def test_singlecolumns_partialnullspartialdistinct(self):
data = pd.DataFrame()
# create and assign columns to df
l1 = [chr(i) for i in range(100)]
l2 = [i for i in range(100)]
l3 = [i / 0.7 for i in range(100)]
for i in range(20):
l1[i] = ""
l2[i] = 0
l3[i] = 0.
for i in range(20, 40):
l1[i] = "zzzzz"
l2[i] = 500
l3[i] = 402.2
random.shuffle(l1)
random.shuffle(l2)
random.shuffle(l3)
data["c1"] = l1
data["c2"] = l2
data["c3"] = l3
df = self.spark.createDataFrame(data)
df = df.withColumn("c1", replace_empty_with_null(df["c1"]))
df = df.withColumn("c2", replace_0_with_null(df["c2"]))
df = df.withColumn("c3", replace_0dot_with_null(df["c3"]))
r1, r2, r3 = deduplication(["c1", "c2", "c3"], df)
self.assertEqual(r1, 61.0)
self.assertEqual(r2, 61.0)
self.assertEqual(r3, 61.0)
def test_wholetable_partialnullspartialdistinct(self):
data = pd.DataFrame()
# create and assign columns to df
l1 = [chr(i) for i in range(100)]
l2 = [i for i in range(100)]
l3 = [i / 0.7 for i in range(100)]
for i in range(20):
l1[i] = ""
for i in range(20, 40):
l2[i] = 0
for i in range(40, 60):
l3[i] = 0.
for i in range(60, 80):
l1[i] = "zzz"
l2[i] = 133
l3[i] = 231.22
data["c1"] = l1
data["c2"] = l2
data["c3"] = l3
df = self.spark.createDataFrame(data)
df = df.withColumn("c1", replace_empty_with_null(df["c1"]))
df = df.withColumn("c2", replace_0_with_null(df["c2"]))
df = df.withColumn("c3", replace_0dot_with_null(df["c3"]))
r = deduplication(df=df)[0]
self.assertEqual(r, 21.0)
def test_wholetable_alldifferent(self):
data = pd.DataFrame()
data["c1"] = [chr(i) for i in range(100)]
data["c2"] = [i for i in range(100)]
data["c3"] = [i / 0.7 for i in range(100)]
df = self.spark.createDataFrame(data)
r = deduplication(df=df)[0]
self.assertEqual(r, 100.0)
def test_wholetable_allsame(self):
data = pd.DataFrame()
data["c1"] = [chr(0) for _ in range(100)]
data["c2"] = [10 for _ in range(100)]
data["c3"] = [20 / 0.7 for _ in range(100)]
df = self.spark.createDataFrame(data)
r = deduplication(df=df)[0]
self.assertEqual(r, 1.0)
def test_wholetable_empty(self):
data = pd.DataFrame()
data["c1"] = []
data["c2"] = []
schema = [StructField("c1", IntegerType(), True), StructField("c2", StringType(), True)]
df = self.spark.createDataFrame(data, StructType(schema))
r = deduplication(df=df)[0]
self.assertEqual(r, 100.)
def test_singlecolumns_empty(self):
data = pd.DataFrame()
data["c1"] = []
data["c2"] = []
schema = [StructField("c1", IntegerType(), True), StructField("c2", StringType(), True)]
df = self.spark.createDataFrame(data, StructType(schema))
r1, r2 = deduplication(["c1", "c2"], df)
self.assertEqual(r1, 100.)
self.assertEqual(r2, 100.)
def test_singlecolumns_alldifferent(self):
data = pd.DataFrame()
data["c1"] = [chr(i) for i in range(100)]
data["c2"] = [i for i in range(100)]
data["c3"] = [i / 0.7 for i in range(100)]
df = self.spark.createDataFrame(data)
r1, r2, r3 = deduplication(["c1", "c2", "c3"], df)
self.assertEqual(r1, 100.0)
self.assertEqual(r2, 100.0)
self.assertEqual(r3, 100.0)
def test_singlecolumns_allsame(self):
data = pd.DataFrame()
data["c1"] = [chr(0) for _ in range(100)]
data["c2"] = [10 for _ in range(100)]
data["c3"] = [20 / 0.7 for _ in range(100)]
df = self.spark.createDataFrame(data)
r1, r2, r3 = deduplication(["c1", "c2", "c3"], df)
self.assertEqual(r1, 1.0)
self.assertEqual(r2, 1.0)
self.assertEqual(r3, 1.0)
def test_wholetable_allnull(self):
data = pd.DataFrame()
data["c1"] = [chr(i) for i in range(100)]
data["c2"] = [i for i in range(100)]
data["c3"] = [i / 0.7 for i in range(100)]
df = self.spark.createDataFrame(data)
df = df.withColumn("c1", replace_every_string_with_null(df["c1"]))
df = df.withColumn("c2", replace_every_int_with_null(df["c2"]))
df = df.withColumn("c3", replace_every_float_with_null(df["c3"]))
r = deduplication(df=df)[0]
self.assertEqual(r, 0.0)
def test_singlecolumns_allnull(self):
data = pd.DataFrame()
data["c1"] = [chr(i) for i in range(100)]
data["c2"] = [i for i in range(100)]
data["c3"] = [i / 0.7 for i in range(100)]
df = self.spark.createDataFrame(data)
df = df.withColumn("c1", replace_every_string_with_null(df["c1"]))
df = df.withColumn("c2", replace_every_int_with_null(df["c2"]))
df = df.withColumn("c3", replace_every_float_with_null(df["c3"]))
r1, r2, r3 = deduplication(["c1", "c2", "c3"], df)
self.assertEqual(r1, 0.0)
self.assertEqual(r2, 0.0)
self.assertEqual(r3, 0.0)
def test_wholetable_partialnulls(self):
data = pd.DataFrame()
# create and assign columns to df
l1 = [chr(i) for i in range(100)]
l2 = [10 for _ in range(100)]
l3 = [0.7 for _ in range(100)]
for i in range(20):
l1[i] = ""
random.shuffle(l1)
data["c1"] = l1
data["c2"] = l2
data["c3"] = l3
df = self.spark.createDataFrame(data)
df = df.withColumn("c1", replace_empty_with_null(df["c1"]))
r = deduplication(df=df)[0]
self.assertEqual(r, 80.0)
def test_wholetable_partial(self):
data = pd.DataFrame()
# create and assign columns to df
l1 = [chr(i) for i in range(100)]
l2 = [i for i in range(100)]
l3 = [i / 0.7 for i in range(100)]
for i in range(20):
l1[i] = ""
l2[i] = 0
l3[i] = 0.
