def test_should_check_all_given_columns_separately(spark_session):
df = spark_session.createDataFrame(
[["a", "12"], ["abcde", "56"], ["def", "123"]],
schema=two_string_columns_schema)
expected_correct = spark_session.createDataFrame(
[], schema=two_string_columns_schema)
expected_errors = spark_session.createDataFrame(
[["a", "12"], ["abcde", "56"], ["def", "123"]],
schema=two_string_columns_schema)
result = ValidateSparkDataFrame(spark_session, df) \
.has_length_between("col1", 2, 4) \
.has_length_between("col2", 1, 2) \
.execute()
AssertDf(result.correct_data, order_by_column="col1") \
.contains_exactly(expected_correct.toPandas()) \
.has_columns(["col1", "col2"])
AssertDf(result.erroneous_data, order_by_column="col2") \
.contains_exactly(expected_errors.toPandas()) \
.has_columns(["col1", "col2"])
assert result.errors == [
ValidationError("col1", "text_length", 2),
ValidationError("col2", "text_length", 1)
]
def test_min_should_check_all_given_columns_separately(spark_session):
df = spark_session.createDataFrame([[5, 1], [10, 2], [15, 3]],
schema=two_integer_columns_schema)
expected_correct = spark_session.createDataFrame(
[], schema=two_integer_columns_schema)
expected_errors = spark_session.createDataFrame(
[[5, 1], [10, 2], [15, 3]], schema=two_integer_columns_schema)
result = ValidateSparkDataFrame(spark_session, df) \
.is_min("col1", 20) \
.is_min("col2", 5) \
.execute()
AssertDf(result.correct_data, order_by_column="col1") \
.contains_exactly(expected_correct.toPandas()) \
.has_columns(["col1", "col2"])
AssertDf(result.erroneous_data, order_by_column="col2") \
.contains_exactly(expected_errors.toPandas()) \
.has_columns(["col1", "col2"])
assert result.errors == [
ValidationError("col1", "min", 3),
ValidationError("col2", "min", 3)
]
def test_should_return_rows_that_pass_all_checks_and_reject_rows_that_violate_any_test(spark_session):
not_between = [25, 1]
max_exceeded = [3, 30]
correct = [3, 15]
less_than_min = [1, 15]
both_wrong = [7, 30]
df = spark_session.createDataFrame([not_between, max_exceeded, correct, less_than_min, both_wrong], schema=two_integer_columns_schema)
expected_correct = spark_session.createDataFrame([correct], schema=two_integer_columns_schema)
expected_errors = spark_session.createDataFrame([not_between, max_exceeded, less_than_min, both_wrong], schema=two_integer_columns_schema)
result = ValidateSparkDataFrame(spark_session, df) \
.is_between("col1", 0, 5) \
.is_min("col1", 3) \
.is_max("col2", 20) \
.execute()
AssertDf(result.correct_data, order_by_column="col1") \
.contains_exactly(expected_correct.toPandas()) \
.has_columns(["col1", "col2"])
AssertDf(result.erroneous_data, order_by_column="col2") \
.contains_exactly(expected_errors.toPandas()) \
.has_columns(["col1", "col2"])
assert result.errors == [ValidationError("col1", "between", 2), ValidationError("col1", "min", 1), ValidationError("col2", "max", 2)]
def check(self, *, actual: ValidationResult, expected_correct: DataFrame,
expected_erroneous: DataFrame):
if expected_correct.count() == 0:
AssertDf(actual.correct_data) \
.is_empty() \
.has_columns(expected_correct.columns)
else:
AssertDf(actual.correct_data, order_by_column=self.column_name) \
.contains_exactly(expected_correct.toPandas()) \
.has_columns(expected_correct.columns)
if expected_erroneous.count() == 0:
AssertDf(actual.erroneous_data) \
.is_empty() \
.has_columns(expected_erroneous.columns)
else:
AssertDf(actual.erroneous_data, order_by_column=self.column_name) \
.contains_exactly(expected_erroneous.toPandas()) \
.has_columns(expected_erroneous.columns)
if expected_erroneous.count() == 0:
assert actual.errors == []
else:
assert actual.errors == [
ValidationError(self.column_name, self.constraint_name,
expected_erroneous.count())
]
def test_should_reject_all_rows_if_all_are_the_same(spark_session):
df = spark_session.createDataFrame([["abc"], ["abc"], ["abc"]], schema=single_string_column_schema)
expected_errors = spark_session.createDataFrame([["abc"]], schema=single_string_column_schema)
result = ValidateSparkDataFrame(spark_session, df) \
.is_unique("col1") \
.execute()
AssertDf(result.correct_data) \
.is_empty() \
.has_columns(["col1"])
