def test_from_pandas():
pd_df = pd.DataFrame({
"x": [1, 3, 5, 7, 9],
"y": [2, 4, 6, 8, 10],
"z": [None, "a", "b", "c", "abc"]
})
ge_df = ge.from_pandas(pd_df)
assert isinstance(ge_df, ge.data_asset.DataAsset)
assert list(ge_df.columns) == ["x", "y", "z"]
assert list(ge_df["x"]) == list(pd_df["x"])
assert list(ge_df["y"]) == list(pd_df["y"])
assert list(ge_df["z"]) == list(pd_df["z"])
# make an empty subclass to test dataset_class argument
class CustomPandasDataset(ge.dataset.PandasDataset):
pass
ge_df_custom = ge.from_pandas(pd_df, dataset_class=CustomPandasDataset)
assert not isinstance(ge_df, CustomPandasDataset)
assert isinstance(ge_df_custom, CustomPandasDataset)
assert list(ge_df_custom.columns) == ["x", "y", "z"]
assert list(ge_df_custom["x"]) == list(pd_df["x"])
assert list(ge_df_custom["y"]) == list(pd_df["y"])
assert list(ge_df_custom["z"]) == list(pd_df["z"])
def test_from_pandas():
pd_df = pd.DataFrame({
'x': [1, 3, 5, 7, 9],
'y': [2, 4, 6, 8, 10],
'z': [None, 'a', 'b', 'c', 'abc']
})
ge_df = ge.from_pandas(pd_df)
assert isinstance(ge_df, ge.dataset.Dataset)
assert list(ge_df.columns) == ['x', 'y', 'z']
assert list(ge_df['x']) == list(pd_df['x'])
assert list(ge_df['y']) == list(pd_df['y'])
assert list(ge_df['z']) == list(pd_df['z'])
# make an empty subclass to test dataset_class argument
class CustomPandasDataset(ge.dataset.PandasDataset):
pass
ge_df_custom = ge.from_pandas(pd_df, dataset_class=CustomPandasDataset)
assert not isinstance(ge_df, CustomPandasDataset)
assert isinstance(ge_df_custom, CustomPandasDataset)
assert list(ge_df_custom.columns)==['x', 'y', 'z']
assert list(ge_df_custom['x'])==list(pd_df['x'])
assert list(ge_df_custom['y'])==list(pd_df['y'])
assert list(ge_df_custom['z'])==list(pd_df['z'])
def _get_expectations(
self, field_mapping: FieldMapping, field_name: str
) -> ge.dataset.Dataset:
field_mapping_ge = ge.from_pandas(field_mapping.get_field_mapping_df())
field_mapping_ge.set_default_expectation_argument("result_format", "COMPLETE")
# Shape
shape_expectation = field_mapping_ge.expect_table_columns_to_match_ordered_list(
common.COLUMN_NAMES
)
# If the shape isn't correct then the following expectations will raise exceptions
if not shape_expectation["success"]:
return field_mapping_ge
field = self.table_schema.get_field(field_name)
# Check that there aren't multiple mappings for the same input value
field_mapping_ge.expect_column_values_to_be_unique(common.INPUT_COLUMN_NAME)
# Check that the mapped values are all part of the available options (blank value is also valid)
valid_mapped_values = self._get_valid_mappings_for_field(field) + [""]
field_mapping_ge.expect_column_values_to_be_in_set(
common.OUTPUT_COLUMN_NAME, valid_mapped_values
)
# Check that the approved column values are all "Yes", "No", or blank (Assumed to be no)
field_mapping_ge.expect_column_values_to_be_in_set(
common.APPROVED_COLUMN_NAME, common.VALID_APPROVED_VALUES + ["None"]
)
return field_mapping_ge
def _do_expectation(_info, df):
ge_df = ge.from_pandas(df)
ge_result = ge_callback(ge_df)
check.invariant('success' in ge_result)
return ExpectationResult(
success=ge_result['success'],
metadata_entries=[
EventMetadataEntry.json(label='result', data=ge_result)
],
)