data["c1"] = l1
data["c2"] = l2
data["c3"] = l3
df = self.spark.createDataFrame(data)
r = deduplication(df=df)[0]
self.assertEqual(r, 81.0)
def test_singlecolumns_partial(self):
data = pd.DataFrame()
# create and assign columns to df
l1 = [chr(i) for i in range(100)]
l2 = [i for i in range(100)]
l3 = [i / 0.7 for i in range(100)]
for i in range(20):
l1[i] = ""
l2[i] = 0
l3[i] = 0.
random.shuffle(l1)
random.shuffle(l2)
random.shuffle(l3)
data["c1"] = l1
data["c2"] = l2
data["c3"] = l3
df = self.spark.createDataFrame(data)
r1, r2, r3 = deduplication(["c1", "c2", "c3"], df)
self.assertEqual(r1, 81.0)
self.assertEqual(r2, 81.0)
self.assertEqual(r3, 81.0)
def test_groping_multile_columns(self):
data = pd.DataFrame()
c1 = [0, 0, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 5]
c2 = ["a", "a", "b", "b", "c", "c", "d", "d", "d", "d", "a", "a", "a"]
c3 = [0.0, 0.0, 0.1, 0.1, 2.2, 2.2, 2.2, 3.1, 3.2, 3.3, 40, 40, 50]
c4 = [
10.0, 20.0, 10.0, 20.0, 10.0, 20.0, 10.0, 20.0, 10.0, 20.0, 10.0,
20.0, 10.0
]
c5 = ["09:10:10" for _ in range(10)]
c5.extend(["00:11:10" for _ in range(3)])
data["c1"] = c1
data["c2"] = c2
data["c3"] = c3
data["c4"] = c4
data["c5"] = c5
df = self.spark.createDataFrame(data)
task = Task()
task.add(completeness())
task.add(completeness([0, 1, 2]))
task.add(deduplication([0, 1]))
task.add(deduplication())
task.add(timeliness(["c5"], value="10:10:10", timeFormat="ss:mm:HH"))
task.add(completeness())
condition1 = {"column": "c3", "operator": "lt", "value": 50}
condition2 = {"column": "c3", "operator": "gt", "value": 1.0}
conditions = [condition1, condition2]
task.add(rule(conditions))
condition1 = {"column": "c5", "operator": "eq", "value": "00:11:10"}
conditions = [condition1]
task.add(rule(conditions))
condition1 = {"column": "c3", "operator": "lt", "value": 50}
condition2 = {"column": "c3", "operator": "gt", "value": 1.0}
conditions = [condition1, condition2]
having1 = {
"column": "*",
"operator": "gt",
"value": 1,
"aggregator": "count"
}
having2 = {
"column": "c4",
"operator": "eq",
"value": 50 / 3,
"aggregator": "avg"
}
havings = [having1, having2]
task.add(grouprule([0, "c2"], havings, conditions))
result = task.run(df)
# c1
r = result[0]["scores"][0]
self.assertEqual(r, 100.)
# c2
r1, r2, r3 = result[1]["scores"]
self.assertEqual(r1, 100.)
self.assertEqual(r2, 100.)
self.assertEqual(r3, 100.)
# d1
r1, r2 = result[2]["scores"]
self.assertEqual(r1, (6 / 13) * 100)
self.assertEqual(r2, (4 / 13) * 100)
# d2
r = result[3]["scores"][0]
self.assertEqual(r, 100.)
# t
r = result[4]["scores"][0]
self.assertEqual(r, (10 / 13) * 100)
# c3
r = result[5]["scores"][0]
self.assertEqual(r, 100.)
# r1
r = result[6]["scores"][0]
self.assertEqual(r, (8 / 13) * 100)
# r2
r = result[7]["scores"][0]
self.assertEqual(r, (3 / 13) * 100)
# gr1
r = result[8]["scores"][0]
self.assertEqual(r, 25.0)
#!/usr/bin/python3
from pyspark.sql import SparkSession
from haychecker.dhc.metrics import deduplication
spark = SparkSession.builder.appName("deduplication_example").getOrCreate()
df = spark.read.format("csv").option(
"header", "true").load("examples/resources/employees.csv")
df.show()
r1, r2 = deduplication(["title", "city"], df)
print("Deduplication title: {}, deduplication city: {}".format(r1, r2))
task1 = deduplication(["title", "city"])
task2 = deduplication(["lastName"])
task3 = task1.add(task2)
result = task3.run(df)
r1, r2 = result[0]["scores"]
r3 = result[1]["scores"][0]
print(
"Deduplication title: {}, deduplication city: {}, deduplication lastName: {}"
.format(r1, r2, r3))