)
}),
stride=1,
window_size=2,
start=3,
slack=4,
),
{
"input": {
"timeSeries": {
"ts1": {
"id": 5
}
}
},
"output": {
"timeSeries": {
"ts1": {
"id": 123,
"offset": 100
}
}
},
"stride": 1,
"windowSize": 2,
"start": 3,
"slack": 4,
},
),
TestCase(
"schedule_data_spec_with_aggregates",
ScheduleDataSpec(
input=ScheduleInputSpec(
time_series={
"ts1":
ScheduleInputTimeSeriesSpec(
id=5, aggregate="average", granularity="3h"),
"ts2":
ScheduleInputTimeSeriesSpec(
id=6, aggregate="max", granularity="10m"),
}),
output=ScheduleOutputSpec(),
stride=60 * 60 * 1000,
window_size=3 * 60 * 60 * 1000,
start=123 * 60 * 60 * 1000,
),
{
"input": {
"timeSeries": {
"ts1": {
"id": 5,
"aggregate": "average",
"granularity": "3h"
},
"ts2": {
"id": 6,
"aggregate": "max",
"granularity": "10m"
},
}
},
"output": {},
"stride": 60 * 60 * 1000,
"windowSize": 3 * 60 * 60 * 1000,
"start": 123 * 60 * 60 * 1000,
"slack": 0,
},
),
]
invalid_test_cases = [
InvalidTestCase(
name="file_missing_identifier",
type=FileSpec,
constructor=lambda: FileSpec(),
primitive={},
errors={
"id": ["Exactly one of id and external_id must be specified."],
"external_id":
["Exactly one of id and external_id must be specified."],
},
),
InvalidTestCase(
name="file_spec_id_and_external_id_specified",
type=FileSpec,
constructor=lambda: FileSpec(id=1, external_id="abc"),
primitive={
"id": 1,
"externalId": "abc"
},
errors={
"id": ["Exactly one of id and external_id must be specified."],
"external_id":
["Exactly one of id and external_id must be specified."],
},
),
InvalidTestCase(
name="time_series_missing_id_and_external_id",
type=TimeSeriesSpec,
constructor=lambda: TimeSeriesSpec(id=None, start=1, end=2),
primitive={
"start": 1,
"end": 2
},
errors={
"id": ["Exactly one of id and external_id must be specified."],
"external_id":
["Exactly one of id and external_id must be specified."],
},
),
InvalidTestCase(
name="time_series_id_and_external_id_specified",
type=TimeSeriesSpec,
constructor=lambda: TimeSeriesSpec(
id=1, external_id="abc", start=1, end=2),
primitive={
"start": 1,
"end": 2
},
errors={
"external_id":
["Exactly one of id and external_id must be specified."],
"id": ["Exactly one of id and external_id must be specified."],
},
),
InvalidTestCase(
name="time_series_missing_fields",
type=TimeSeriesSpec,
constructor=None,
primitive={},
errors={
"id": ["Exactly one of id and external_id must be specified."],
"external_id":
["Exactly one of id and external_id must be specified."],
"start": ["Missing data for required field."],
"end": ["Missing data for required field."],
},
),
InvalidTestCase(
name="time_series_aggregates_but_not_granularity",
type=TimeSeriesSpec,
constructor=lambda: TimeSeriesSpec(
id=6, start=123, end=234, aggregate="average"),
primitive={
"id": 6,
"start": 123,
"end": 234,
"aggregate": "average"
},
errors={
"granularity":
["granularity must be specified for aggregates."]
},
),
InvalidTestCase(
name="time_series_aggregates_but_include_outside_points",
type=TimeSeriesSpec,
constructor=lambda: TimeSeriesSpec(id=6,
start=123,
end=234,
aggregate="average",
granularity="1m",
include_outside_points=True),
primitive={
"id": 6,
"start": 123,
"end": 234,
"aggregate": "average",
"granularity": "1m",
"includeOutsidePoints": True,
},
errors={
"includeOutsidePoints":
["Can't include outside points for aggregates."]
},
),
InvalidTestCase(
name="time_series_invalid_aggregate_function",
type=TimeSeriesSpec,
constructor=lambda: TimeSeriesSpec(
id=6, start=123, end=234, aggregate="avg", granularity="1m"),
primitive={
"id": 6,
"start": 123,
"end": 234,
"aggregate": "avg",
"granularity": "1m"
},
errors={
"aggregate":
["Not a valid aggregate function. Cannot use shorthand name."],
"granularity":
["granularity can only be specified for aggregates."],
},
),
InvalidTestCase(
name="time_series_not_aggregate_but_granularity",
type=TimeSeriesSpec,
constructor=lambda: TimeSeriesSpec(
id=6, start=123, end=234, granularity="1m"),
primitive={
"id": 6,
"start": 123,
"end": 234,
"granularity": "1m"
},
errors={
"granularity":
["granularity can only be specified for aggregates."]