def _get_expectations(self, field_mapping: FieldMapping) -> ge.dataset.Dataset:
field_mapping_ge = ge.from_pandas(field_mapping.get_field_mapping_df())
field_mapping_ge.set_default_expectation_argument("result_format", "COMPLETE")
# Check that the approved column values are all "Yes"
field_mapping_ge.expect_column_values_to_be_in_set(
common.APPROVED_COLUMN_NAME, [common.APPROVED]
)
return field_mapping_ge
def check_iris_data(df: DataFrame):
gdf = ge.from_pandas(df)
result = gdf.expect_column_values_to_be_in_set(
"species", ["setosa", "virginica", "versicolor"], mostly=0.99
)
if not result.success:
raise Exception("iris data is no good")
return df
def validate_with_great_expectations(
self,
dataframe: pd.DataFrame,
expectation_suite: TypeVar("ge.core.ExpectationSuite"),
ge_validate_kwargs: Optional[Dict[Any, Any]] = {},
):
report = ge.from_pandas(
dataframe, expectation_suite=expectation_suite
).validate(**ge_validate_kwargs)
return report
def test_from_pandas(self):
pd_df = pd.DataFrame({
'x': [1, 3, 5, 7, 9],
'y': [2, 4, 6, 8, 10],
'z': [None, 'a', 'b', 'c', 'abc']
})
ge_df = ge.from_pandas(pd_df)
self.assertIsInstance(ge_df, ge.dataset.Dataset)
self.assertEquals(list(ge_df.columns), ['x', 'y', 'z'])
self.assertEquals(list(ge_df['x']), list(pd_df['x']))
self.assertEquals(list(ge_df['y']), list(pd_df['y']))
self.assertEquals(list(ge_df['z']), list(pd_df['z']))
def sanity_check(date, bucket):
s3 = s3fs.S3FileSystem()
output = ge.from_pandas(
pq.ParquetDataset("s3://" + bucket + "/usp/output/date=" + date,
filesystem=s3).read_pandas().to_pandas())
output.expect_column_values_to_be_in_set('accommodation_ns', [100])
status = output.validate()
print(status)
if not status['success']:
raise ValueError
def _file_passes(_info, df):
with open(file_path) as ff:
expt_config = json.load(ff)
# This is necessary because ge ends up coercing a type change
# on the same dataframe instance, changing the type. But a
# ge can't be copied, because it causes an error.
# The error is
# AttributeError: 'PandasDataset' object has no attribute 'discard_subset_failing_expectations'
# See https://github.com/great-expectations/great_expectations/issues/342
df_copy = copy.deepcopy(df)
ge_df = ge.from_pandas(df_copy, expt_config)
validate_result = ge_df.validate()
check.invariant('success' in validate_result)
check.invariant('results' in validate_result)
return ExpectationResult(success=validate_result['success'],
result_context=validate_result)
def data_qual(df: pd.DataFrame):
"""
Функция для тестирования данных в pipeline
:param df:
:return:
"""
df = ge.from_pandas(df)
# создаем проверки
result = df.expect_column_values_to_be_in_set('Subscriber_Type',
list(df['Subscriber_Type'].unique()),
mostly=.95)
if not result['success']:
err = result["exception_info"]
raise Exception(f"You get unexpected data in Subscriber_Type column\n{err}")
def test_dataframe(df):
"""
Test the resulting dataframe to ensure data integrity. Can be expanded
if further tests are needed.
"""
print("Performing final validation test on dataframe")
ge_df = ge.from_pandas(df)
# Check nulls
columns_not_null = df.columns
for column in columns_not_null:
result = ge_df.expect_column_values_to_not_be_null(column)
assert (result["success"] == True
), "Coluns contain null value. Please check data."
