问题
I have a PySpark dataframe like this
+----------+--------+----------+----------+
|id_ | p |d1 | d2 |
+----------+--------+----------+----------+
| 1 | A |2018-09-26|2018-10-26|
| 2 | B |2018-06-21|2018-07-19|
| 2 | B |2018-08-13|2018-10-07|
| 2 | B |2018-12-31|2019-02-27|
| 2 | B |2019-05-28|2019-06-25|
| 3 |C |2018-06-15|2018-07-13|
| 3 |C |2018-08-15|2018-10-09|
| 3 |C |2018-12-03|2019-03-12|
| 3 |C |2019-05-10|2019-06-07|
| 4 | A |2019-01-30|2019-03-01|
| 4 | A |2019-05-30|2019-07-25|
| 5 |C |2018-09-19|2018-10-17|
-------------------------------------------
From this I want to create and populate another Pyspark dataframe which have n columns ranging from min(d1) to max(d2) and each column is a date in that range.
I want to populate this dataframe with 1 and 0 for each row.
For row 1 I want populate all the days in the range min(d1 of row 1) to max(d1 of row 1) with 1 and rest columns with a 0. Similarly for all the rows in the dataframe.
I was doing something like this in pandas for this purpose.
result = pd.DataFrame(data = 0, columns=pd.period_range(data['d1'].min(), data['d2'].max(), freq='D'), index=data.index)
for c in result.columns:
result[c] = np.where((c.d2>=data.d1)&(c.d1 <= data.d2), 1, 0)
How to do the same in PySpark.?
回答1:
One way with list comprehension:
UPDATE: per request, adjusted d1, d2 fields from StringType to DateType.
Setup data, modules:
import pandas as pd
from pyspark.sql import functions as F
#... skip the code to initialize SparkSession spark and df
# if d1 and d2 were read as String, convert them to Date using the following.
# Or if the data were already imported with explicit schema or inferSchema=True when running read.csv(), then skip the following:
df = df.withColumn('d1', F.to_date('d1')) \
.withColumn('d2', F.to_date('d2'))
>>> df.show()
+---+---+----------+----------+
|id_| p| d1| d2|
+---+---+----------+----------+
| 1| A|2018-09-26|2018-10-26|
| 2| B|2018-06-21|2018-07-19|
| 2| B|2018-08-13|2018-10-07|
| 2| B|2018-12-31|2019-02-27|
| 2| B|2019-05-28|2019-06-25|
| 3| C|2018-06-15|2018-07-13|
| 3| C|2018-08-15|2018-10-09|
| 3| C|2018-12-03|2019-03-12|
| 3| C|2019-05-10|2019-06-07|
| 4| A|2019-01-30|2019-03-01|
| 4| A|2019-05-30|2019-07-25|
| 5| C|2018-09-19|2018-10-17|
+---+---+----------+----------+
>>> df.printSchema()
root
|-- id_: string (nullable = true)
|-- p: string (nullable = true)
|-- d1: date (nullable = true)
|-- d2: date (nullable = true)
Get min(d1) for start_date and max(d2) for end_date:
d = df.select(F.min('d1').alias('start_date'), F.max('d2').alias('end_date')).first()
>>> d
Row(start_date=datetime.date(2018, 6, 15), end_date=datetime.date(2019, 7, 25))
Get a list of dates and convert them to Date
cols = [ c.to_timestamp().date() for c in pd.period_range(d.start_date, d.end_date, freq='D') ]
>>> cols
[datetime.date(2018, 6, 15),
datetime.date(2018, 6, 16),
...
