Apache spark 基于列值联接

我使用的是spark-sql-2.4.1v 如何进行各种连接取决于列的值

样本数据

val data = List(
  ("20", "score", "school",  14 ,12),
  ("21", "score", "school",  13 , 13),
  ("22", "rate", "school",  11 ,14)
 )
val df = data.toDF("id", "code", "entity", "value1","value2")

+---+-----+------+------+------+
| id| code|entity|value1|value2|
+---+-----+------+------+------+
| 20|score|school|    14|    12|
| 21|score|school|    13|    13|
| 22| rate|school|    11|    14|
| 21| rate|school|    13|    12|

基于“code”列值，我需要与其他各种表进行连接

val rateDs = // val data1= List(
  ("22", 11 ,A),
  ("22", 14 ,B),
  ("20", 13 ,C),
  ("21", 12 ,C),
  ("21", 13 ,D)
)

val df=data1.toDF（“id”、“映射代码”、“映射值”）

如果“code”列的值是“rate”，我需要加入rateDs 如果“code”列的值是“score”，我需要加入scoreDs

如何在spark中处理这些事情？有什么最佳的方法来实现这一点吗

费率字段的预期结果

+---+-----+------+------+------+
| id| code|entity|value1|value2|
+---+-----+------+------+------+
| 22| rate|school|     A|    B |
| 21| rate|school|     D|    C |

---

例如，您可以简单地加入两次

val data = List(
  ("20", "score", "school",  14 , 12),
  ("21", "score", "school",  13 , 13),
  ("22", "rate", "school",  11 , 14),
  ("21", "rate", "school",  13 , 12)    
 )
val df = data.toDF("id", "code", "entity", "value1","value2")

val data1 = List(
  ("22", 11 ,"A"),
  ("22", 14 ,"B"),
  ("20", 13 ,"C"),
  ("21", 12 ,"C"),
  ("21", 13 ,"D")
)
val rateDF = data1.toDF("id", "map_code","map_val")

df.as("a")
  .join(rateDF.as("b"),
       col("a.code") === lit("rate") 
        && col("a.id") === col("b.id") 
        && col("a.value1") === col("b.map_code"), "inner")
  .join(rateDF.as("c"),
       col("a.code") === lit("rate") 
        && col("a.id") === col("c.id") 
        && col("a.value2") === col("c.map_code"), "inner")
  .select(col("a.id"), col("a.code"), col("a.entity"), col("b.map_val").as("value1"), col("c.map_val").as("value2"))
  .show(false)

+---+----+------+------+------+
|id |code|entity|value1|value2|
+---+----+------+------+------+
|22 |rate|school|A     |B     |
|21 |rate|school|D     |C     |
+---+----+------+------+------+

---

嗯，这看起来有点脏，但我不知道多个列…

您可以筛选出两个数据帧，与其他数据帧连接并合并它们again@koiralo谢谢，可以使用“when”子句吗？谢谢，可以使用“when”子句吗？我想，这种连接会影响性能。当您连接表时，不建议这样做。这里所做的合并（“b.value1”，“c.value1”）？？当

code=rate

时，b.value1将不为空，c.value1将为空，当

code=score

时，它将被反转。因此，coalesce将这两个结果收集为一列，但这取决于您，这只是一个示例。在基于col（“a.id”）===col（“b.id”），“left”）条件进行连接后，我需要循环多个列值，即“rateDs”中“df”的“value_1”和“value_2”…如何处理？

val data = List(
  ("20", "score", "school",  14 , 12),
  ("21", "score", "school",  13 , 13),
  ("22", "rate", "school",  11 , 14),
  ("21", "rate", "school",  13 , 12)    
 )
val df = data.toDF("id", "code", "entity", "value1","value2")

val data1 = List(
  ("22", 11 ,"A"),
  ("22", 14 ,"B"),
  ("20", 13 ,"C"),
  ("21", 12 ,"C"),
  ("21", 13 ,"D")
)
val rateDF = data1.toDF("id", "map_code","map_val")

df.as("a")
  .join(rateDF.as("b"),
       col("a.code") === lit("rate") 
        && col("a.id") === col("b.id") 
        && col("a.value1") === col("b.map_code"), "inner")
  .join(rateDF.as("c"),
       col("a.code") === lit("rate") 
        && col("a.id") === col("c.id") 
        && col("a.value2") === col("c.map_code"), "inner")
  .select(col("a.id"), col("a.code"), col("a.entity"), col("b.map_val").as("value1"), col("c.map_val").as("value2"))
  .show(false)

+---+----+------+------+------+
|id |code|entity|value1|value2|
+---+----+------+------+------+
|22 |rate|school|A     |B     |
|21 |rate|school|D     |C     |
+---+----+------+------+------+