# Ensure result is only between 0 and 1
result = ge_df.expect_column_values_to_be_between("result",
min_value=0,
max_value=1)
assert result[
"success"] == True, "Results not between 0 and 1, please check data"
# Make sure the gold delta does not have outliders
gold_delta = [
"golddiffat10",
"golddiffat15",
"teamgolddiffat10",
"teamgolddiffat15",
]
for column in gold_delta:
result = ge_df.expect_column_values_to_be_between(column,
min_value=-20000,
max_value=20000)
assert result[
"success"] == True, "Gold diff values too high, please check data"
print("Data validation passed.")
return df
def test_failure_map_values_unique(self):
dataset_ge = ge.from_pandas(TEST_DATA)
dataset_ge.set_default_expectation_argument("result_format",
"COMPLETE")
dataset_ge.expect_column_values_to_be_unique("column_one")
failure_map = common.ge_results_to_failure_map(
{"dataset": dataset_ge.validate()})
self.assertEqual(
{
"dataset": {
common.EXPECT_VALUES_UNIQUE_KEY: {
common.COLUMN_NAME_KEY: "column_one",
common.FAILED_VALUES_KEY: ["value1"],
}
}
},
failure_map,
)
def test_from_pandas_expectations_config():
# Logic mostly copied from TestValidation.test_validate
def load_ge_config(file):
with open(file) as f:
return json.load(f)
my_expectations_config = load_ge_config(
"./tests/test_sets/titanic_expectations.json")
pd_df = pd.read_csv("./tests/test_sets/Titanic.csv")
my_df = ge.from_pandas(pd_df, expectations_config=my_expectations_config)
my_df.set_default_expectation_argument("result_format", "COMPLETE")
results = my_df.validate(catch_exceptions=False)
expected_results = load_ge_config(
"./tests/test_sets/expected_results_20180303.json")
assertDeepAlmostEqual(results, expected_results)
def _get_shape_expectations(self,
dataset: pd.DataFrame) -> ge.dataset.Dataset:
"""
Validates dataset shape.
Validations:
- Dataset columns names are a subset of mapped column names and table_schema field names
"""
dataset_ge = ge.from_pandas(dataset, dataset_class=ShapePandasDataset)
dataset_ge.set_default_expectation_argument("result_format",
"COMPLETE")
valid_field_names: List[str] = table_schema.get_valid_field_names(
self.table_schema, self.row_format)
valid_cols: List[str] = list(
self.column_mapping.keys()) + valid_field_names
dataset_ge.expect_table_columns_to_be_in_set(valid_cols)
dataset_ge.expect_named_cols()
return dataset_ge
def test_failure_map_values_in_set(self):
dataset_ge = ge.from_pandas(TEST_DATA)
dataset_ge.set_default_expectation_argument("result_format",
"COMPLETE")
dataset_ge.expect_column_values_to_be_in_set(
"column_two", set(["invalid_value", "row1_col2"]))
failure_map = common.ge_results_to_failure_map(
{"dataset": dataset_ge.validate()})
self.assertEqual(
{
"dataset": {
common.EXPECT_VALUES_IN_SET_KEY: {
common.COLUMN_NAME_KEY: "column_two",
common.FAILED_VALUES_KEY: ["row2_col2"],
}
}
},
failure_map,
)
def test_failure_map_values(self):
dataset_ge = ge.from_pandas(TEST_DATA,
dataset_class=ShapePandasDataset)
dataset_ge.set_default_expectation_argument("result_format",
"COMPLETE")
dataset_ge.expect_table_columns_to_be_in_set(
set(["column_one", "other_column"]))
failure_map = common.ge_results_to_failure_map(
{"dataset": dataset_ge.validate()})
self.assertEqual(
{
"dataset": {
common.EXPECT_COLUMNS_IN_SET_KEY: {
common.FAILED_VALUES_KEY: ["column_two"]
}
}
},
failure_map,
)
def duplicate_and_obfuscuate(df):
df_a = df.copy()
df_b = df.copy()
df_a["group"] = "a"
df_b["group"] = "b"
for column in df_b.columns:
if column == "group":
continue
if df_b[column].dtype in ["int", "float"]:
df_b[column] += np.max(df_b[column])
continue
if df_b[column].dtype == "object" and all(
[(type(elem) == str) or (elem is None) for elem in df_b[column]]
):
df_b[column] = df_b[column].astype(str) + "__obfuscate"
continue
return ge.from_pandas(pd.concat([df_a, df_b], ignore_index=True, sort=False))
def __init__(self, df, meta_data):
"""
Takes a pandas dataframe and a table meta data object and checks
the values in the dataframe against the meta data.