},
),
InvalidTestCase(
name="time_series_invalid_granularity",
type=TimeSeriesSpec,
constructor=lambda: TimeSeriesSpec(
id=6, start=123, end=234, granularity="bla"),
primitive={
"id": 6,
"start": 123,
"end": 234,
"granularity": "bla"
},
errors={
"granularity": [
"Invalid granularity format: `bla`. Must be on format <integer>(s|m|h|d). E.g. '5m', '3h' or '1d'."
]
},
),
InvalidTestCase(
name="data_spec_nested_errors",
type=DataSpec,
constructor=lambda: DataSpec(files={"f1": FileSpec(id=None)}),
primitive={"files": {
"f1": {}
}},
errors={
"files": {
"f1": {
"value": {
"external_id": [
"Exactly one of id and external_id must be specified."
],
"id": [
"Exactly one of id and external_id must be specified."
],
}
}
}
},
),
InvalidTestCase(
name="data_spec_with_metadata_invalid_type",
type=DataSpec,
constructor=lambda: DataSpec(metadata=DataSpecMetadata(
ScheduleSettings(stride="1m", window_size=1, start=1, end=1))),
primitive={
"metadata": {
"scheduleSettings": {
"stride": "1m",
"windowSize": 1,
"start": 1,
"end": 1
}
}
},
errors={
"metadata": {
"scheduleSettings": {
"stride": ["Not a valid integer."]
}
}
},
),
InvalidTestCase(
name="data_spec_invalid_alias",
type=DataSpec,
constructor=lambda: DataSpec(
files={"1f": FileSpec(id=1)},
time_series={"F1": TimeSeriesSpec(id=1, start=0, end=10)}),
primitive={
"files": {
"1f": {
"id": 1
}
},
"timeSeries": {
"F1": {
"id": 1,
"start": 0,
"end": 10
}
}
},
errors={
"files": {
"1f": {
"key": [
"Invalid alias. Must be 1 to 50 lowercase alphanumeric characters or `_`. Must start with a letter and cannot end with `_` ."
]
}
},
"timeSeries": {
"F1": {
"key": [
"Invalid alias. Must be 1 to 50 lowercase alphanumeric characters or `_`. Must start with a letter and cannot end with `_` ."
]
}
},
},
),
# TODO add when Marshmallow fixes inconsistencies (https://github.com/marshmallow-code/marshmallow/issues/1132)
# InvalidTestCase(
# name="schedule_data_spec_nested_errors",
# type=ScheduleDataSpec,
# constructor=lambda: ScheduleDataSpec(
# stride="1m", window_size="5m", time_series={"ts1": ScheduleTimeSeriesSpec(id="abc")}
# ),
# primitive={"stride": "1m", "windowSize": "5m", "timeSeries": {"ts1": {"id": "abc"}}},
# errors={"timeSeries": {"ts1": {"value": {"id": ["Not a valid integer."]}}}},
# ),
InvalidTestCase(
name="schedule_output_time_series_spec_missing_fields",
type=ScheduleOutputTimeSeriesSpec,
constructor=lambda: ScheduleOutputTimeSeriesSpec(offset=None),
primitive={},
errors={
"id": ["Exactly one of id and external_id must be specified."],
"external_id":
["Exactly one of id and external_id must be specified."],
"offset": ["Missing data for required field."],
},
),
InvalidTestCase(
name=
"schedule_output_time_series_spec_id_and_external_id_specified",
type=ScheduleOutputTimeSeriesSpec,
constructor=lambda: ScheduleOutputTimeSeriesSpec(
id=1, external_id="abc", offset=0),
primitive={
"id": 1,
"externalId": "abc",
"offset": 0
},
errors={
"id": ["Exactly one of id and external_id must be specified."],
"external_id":
["Exactly one of id and external_id must be specified."],
},
),
InvalidTestCase(
name="schedule_input_time_series_spec_missing_fields",
type=ScheduleInputTimeSeriesSpec,
constructor=lambda: ScheduleInputTimeSeriesSpec(),
primitive={},
errors={
"id": ["Exactly one of id and external_id must be specified."],
"external_id":
["Exactly one of id and external_id must be specified."],
},
),
InvalidTestCase(
name="schedule_input_time_series_spec_id_and_external_id_specified",
type=ScheduleInputTimeSeriesSpec,
constructor=lambda: ScheduleInputTimeSeriesSpec(id=1,
external_id="abc"),
primitive={
"id": 1,
"externalId": "abc"
},
errors={
"id": ["Exactly one of id and external_id must be specified."],
"external_id":
["Exactly one of id and external_id must be specified."],
},
),
InvalidTestCase(
name="schedule_data_spec_missing_fields",
type=ScheduleDataSpec,
constructor=None,
primitive={"start": 123},
errors={
"input": ["Missing data for required field."],
"output": ["Missing data for required field."],
"windowSize": ["Missing data for required field."],
"stride": ["Missing data for required field."],
},
),
InvalidTestCase(
name="schedule_data_spec_invalid_stride_window_size",
type=ScheduleDataSpec,
constructor=None,
primitive={
"input": {},
"output": {},
"windowSize": 0,
"stride": -1,
"start": 123
},
errors={
"stride": ["Must be at least 1."],
"windowSize": ["Must be at least 1."]