datetime.date(2019, 7, 23),
datetime.date(2019, 7, 24),
datetime.date(2019, 7, 25)]
Use list comprehension and when() function
Use list comprehension to iterate all dates in cols, F.when(condition,1).otherwise(0) to set up the column values and str(c) for the column names (alias):
result = df.select('id_', *[ F.when((df.d1 <= c)&(df.d2 >= c),1).otherwise(0).alias(str(c)) for c in cols ])
# check data in some columns
result.select('id_', str(d.start_date), '2019-01-01', str(d.end_date)).show()
+---+----------+----------+----------+
|id_|2018-06-15|2019-01-01|2019-07-25|
+---+----------+----------+----------+
| 1| 0| 0| 0|
| 2| 0| 0| 0|
| 2| 0| 0| 0|
| 2| 0| 1| 0|
| 2| 0| 0| 0|
| 3| 1| 0| 0|
| 3| 0| 0| 0|
| 3| 0| 1| 0|
| 3| 0| 0| 0|
| 4| 0| 0| 0|
| 4| 0| 0| 1|
| 5| 0| 0| 0|
+---+----------+----------+----------+
回答2:
Here is an approach (I have taken only few rows and small date ranges to print the output here)
- Identify min and max date (Please note that i have taken min from d1 and max form d2, but you can change if min d2 is less than min d1 for your data)
- Construct the struct, with dates as struct fields
- Populate the struct fields using udf
- Expand the struct by selecting struct.*
from pyspark.sql import SparkSession,Row
import pyspark.sql.functions as F
import pyspark.sql.types as T
import datetime
def fill_dates(d1, d2, start_date, no_of_date_cols):
start_date = datetime.datetime.strptime(start_date, '%Y-%m-%d')
d1 = datetime.datetime.strptime(d1, '%Y-%m-%d')
d2 = datetime.datetime.strptime(d2, '%Y-%m-%d')
cols = {}
for x in range(0, no_of_date_cols):
col = (start_date + datetime.timedelta(days=x)).strftime('%Y-%m-%d')
if datetime.datetime.strptime(col, '%Y-%m-%d') >= d1 and datetime.datetime.strptime(col, '%Y-%m-%d') <= d2:
cols[col] = 1
else:
cols[col] = 0
return cols
spark = SparkSession \
.builder \
.appName("Filling_Dates_Cols") \
.config("master", "local") \
.getOrCreate()
df = spark.createDataFrame([
[1, 'A', '2018-09-26', '2018-09-28'],
[2, 'B', '2018-09-20', '2018-09-22'],
[2, 'B', '2018-09-23', '2018-09-26'],
[3, 'C', '2018-09-15', '2018-09-26']
], schema=['id', 'p', 'd1', 'd2'])
min_max_dates = df.select(
F.min('d1').alias('min'),
F.max('d2').alias('max')
).collect()[0]
min_date = min_max_dates[0]
max_date = min_max_dates[1]
d1 = datetime.datetime.strptime(min_date, '%Y-%m-%d')
d2 = datetime.datetime.strptime(max_date, '%Y-%m-%d')
no_of_date_cols = (d2 - d1).days + 1
schema = T.StructType()
for x in range(0, no_of_date_cols):
new_col = (d1 + datetime.timedelta(days=x)).strftime('%Y-%m-%d')
schema = schema.add(new_col, T.IntegerType())
fill_dates_udf = F.udf(fill_dates, schema)
df = df.withColumn(
'dates',
fill_dates_udf(F.col('d1'), F.col('d2'), F.lit(min_date), F.lit(no_of_date_cols))
)
df.select('id', 'p', 'd1', 'd2', 'dates.*').show()
which results in
+---+---+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+
| id| p| d1| d2|2018-09-15|2018-09-16|2018-09-17|2018-09-18|2018-09-19|2018-09-20|2018-09-21|2018-09-22|2018-09-23|2018-09-24|2018-09-25|2018-09-26|2018-09-27|2018-09-28|
+---+---+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+
| 1| A|2018-09-26|2018-09-28| 0| 0| 0| 0| 0| 0| 0| 0| 0| 0| 0| 1| 1| 1|
| 2| B|2018-09-20|2018-09-22| 0| 0| 0| 0| 0| 1| 1| 1| 0| 0| 0| 0| 0| 0|
| 2| B|2018-09-23|2018-09-26| 0| 0| 0| 0| 0| 0| 0| 0| 1| 1| 1| 1| 0| 0|
| 3| C|2018-09-15|2018-09-26| 1| 1| 1| 1| 1| 1| 1| 1| 1| 1| 1| 1| 0| 0|
+---+---+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+----------+
来源：https://stackoverflow.com/questions/57070594/make-and-populate-a-pyspark-dataframe-with-columns-as-period-range