PySpark Dataframes:带条件的完全外部连接

PySpark Dataframes: Full Outer Join with a condition

我有以下2个数据框-

dataframe_a
+----------------+---------------+
|         user_id|         domain|
+----------------+---------------+
|            josh|     wanadoo.fr|
|        samantha|     randomn.fr|
|             bob|    eidsiva.net|
|           dylan|    vodafone.it|
+----------------+---------------+

dataframe_b
+----------------+---------------+
|         user_id|         domain|
+----------------+---------------+
|            josh|  oldwebsite.fr|
|        samantha|     randomn.fr|
|           dylan|      oldweb.it|
|            ryan|      chicks.it|
+----------------+---------------+

我想进行完全外部联接,但保留 dataframe_adomain 列中的值,以防我为单个 user_id 获得 2 个不同的域。所以,我想要的数据框看起来像-

desired_df
+----------------+---------------+
|         user_id|         domain|
+----------------+---------------+
|            josh|     wanadoo.fr|
|        samantha|     randomn.fr|
|             bob|    eidsiva.net|
|           dylan|    vodafone.it|
|            ryan|      chicks.it|
+----------------+---------------+

我想我可以做一些像-

desired_df = dataframe_a.join(dataframe_b, ["user_id"], how="full_outer").drop(dataframe_b.domain)

但我担心这是否会在我想要的数据框中给我 ryan。这是正确的方法吗?

不,执行 full_outer 连接将留下所需的数据框,其中与 ryan 对应的域名为 null value.No 对上述给定数据框的连接操作类型将为您提供所需的输出.

您将要使用 'coalesce'。在您当前的解决方案中,ryan 将在生成的数据框中,但剩余的 dataframe_a.domain 列为空值。

joined_df = dataframe_a.join(dataframe_b, ["user_id"], how="full_outer")
+----------------+---------------+---------------+
|         user_id|         domain|         domain|
+----------------+---------------+---------------+
|            josh|     wanadoo.fr|  oldwebsite.fr|
|        samantha|     randomn.fr|     randomn.fr|
|             bob|    eidsiva.net|               |
|           dylan|    vodafone.it|      oldweb.it|
|            ryan|               |      chicks.it|
+----------------+---------------+---------------+

'coalesce' 允许您指定优先顺序,但会跳过空值。

import pyspark.sql.functions as F
joined_df = joined_df.withColumn(
  "preferred_domain",
  F.coalesce(dataframe_a.domain, dataframe_b.domain)
)
joined_df = joined_df.drop(dataframe_a.domain).drop(dataframe_b.domain)

给予

+----------------+----------------+
|         user_id|preferred_domain|
+----------------+----------------+
|            josh|      wanadoo.fr|
|        samantha|      randomn.fr|
|             bob|     eidsiva.net|
|           dylan|     vodafone.it|
|            ryan|       chicks.it|
+----------------+----------------+