AssertDf(result.erroneous_data, order_by_column="col1") \
.contains_exactly(expected_errors.toPandas()) \
.has_columns(["col1"])
assert result.errors == [ValidationError("col1", "unique", 3)]
def test_not_null_should_check_all_given_columns_separately_even_if_all_of_them_are_defined_at_once(spark_session):
df = spark_session.createDataFrame([["abc", None], [None, "456"], [None, None]], schema=two_string_columns_schema)
expected_errors = spark_session.createDataFrame([["abc", None], [None, "456"], [None, None]], schema=two_string_columns_schema)
result = ValidateSparkDataFrame(spark_session, df) \
.are_not_null(["col1", "col2"]) \
.execute()
AssertDf(result.correct_data) \
.is_empty() \
.has_columns(["col1", "col2"])
AssertDf(result.erroneous_data, order_by_column=["col1", "col2"]) \
.contains_exactly(expected_errors.toPandas()) \
.has_columns(["col1", "col2"])
assert result.errors == [ValidationError("col1", "not_null", 2), ValidationError("col2", "not_null", 2)]
def test_uniqueness_of_other_columns_is_ignored(spark_session):
df = spark_session.createDataFrame([["abc", "123"], ["abc", "456"], ["def", "123"]], schema=two_string_columns_schema)
expected_correct = spark_session.createDataFrame([["def", "123"]], schema=two_string_columns_schema)
expected_errors = spark_session.createDataFrame([["abc", "123"], ["abc", "456"]], schema=two_string_columns_schema)
result = ValidateSparkDataFrame(spark_session, df) \
.is_unique("col1") \
.execute()
AssertDf(result.correct_data, order_by_column="col1") \
.contains_exactly(expected_correct.toPandas()) \
.has_columns(["col1", "col2"])
AssertDf(result.erroneous_data, order_by_column="col2") \
.contains_exactly(expected_errors.toPandas()) \
.has_columns(["col1", "col2"])
assert result.errors == [ValidationError("col1", "unique", 2)]
def test_uniqueness_should_check_all_given_columns_separately_when_defining_all_columns_at_once(spark_session):
df = spark_session.createDataFrame([["abc", "123"], ["abc", "456"], ["def", "123"]], schema=two_string_columns_schema)
expected_correct = spark_session.createDataFrame([], schema=two_string_columns_schema)
expected_errors = spark_session.createDataFrame([["abc", "123"], ["abc", "456"], ["def", "123"]], schema=two_string_columns_schema)
result = ValidateSparkDataFrame(spark_session, df) \
.are_unique(["col1", "col2"]) \
.execute()
AssertDf(result.correct_data, order_by_column="col1") \
.contains_exactly(expected_correct.toPandas()) \
.has_columns(["col1", "col2"])
AssertDf(result.erroneous_data, order_by_column="col2") \
.contains_exactly(expected_errors.toPandas()) \
.has_columns(["col1", "col2"])
assert result.errors == [ValidationError("col1", "unique", 2), ValidationError("col2", "unique", 2)]
def test_between_ignores_the_other_column(spark_session):
df = spark_session.createDataFrame([[5, 8], [10, 20], [15, 8]],
schema=two_integer_columns_schema)
expected_correct = spark_session.createDataFrame(
[[5, 8], [10, 20]], schema=two_integer_columns_schema)
expected_errors = spark_session.createDataFrame(
[[15, 8]], schema=two_integer_columns_schema)
result = ValidateSparkDataFrame(spark_session, df) \
.is_between("col1", 5, 10) \
.execute()
AssertDf(result.correct_data, order_by_column="col1") \
.contains_exactly(expected_correct.toPandas()) \
.has_columns(["col1", "col2"])
AssertDf(result.erroneous_data, order_by_column="col2") \
.contains_exactly(expected_errors.toPandas()) \
.has_columns(["col1", "col2"])
assert result.errors == [ValidationError("col1", "between", 1)]
def test_spark_sql_operation(spark_session):
df_schema = StructType([
StructField("col1", StringType()),
StructField("col2", IntegerType())
])
test_list = [["v1", 1], ["v1", 2], ["v2", 3]]
df: DataFrame = spark_session.createDataFrame(test_list, schema=df_schema)
aggregated = df.groupby("col1").sum("col2").orderBy('col1')
AssertDf(aggregated) \
.contains_exactly(pd.DataFrame([['v1', 3], ['v2', 3]], columns=['col1', 'sum(col2)']).sort_values('col1')) \
.has_columns(["col1", "sum(col2)"]) \
.has_n_rows(2)
def test_empty_dataframe(spark_session):
df_schema = StructType([StructField("col1", StringType())])
df = spark_session.createDataFrame([], schema=df_schema)
AssertDf(df).is_empty()