def test_resample_millis(self):
"""Test of resampling for millisecond windows"""
schema = StructType([
StructField("symbol", StringType()),
StructField("date", StringType()),
StructField("event_ts", StringType()),
StructField("trade_pr", FloatType()),
StructField("trade_pr_2", FloatType())
])
expectedSchema = StructType([
StructField("symbol", StringType()),
StructField("event_ts", StringType()),
StructField("floor_trade_pr", FloatType()),
StructField("floor_date", StringType()),
StructField("floor_trade_pr_2", FloatType())
])
expectedSchemaMS = StructType([
StructField("symbol", StringType()),
StructField("event_ts", StringType(), True),
StructField("date", DoubleType()),
StructField("trade_pr", DoubleType()),
StructField("trade_pr_2", DoubleType())
])
data = [["S1", "SAME_DT", "2020-08-01 00:00:10.12345", 349.21, 10.0],
["S1", "SAME_DT", "2020-08-01 00:00:10.123", 340.21, 9.0],
["S1", "SAME_DT", "2020-08-01 00:00:10.124", 353.32, 8.0]]
expected_data_ms = [[
"S1", "2020-08-01 00:00:10.123", None, 344.71, 9.5
], ["S1", "2020-08-01 00:00:10.124", None, 353.32, 8.0]]
# construct dataframes
df = self.buildTestDF(schema, data)
dfExpected = self.buildTestDF(expectedSchemaMS, expected_data_ms)
# convert to TSDF
tsdf_left = TSDF(df, partition_cols=["symbol"])
# 30 minute aggregation
resample_ms = tsdf_left.resample(freq="ms", func="mean").df.withColumn(
"trade_pr", F.round(F.col('trade_pr'), 2))
int_df = TSDF(tsdf_left.df.withColumn(
"event_ts",
F.col("event_ts").cast("timestamp")),
partition_cols=['symbol'])
interpolated = int_df.interpolate(freq='ms',
func='floor',
method='ffill')
self.assertDataFramesEqual(resample_ms, dfExpected)
def test_interpolation_using_custom_params(self):
"""Verify that by specifying optional paramters it will change the result of the interpolation based on those modified params."""
self.buildTestingDataFrame()
expected_data = [
["A", "A-1", "2020-01-01 00:00:00", 0.0, False, False],
["A", "A-1", "2020-01-01 00:00:30", 1.0, True, True],
["A", "A-1", "2020-01-01 00:01:00", 2.0, False, False],
["A", "A-1", "2020-01-01 00:01:30", 3.0, False, True],
["A", "A-1", "2020-01-01 00:02:00", 4.0, False, True],
["A", "A-1", "2020-01-01 00:02:30", 5.0, True, True],
["A", "A-1", "2020-01-01 00:03:00", 6.0, True, True],
["A", "A-1", "2020-01-01 00:03:30", 7.0, False, True],
["A", "A-1", "2020-01-01 00:04:00", 8.0, False, False],
["A", "A-1", "2020-01-01 00:04:30", 9.0, True, True],
["A", "A-1", "2020-01-01 00:05:00", 10.0, True, True],
["A", "A-1", "2020-01-01 00:05:30", 11.0, False, False],
]
expected_schema = StructType([
StructField("partition_a", StringType()),
StructField("partition_b", StringType()),
StructField("other_ts_col", StringType(), False),
StructField("value_a", DoubleType()),
StructField("is_ts_interpolated", BooleanType(), False),
StructField("is_interpolated_value_a", BooleanType(), False),
])
# Modify input DataFrame using different ts_col
expected_df: DataFrame = self.buildTestDF(expected_schema,
expected_data,
ts_cols=["other_ts_col"])
input_tsdf = TSDF(
self.simple_input_tsdf.df.withColumnRenamed(
"event_ts", "other_ts_col"),
partition_cols=["partition_a", "partition_b"],
ts_col="other_ts_col",
)
actual_df: DataFrame = input_tsdf.interpolate(
ts_col="other_ts_col",
show_interpolated=True,
partition_cols=["partition_a", "partition_b"],
target_cols=["value_a"],
freq="30 seconds",
func="mean",
method="linear",
).df
assert_df_equality(expected_df, actual_df, ignore_nullable=True)
