首先，你可以创建两个 Dataframe ，一个有空值，另一个没有空值，然后在空值的 Dataframe 上，你可以使用apache spark中的randomSplit函数将其拆分为2个 Dataframe ，使用你指定的比率，最后你可以将3个 Dataframe 合并以获得想要的结果：

spark = SparkSession.builder.appName("DateDataFrame").getOrCreate()
data = [
    ("a1", "Engalnd", None),
    ("a2", "Engalnd", None),
    ("b1", "Engalnd", None),
    ("b2", "Engalnd", None),
    ("c1", "Engalnd", None),
    ("c2", "Engalnd", None),
    ("er3", "Engalnd", None),
    ("po9", "Engalnd", None),
    ("ee4", "Engalnd", None),
    ("e4", "Engalnd", None),
    ("t5", "Engalnd", None),
    ("u8", "Engalnd", None),
    ("r4", "Engalnd", None),
    ("zx1", "Engalnd", None),
    ("11d", "Engalnd", None),
    ("22", "Engalnd", None),
    ("2p", "Engalnd", None),
    ("3jk", "Engalnd", None),
    ("56h", "Engalnd", None),
    ("a78", "Engalnd", None),
    ("xxx", "Engalnd", "value1"),
    ("zzz", "Engalnd", "value2"),
]
df = spark.createDataFrame(data, ['Id_a', 'Country', 'Type'])

missingTypeDf = df.filter(col("Type").isNull())
notMissingTypeDf = df.filter(~col("Type").isNull())

from pyspark.sql.functions import rand
fractions = [0.8, 0.2]

# Split the DataFrame
chunkDf1, chunkDf2 = missingTypeDf.randomSplit(fractions, seed = 13)

chunkDf1 = chunkDf1.withColumn("Type", lit("R"))
chunkDf2 = chunkDf2.withColumn("Type", lit("NR"))

resultDf = notMissingTypeDf.unionByName(chunkDf1).unionByName(chunkDf2)

resultDf.show(200, truncate=False)

+----+-------+------+
|Id_a|Country|Type  |
+----+-------+------+
|xxx |Engalnd|value1|
|zzz |Engalnd|value2|
|a1  |Engalnd|R     |
|a2  |Engalnd|R     |
|b1  |Engalnd|R     |
|b2  |Engalnd|R     |
|c1  |Engalnd|R     |
|er3 |Engalnd|R     |
|po9 |Engalnd|R     |
|ee4 |Engalnd|R     |
|e4  |Engalnd|R     |
|t5  |Engalnd|R     |
|u8  |Engalnd|R     |
|r4  |Engalnd|R     |
|zx1 |Engalnd|R     |
|22  |Engalnd|R     |
|2p  |Engalnd|R     |
|3jk |Engalnd|R     |
|c2  |Engalnd|NR    |
|11d |Engalnd|NR    |
|56h |Engalnd|NR    |
|a78 |Engalnd|NR    |
+----+-------+------+