在 pyspark 中删除所有重复实例

Question

我试着搜索这个，但我能找到的最接近的是 this。但它并没有给我我想要的。我想删除数据框中所有重复的实例。例如，如果我有一个数据框

   Col1   Col2   Col3
   Alice  Girl   April
   Jean   Boy    Aug
   Jean   Boy    Sept

我想根据 Col1 和 Col2 删除 all 重复项，以便我得到

  Col1   Col2  Col3
  Alice  Girl  April

有什么办法吗？

此外，如果我有大量这样的列：

   Col1   Col2   Col3  .... Col n
   Alice  Girl   April .... Apple
   Jean   Boy    Aug   .... Orange
   Jean   Boy    Sept  .... Banana

如何仅按 Col1 和 Col2 分组，但仍保留其余列？

谢谢

Answer 1

from pyspark.sql import functions as F
# Sample Dataframe
df = sqlContext.createDataFrame([
    ["Alice", "Girl","April"],
   ["Jean","Boy", "Aug"],
   ["Jean","Boy","Sept"]
], 
    ["Col1","Col2", "Col3"])

# Group by on required column and select rows where count is 1.
df2 = (df
       .groupBy(["col1", "col2"])
       .agg(
           F.count(F.lit(1)).alias('count'), 
           F.max("col3").alias("col3")).where("count = 1")).drop("count")

df2.show(10, False)

输出：

+-----+----+-----+
|col1 |col2|col3 |
+-----+----+-----+
|Alice|Girl|April|
+-----+----+-----+

对编辑版本的回应

df = sqlContext.createDataFrame([
    ["Alice", "Girl","April", "April"],
    ["Jean","Boy", "Aug", "XYZ"],
    ["Jean","Boy","Sept", "IamBatman"]
], 
    ["col1","col2", "col3", "newcol"])

groupingcols = ["col1", "col2"]
othercolumns = [F.max(col).alias(col) for col in df.columns if col not in groupingcols]

df2 = (df
       .groupBy(groupingcols)
       .agg(F.count(F.lit(1)).alias('count'), *othercolumns)
       .where("count = 1")
       .drop("count"))

df2.show(10, False)

输出：

+-----+----+-----+------+
|col1 |col2|col3 |newcol|
+-----+----+-----+------+
|Alice|Girl|April|April |
+-----+----+-----+------+

在 pyspark 中删除所有重复实例

drop all instances of duplicates in pyspark

pyspark

pyspark-sql

对编辑版本的回应