"""
if not isinstance(df, pd.DataFrame):
raise TypeError("df must be a pandas dataframe object")
self.meta_data = meta_data
self.validate_meta_data()
self.meta_cols = meta_data["columns"]
# Placeholer for proper schema check
if not isinstance(self.meta_cols, list):
raise TypeError("meta_cols must be a list of objects")
# This never fails, but the resultant types are not guaranteed to be correct
df = impose_metadata_types_on_pd_df(df, meta_data)
self.df_ge = ge.from_pandas(df)
self.vlog = ValidationLog(self)
def _get_expectations(self, column_mapping: pd.DataFrame) -> ge.dataset.Dataset:
"""
Returns great_expectations object for a pd.DataFrame with expectations attached.
If not all expectations have been satisfied, this function may fail early for
readability of the failed expectation(s).
Expectations:
- Has columns matching exactly to COLUMN_NAMES
- Does not have any repeated local column names
- All supposed GII columns are valid ones
"""
column_mapping_ge = ge.from_pandas(column_mapping)
column_mapping_ge.set_default_expectation_argument("result_format", "COMPLETE")
# Shape
shape_expectation = column_mapping_ge.expect_table_columns_to_match_ordered_list(
COLUMN_NAMES
)
# If the shape isn't correct then the following expectations will raise exceptions
if not shape_expectation["success"]:
return column_mapping_ge
# Check that there are no repeats of an internal column name
column_mapping_ge.expect_column_values_to_be_unique(
INTERNAL_COLUMN_NAME_COLUMN_NAME
)
# Check that all Mission Impact field names are valid
valid_field_names = table_schema.get_valid_field_names(
self.table_schema, self.row_format
)
column_mapping_ge.expect_column_values_to_be_in_set(
MI_FIELD_NAME_COLUMN_NAME, valid_field_names
)
return column_mapping_ge
def test_failure_map_column_ordered_list(self):
dataset_ge = ge.from_pandas(TEST_DATA)
dataset_ge.set_default_expectation_argument("result_format",
"COMPLETE")
dataset_ge.expect_table_columns_to_match_ordered_list(
["column_one", "other_column"])
failure_map = common.ge_results_to_failure_map(
{"dataset": dataset_ge.validate()})
self.assertEqual(
{
"dataset": {
common.EXPECT_COLUMNS_MATCH_KEY: {
common.EXPECTED_ORDERED_LIST_KEY: [
"column_one",
"other_column",
],
common.FAILED_VALUES_KEY: ["column_two"],
}
}
},
failure_map,
)
def _do_expectation(_info, df):
ge_df = ge.from_pandas(df)
ge_result = ge_callback(ge_df)
check.invariant('success' in ge_result)
return ExpectationResult(success=ge_result['success'],
result_context=ge_result)
RVMS_Current_Budgeted_CPU['ModelName'] = RVMS_Current_Budgeted_CPU.apply(
getModelName, axis='columns')
RVMS_Current_Budgeted_CPU['DestCode'] = RVMS_Current_Budgeted_CPU.apply(
getDestCode, axis='columns')
# %%
RVMS_Current_Budgeted_CPU.head()
# %% [markdown]
# ### Perform data validation checks using Great Expectations library
# %% [markdown]
# #### Create Great Expectations dataframe from pandas dataframe:
# %%
ge_df = ge.from_pandas(RVMS_Current_Budgeted_CPU)
# %% [markdown]
# #### Check Model Years are between 1994 and 2099
# %%
if ge_df.expect_column_values_to_be_between(column="ModelYear",
min_value='1994',
max_value='2099')['success']:
print('Passed Model Year Check')
else:
print('FAILED Model Year Check')
toaster = ToastNotifier()
toaster.show_toast("### Check Status ###",
"FAILED Model Year Check",
icon_path="images/honda_logo.ico",
def test_errors_warnings_validation_operator_run_slack_query(
basic_data_context_config_for_validation_operator, tmp_path_factory,
filesystem_csv_4):
#####
#####
#
# WARNING: PY2 SUPPORT IS UNTESTED BECAUSE OF DICTIONARY ORDER ISSUES NOT YET RESOLVED
#
#####
#####
if PY2:
pytest.skip(
"skipping test_errors_warnings_validation_operator_run_slack_query in py2"
)
project_path = str(tmp_path_factory.mktemp('great_expectations'))
