Spark DataFrame join后移除重复的列

书香恋仁心2023-11-16 16:48

在Spark，两个DataFrame做join操作后，会出现重复的列。例如：

java 复制代码

 Dataset<Row> moviesWithRating = moviesDF
                .join(averageRatingMoviesDF,
                        moviesDF.col("movieId").equalTo(averageRatingMoviesDF.col("movieId")));

其schema如下：

java 复制代码

//moviesWithRating.printSchema();
        /**
         * root
         *  |-- _id: struct (nullable = true)
         *  |    |-- oid: string (nullable = true)
         *  |-- actors: string (nullable = true)
         *  |-- description: string (nullable = true)
         *  |-- directors: string (nullable = true)
         *  |-- genres: string (nullable = true)
         *  |-- issue: string (nullable = true)
         *  |-- language: string (nullable = true)
         *  |-- movieId: integer (nullable = true)
         *  |-- shoot: string (nullable = true)
         *  |-- timeLong: string (nullable = true)
         *  |-- title: string (nullable = true)
         *  |-- movieId: integer (nullable = true)
         *  |-- avgRating: double (nullable = true)
         */

我们在继续操作这个DataFrame时，可能就会报错，如下：org.apache.spark.sql.AnalysisException: Reference 'movieId' is ambiguous

解决方案有两种方法可以用来移除重复的列

方法一：join表达式使用字符串数组（用于join的列）

java 复制代码

Seq<String> joinColumns = JavaConversions.asScalaBuffer(Arrays.asList("movieId", "movieId")).toList();
        Dataset<Row> moviesWithRating = moviesDF
                .join(
                        averageRatingMoviesDF,
                        joinColumns,
                        "inner");

这里DataFrame moviesDF和averageRatingMoviesDF使用了movieId和movieId两列来做join，返回的结果会对这两列去重

scala解决方案：

scala 复制代码

df1.join(df2, Seq("id","name"),"left")  // df1和df2使用了id和name两列来做join,返回的结果会对这两列去

方法二：使用select返回指定的列

java 复制代码

Dataset<Row> moviesWithRating = moviesDF
                .join(averageRatingMoviesDF,
                        moviesDF.col("movieId").equalTo(averageRatingMoviesDF.col("movieId")))
                .select(
                        moviesDF.col("movieId"),

                        col("actors"),
                        col("description"),
                        col("directors"),
                        col("genres"),
                        col("issue"),
                        col("language"),
                        col("shoot"),
                        col("timeLong"),
                        col("title"),
                        col("avgRating")
                );

说明：

如果列较少, 推荐使用第二种.

如果列较多, 推荐使用第一种.