},
),
InvalidTestCase(
name=
"schedule_data_spec_invalid_stride_window_size_start_multiple_of_largest_granularity_unit",
type=ScheduleDataSpec,
constructor=None,
primitive={
"input": {
"timeSeries": {
"ts1": {
"id": 5,
"aggregate": "average",
"granularity": "3h"
},
"ts2": {
"id": 6,
"aggregate": "max",
"granularity": "10m"
},
}
},
"output": {},
"windowSize": 123123,
"stride": 123123,
"start": 123123,
},
errors={
"stride": [
"Must be a multiple of the largest granularity unit in the input time series."
],
"windowSize": [
"Must be greater than or equal to the largest granularity of any of aggregated input time series."
],
"start": [
"Must be a multiple of the largest granularity unit in the input time series."
],
},
),
]
@pytest.mark.parametrize("name, obj, primitive", valid_test_cases)
def test_valid_dump(self, name, obj, primitive):
dumped = obj.dump()
assert dumped == primitive
@pytest.mark.parametrize("name, obj, primitive", valid_test_cases)
def test_valid_load(self, name, obj, primitive):
loaded = obj.__class__.load(primitive)
assert loaded == obj
@pytest.mark.parametrize("name, obj, primitive", valid_test_cases)
def test_valid_json_serializable(self, name, obj, primitive):
json_data = obj.to_json()
from_json = obj.__class__.from_json(json_data)
assert from_json == obj, "\nFrom JSON: ({})\n{}\nObj:({})\n{}\n".format(
type(from_json), from_json, type(obj), obj)
def test_load_with_unknown_fields(self):
primitive = {
"timeSeries": {
"ts1": {
"id": 6,
"start": 123,
"end": 234
},
"ts2": {
"id": 7,
"start": 1234,
"end": 2345
}
},
"files": {
"f1": {
"id": 3
},
"f2": {
"id": 4
}
},
"metadata": {
"scheduleSettings": {
"stride": 1,
"windowSize": 1,
"start": 1,
"end": 2
}
},
"unkown_field": "blabla",
}
ds = DataSpec.load(primitive)
assert {"ts1", "ts2"} == set(ds.time_series.keys())
assert {"f1", "f2"} == set(ds.files.keys())
assert {
"stride": 1,
"windowSize": 1,
"start": 1,
"end": 2
} == ds.metadata.schedule_settings.dump()
assert "unkown_field" not in ds.dump()
def remove_defaultdict_in_errors(self, d):
return json.loads(json.dumps(d))
@pytest.mark.parametrize("name, type, constructor, primitive, errors",
invalid_test_cases)
def test_invalid(self, name, type, constructor, primitive, errors):
should_fail = {
"load": lambda: type.load(primitive),
"from_json": lambda: type.from_json(json.dumps(primitive))
}
if constructor is not None:
if type in (DataSpec, ScheduleDataSpec):
should_fail["constructor"] = constructor
else:
should_fail["dump"] = constructor().dump
should_fail["to_json"] = constructor().to_json
for method_name, method in should_fail.items():
with pytest.raises(SpecValidationError) as excinfo:
method()
actual_errors = self.remove_defaultdict_in_errors(
excinfo.value.errors)
if actual_errors != errors:
pytest.fail(
"\nMethod: {}\nErrors:\n{}\nExpected:\n{}\n".format(
name, actual_errors, errors))
@pytest.mark.parametrize("name, obj, primitive", valid_test_cases)
def test_valid_str_repr(self, name, obj, primitive):
assert str(obj) == obj.to_json()