# NOTE: This setup is almost identical to test_DefaultDataContextAwareValidationOperator.
# Consider converting to a single fixture.
data_context = ConfigOnlyDataContext(
basic_data_context_config_for_validation_operator,
project_path,
)
data_context.add_datasource("my_datasource",
class_name="PandasDatasource",
base_directory=str(filesystem_csv_4))
data_context.create_expectation_suite(
data_asset_name="my_datasource/default/f1",
expectation_suite_name="failure")
df = data_context.get_batch("my_datasource/default/f1",
"failure",
batch_kwargs=data_context.yield_batch_kwargs(
"my_datasource/default/f1"))
df.expect_column_values_to_be_between(column="x", min_value=1, max_value=9)
failure_expectations = df.get_expectation_suite(
discard_failed_expectations=False)
data_context.save_expectation_suite(
failure_expectations,
data_asset_name="my_datasource/default/f1",
expectation_suite_name="failure")
data_context.create_expectation_suite(
data_asset_name="my_datasource/default/f1",
expectation_suite_name="warning")
df = data_context.get_batch("my_datasource/default/f1",
"warning",
batch_kwargs=data_context.yield_batch_kwargs(
"my_datasource/default/f1"))
df.expect_column_values_to_be_between(column="x", min_value=1, max_value=9)
df.expect_column_values_to_not_be_null(column="y")
warning_expectations = df.get_expectation_suite(
discard_failed_expectations=False)
data_context.save_expectation_suite(
warning_expectations,
data_asset_name="my_datasource/default/f1",
expectation_suite_name="warning")
data_context.save_expectation_suite(
failure_expectations,
data_asset_name="my_datasource/default/f2",
expectation_suite_name="failure")
data_context.save_expectation_suite(
failure_expectations,
data_asset_name="my_datasource/default/f3",
expectation_suite_name="failure")
data_context.save_expectation_suite(
warning_expectations,
data_asset_name="my_datasource/default/f2",
expectation_suite_name="warning")
data_context.save_expectation_suite(
warning_expectations,
data_asset_name="my_datasource/default/f3",
expectation_suite_name="warning")
vo = WarningAndFailureExpectationSuitesValidationOperator(
data_context=data_context,
action_list=[],
slack_webhook="https://hooks.slack.com/services/test/slack/webhook")
my_df_1 = pd.DataFrame({"x": [1, 2, 3, 4, 5], "y": [1, 2, 3, 4, None]})
my_ge_df_1 = ge.from_pandas(my_df_1)
my_ge_df_1._expectation_suite["data_asset_name"] = DataAssetIdentifier(
"my_datasource", "default", "f1")
my_df_2 = pd.DataFrame({"x": [1, 2, 3, 4, 99], "y": [1, 2, 3, 4, 5]})
my_ge_df_2 = ge.from_pandas(my_df_2)
my_ge_df_2._expectation_suite["data_asset_name"] = DataAssetIdentifier(
"my_datasource", "default", "f2")
my_df_3 = pd.DataFrame({"x": [1, 2, 3, 4, 5], "y": [1, 2, 3, 4, 5]})
my_ge_df_3 = ge.from_pandas(my_df_3)
my_ge_df_3._expectation_suite["data_asset_name"] = DataAssetIdentifier(
"my_datasource", "default", "f3")
return_obj = vo.run(
assets_to_validate=[my_ge_df_1, my_ge_df_2, my_ge_df_3],
run_id="test_100")
slack_query = vo._build_slack_query(return_obj)
expected_slack_query = {
