我需要在数据框中添加一个“行号”,但是必须为列中的每个新值重新启动该“行号”。
让我给你看一个例子:
from pyspark.sql import SparkSession
spark = SparkSession.builder.appName('test').getOrCreate()
df = spark.createDataFrame([
('2018-01-01', 'John', 0),
('2018-01-01', 'Paul', 1),
('2018-01-08', 'Paul', 3),
('2018-01-08', 'Pete', 4),
('2018-01-08', 'John', 3),
('2018-01-15', 'Mary', 6),
('2018-01-15', 'Pete', 6),
('2018-01-15', 'John', 6),
('2018-01-15', 'Paul', 6),
], ['str_date', 'name', 'value'])
# Convert str_date to date:
df = df.withColumn('date', to_date(df['str_date'])) \
.select(['date', 'name', 'value'])
# Sort by name and date
df.orderBy(['name', 'date']).show()
## +----------+----+-----+
## | date|name|value|
## +----------+----+-----+
## |2018-01-01|John| 0|
## |2018-01-08|John| 3|
## |2018-01-15|John| 6|
## |2018-01-15|Mary| 6|
## |2018-01-01|Paul| 1|
## |2018-01-08|Paul| 3|
## |2018-01-15|Paul| 6|
## |2018-01-08|Pete| 4|
## |2018-01-15|Pete| 6|
## +----------+----+-----+
因此,我需要添加一个新列,其中包含每个name
的行号:
# Expected result
## +----------+----+-----+------+
## | date|name|value|rowNum|
## +----------+----+-----+------+
## |2018-01-01|John| 0| 1| <- First row for 'John'
## |2018-01-08|John| 3| 2|
## |2018-01-15|John| 6| 3|
## |2018-01-15|Mary| 6| 1| <- First row for 'Mary'
## |2018-01-01|Paul| 1| 1| <- First row for 'Paul'
## |2018-01-08|Paul| 3| 2|
## |2018-01-15|Paul| 6| 3|
## |2018-01-08|Pete| 4| 1| <- First row for 'Pete'
## |2018-01-15|Pete| 6| 2|
## +----------+----+-----+------+
我一直在尝试使用Window
函数,但是遇到了麻烦。你能帮我吗?
注释:
答案 0 :(得分:1)
使用诸如row_number
之类的排名函数来执行此操作。如果在给定日期可以绑定名称,请改用dense_rank
。
from pyspark.sql import Window
from pyspark.sql import functions as f
#Window definition
w = Window.partitionBy(df.name).orderBy(df.date)
res = df.withColumn('rnum',f.row_number().over(w))
res.show()
答案 1 :(得分:0)
Vamsi的答案是正确的。错过了()作为row_number的位置,所以...
w = Window.partitionBy(df.name).orderBy(df.date)
res = df.withColumn('rnum',f.row_number().over(w)) # change after row_number
res.show()