如何将多个列转换为一个单元格中的键值列表 - Pyspark？

Question

我有一个 spark 数据框，如下所示：

+-------------+-------+--------+------+-----------+----+----+----+-------------------------+----+
|uniqueKey    |channel|division|gender|category   |W1  |W3  |W4  |W2                       |W5  |
+-------------+-------+--------+------+-----------+----+----+----+-------------------------+----+
|key1         |Digital|APPAREL |KIDS  |COLLECTIONS|null|null|null|[7000, 181]              |null|
|key2         |Digital|APPAREL |KIDS  |CRICKET    |null|null|null|[10000.3, 699]           |null|
|key3         |Digital|FOOTWEAR|MENS  |COLLECTIONS|null|null|null|[4567, 34]               |null|
+-------------+-------+--------+------+-----------+----+----+----+-------------------------+----+

我需要创建一个如下所示的 Json：

{   
  "uniqueKey": "key1"
  "division": "APPAREL"
  "gender": "KIDS"
  "category": "CRICKET"
  "channel": "DIGITAL"
  "dataRows": [ 
     {    
     "rowId": "Net Sales"
     "dataRow": {
      "W1": 0
      "W2": 10000
      "W3": 0
      "W4": 0
      "W5": 0
    }
  }
  {    
     "rowId": "Sales Units"
     "dataRow": {
      "W1": 0
      "W2": 699
      "W3": 0
      "W4": 0
      "W5": 0
      }
  }
 ]

我试过 pivot，但我不确定我做的是否正确。你能帮我解决这个问题吗？

Answer 1

添加了内嵌注释，检查下面的代码。

对于Week Columns，这里我用的是range(1,5) 你可以增加到52

weekCols = F.struct(map(lambda c: F.coalesce(F.col(c).cast("int"),F.lit(0)),["W"+str(x) for x in range(1,5)]))

df.withColumn("W2",F.explode(F.col("W2")))\
.groupBy(F.col("uniqueKey"),F.col("channel"),F.col("division"),F.col("gender"),F.col("category"))\
.agg(F.collect_list(weekCols.alias("dataRow")).alias("dataRows"))\
.withColumn("dataRows",F.expr("transform(dataRows,(v,i) -> if(i=0,struct('Net Sales' as rowId,v as dataRow),struct('Sales Units' as rowId,v as dataRow)))"))\
.select(F.collect_list(F.to_json(F.struct("*"))).alias("data"))\
.show()

最终输出

+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|data                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                 |
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|[{"uniqueKey":"key1","channel":"Digital","division":"APPAREL","gender":"KIDS","category":"COLLECTIONS","dataRows":[{"rowId":"Net Sales","dataRow":{"col1":0,"col2":7000,"col3":0,"col4":0}},{"rowId":"Sales Units","dataRow":{"col1":0,"col2":181,"col3":0,"col4":0}}]}, {"uniqueKey":"key3","channel":"Digital","division":"FOOTWEAR","gender":"MENS","category":"COLLECTIONS","dataRows":[{"rowId":"Net Sales","dataRow":{"col1":0,"col2":4567,"col3":0,"col4":0}},{"rowId":"Sales Units","dataRow":{"col1":0,"col2":34,"col3":0,"col4":0}}]}, {"uniqueKey":"key2","channel":"Digital","division":"APPAREL","gender":"KIDS","category":"CRICKET","dataRows":[{"rowId":"Net Sales","dataRow":{"col1":0,"col2":10000,"col3":0,"col4":0}},{"rowId":"Sales Units","dataRow":{"col1":0,"col2":699,"col3":0,"col4":0}}]}]|
+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+

格式化Json输出

[
  {
    "uniqueKey": "key1",
    "channel": "Digital",
    "division": "APPAREL",
    "gender": "KIDS",
    "category": "COLLECTIONS",
    "dataRows": [
      {
        "rowId": "Net Sales",
        "dataRow": {
          "col1": 0,
          "col2": 7000,
          "col3": 0,
          "col4": 0
        }
      },
      {
        "rowId": "Sales Units",
        "dataRow": {
          "col1": 0,
          "col2": 181,
          "col3": 0,
          "col4": 0
        }
      }
    ]
  },
  {
    "uniqueKey": "key3",
    "channel": "Digital",
    "division": "FOOTWEAR",
    "gender": "MENS",
    "category": "COLLECTIONS",
    "dataRows": [
      {
        "rowId": "Net Sales",
        "dataRow": {
          "col1": 0,
          "col2": 4567,
          "col3": 0,
          "col4": 0
        }
      },
      {
        "rowId": "Sales Units",
        "dataRow": {
          "col1": 0,
          "col2": 34,
          "col3": 0,
          "col4": 0
        }
      }
    ]
  },
  {
    "uniqueKey": "key2",
    "channel": "Digital",
    "division": "APPAREL",
    "gender": "KIDS",
    "category": "CRICKET",
    "dataRows": [
      {
        "rowId": "Net Sales",
        "dataRow": {
          "col1": 0,
          "col2": 10000,
          "col3": 0,
          "col4": 0
        }
      },
      {
        "rowId": "Sales Units",
        "dataRow": {
          "col1": 0,
          "col2": 699,
          "col3": 0,
          "col4": 0
        }
      }
    ]
  }
]

如何将多个列转换为一个单元格中的键值列表 - Pyspark？

How to convert multiple columns to a list of key values in one cell - Pyspark?

apache-spark

apache-spark-sql

pyspark

pyspark-dataframes