'blocks': [{
'type': 'divider'
}, {
'type': 'section',
'text': {
'type': 'mrkdwn',
'text': '*FailureVsWarning Validation Operator Completed.*'
}
}, {
'type': 'divider'
}, {
'type': 'section',
'text': {
'type': 'mrkdwn',
'text': '*Status*: Failed :x:'
}
}, {
'type': 'section',
'text': {
'type':
'mrkdwn',
'text':
'*Data Asset List:* [my_datasource/default/f1, my_datasource/default/f2, my_datasource/default/f3]'
}
}, {
'type': 'section',
'text': {
''
'type': 'mrkdwn',
'text': '*Failed Data Assets:* [my_datasource/default/f2]'
}
}, {
'type': 'section',
'text': {
'type': 'mrkdwn',
'text': '*Run ID:* test_100'
}
}, {
'type': 'section',
'text': {
'type': 'mrkdwn',
'text': '*Timestamp:* 09/26/2019 13:42:41'
}
}, {
'type': 'divider'
}, {
'type':
'context',
'elements': [{
'type':
'mrkdwn',
'text':
'Learn about FailureVsWarning Validation Operators at https://docs.greatexpectations.io/en/latest/reference/validation_operators/warning_and_failure_expectation_suites_validation_operator.html'
}]
}]
}
# We're okay with system variation in locales (OS X likes 24 hour, but not Travis)
slack_query['blocks'][7]['text']['text'] = \
slack_query['blocks'][7]['text']['text'].replace('09/26/2019 13:42:41', 'LOCALEDATE')
slack_query['blocks'][7]['text']['text'] = \
slack_query['blocks'][7]['text']['text'].replace('09/26/2019 01:42:41 PM', 'LOCALEDATE')
expected_slack_query['blocks'][7]['text']['text'] = \
expected_slack_query['blocks'][7]['text']['text'].replace('09/26/2019 13:42:41', 'LOCALEDATE')
expected_slack_query['blocks'][7]['text']['text'] = \
expected_slack_query['blocks'][7]['text']['text'].replace('09/26/2019 01:42:41 PM', 'LOCALEDATE')
import json
print(json.dumps(slack_query, indent=2))
print(json.dumps(expected_slack_query, indent=2))
assert slack_query == expected_slack_query
import pandas
from pandas_profiling import ProfileReport
import great_expectations as ge
from great_expectations.profile.basic_dataset_profiler import BasicDatasetProfiler
if __name__ == "__main__":
with mlflow.start_run(run_name="check_verify_data") as run:
mlflow.set_tag("mlflow.runName", "check_verify_data")
df = pandas.read_csv("./data/raw/data.csv")
describe_to_dict = df.describe().to_dict()
mlflow.log_dict(describe_to_dict, "describe_data.json")
pd_df_ge = ge.from_pandas(df)
assert pd_df_ge.expect_column_values_to_match_strftime_format(
"Date", "%Y-%m-%d").success == True
assert pd_df_ge.expect_column_values_to_be_of_type(
"High", "float").success == True
assert pd_df_ge.expect_column_values_to_be_of_type(
"Low", "float").success == True
assert pd_df_ge.expect_column_values_to_be_of_type(
"Open", "float").success == True
assert pd_df_ge.expect_column_values_to_be_of_type(
"Close", "float").success == True
assert pd_df_ge.expect_column_values_to_be_of_type(
"Volume", "long").success == True
assert pd_df_ge.expect_column_values_to_be_of_type(
"Adj Close", "float").success == True
def _get_batch_kwargs(df: pd.DataFrame) -> dict:
dataset = ge.from_pandas(df)
return {"dataset": dataset, "datasource": "pandas"}
