当前位置：首页 > article >正文

java spark解决文件读取乱码问题

article 2024/10/28 17:54:41

一、问题

环境为jdk1.8，spark3.2.1，读取hadoop中GB18030编码格式的文件出现乱码。

二、心酸历程

为了解决该问题，尝试过很多种方法，但都没有成功

1、textFile+Configuration方式——乱码

        String filePath = "hdfs:///user/test.deflate";
        //创建SparkSession和SparkContext的实例
        String encoding = "GB18030";

        SparkSession spark = SparkSession.builder()
                .master("local[*]").appName("Spark Example")
                .getOrCreate();

        JavaSparkContext sc = JavaSparkContext.fromSparkContext(spark.sparkContext());
        Configuration entries = sc.hadoopConfiguration();
        entries.set("textinputformat.record.delimiter", "\n");
        entries.set("mapreduce.input.fileinputformat.inputdir",filePath);entries.set("mapreduce.input.fileinputformat.encoding", "GB18030");
        
        JavaRDD<String> rdd = sc.textFile(filePath);

2、spark.read().option方式——乱码

        Dataset<Row> load = spark.read().format("text").option("encoding", "GB18030").load(filePath);

        load.foreach(row -> {
            System.out.println(row.toString());
            System.out.println(new String(row.toString().getBytes(encoding),"UTF-8"));
            System.out.println(new String(row.toString().getBytes(encoding),"GBK"));


        });

3、newAPIHadoopFile+Configuration——乱码

        JavaPairRDD<LongWritable, Text> longWritableTextJavaPairRDD = sc.newAPIHadoopFile(filePath, TextInputFormat.class, LongWritable.class, Text.class, entries );

        System.out.println("longWritableTextJavaPairRDD  count ="+longWritableTextJavaPairRDD.count());
        longWritableTextJavaPairRDD.foreach(k->{

            System.out.println(k._2);
        });

4、newAPIHadoopFile+自定义类——乱码

        JavaPairRDD<LongWritable, Text> longWritableTextJavaPairRDD = sc.newAPIHadoopFile(filePath, GBKInputFormat.class, LongWritable.class, Text.class, entries );

        System.out.println("longWritableTextJavaPairRDD  count ="+longWritableTextJavaPairRDD.count());
        longWritableTextJavaPairRDD.foreach(k->{

            System.out.println(k._2);
        });

代码中GBKInputFormat.class是TextInputFormat.class复制将内部UTF-8修改为GB18030所得

5、newAPIHadoopRDD+自定义类——乱码

        JavaPairRDD<LongWritable, Text> longWritableTextJavaPairRDD1 = sc.newAPIHadoopRDD(entries, GBKInputFormat.class, LongWritable.class, Text.class);
        System.out.println("longWritableTextJavaPairRDD  count ="+longWritableTextJavaPairRDD1.count());
        longWritableTextJavaPairRDD1.foreach(k->{
            System.out.println(k._2());
        });

3、最终解决

上述方法感觉指定的字符编码并没有生效不知道为什么，如有了解原因的还请为我解惑，谢谢
最终解决方案如下

       JavaPairRDD<LongWritable, Text> longWritableTextJavaPairRDD = sc.newAPIHadoopFile(filePath, TextInputFormat.class, LongWritable.class, Text.class, new Configuration());

        System.out.println("longWritableTextJavaPairRDD  count ="+longWritableTextJavaPairRDD.count());
        longWritableTextJavaPairRDD.foreach(k->{
            System.out.println(new String(k._2.copyBytes(), encoding));
        });

        JavaPairRDD<LongWritable, Text> longWritableTextJavaPairRDD1 = sc.newAPIHadoopRDD(entries, TextInputFormat.class, LongWritable.class, Text.class);

        System.out.println("longWritableTextJavaPairRDD  count ="+longWritableTextJavaPairRDD1.count());
        longWritableTextJavaPairRDD1.foreach(k->{
            System.out.println(new String(k._2().copyBytes(),encoding));
            System.out.println(new String(k._2.copyBytes(),encoding));
        });