猪铸件/数据类型 [英] Pig casting / datatypes

问题描述

我正在尝试将关系转储到AVRO文件中，但出现一个奇怪的错误:

I'm trying to dump relation into AVRO file but I'm getting a strange error:

org.apache.pig.data.DataByteArray cannot be cast to java.lang.CharSequence

我不使用DataByteArray(字节数组)，请参见下面的关系描述.

I don't use DataByteArray (bytearray), see description of the relation below.

sensitiveSet: {rank_ID: long,name: chararray,customerId: long,VIN: chararray,birth_date: chararray,fuel_mileage: chararray,fuel_consumption: chararray}

即使我进行显式强制转换，我也会遇到相同的错误:

Even when I do explicit casting I get the same error:

sensitiveSet = foreach sensitiveSet generate (long) $0, (chararray) $1, (long) $2, (chararray) $3, (chararray) $4, (chararray) $5, (chararray) $6;

STORE sensitiveSet INTO 'testOut2222.avro'
USING org.apache.pig.piggybank.storage.avro.AvroStorage('no_schema_check', 'schema', '{"type":"record","name":"xxxx","namespace":"","fields":[{"name":"rank_ID","type":"long"},{"name":"name","type":"string","store":"no","sensitive":"na"},{"name":"customerId","type":"string","store":"yes","sensitive":"yes"},{"name":"VIN","type":"string","store":"yes","sensitive":"yes"},{"name":"birth_date","type":"string","store":"yes","sensitive":"no"},{"name":"fuel_mileage","type":"string","store":"yes","sensitive":"no"},{"name":"fuel_consumption","type":"string","store":"yes","sensitive":"no"}]}');

已

我试图定义一个输出模式，该模式应该是一个包含另外两个元组的元组，即stats:tuple(c:tuple(),d:tuple).

I'm trying to define an output schema which should be a Tuple that contains another two tuples, i.e. stats:tuple(c:tuple(),d:tuple).

下面的代码无法正常工作.它以某种方式产生结构为:

The code below doesn't work as it was intended. It somehow produces structure as:

stats:tuple(b:tuple(c:tuple(),d:tuple()))

下面是describe产生的输出.

sourceData: {com.mortardata.pig.dataspliter_36: (stats: ((name: chararray,customerId: chararray,VIN: chararray,birth_date: chararray,fuel_mileage: chararray,fuel_consumption: chararray),(name: chararray,customerId: chararray,VIN: chararray,birth_date: chararray,fuel_mileage: chararray,fuel_consumption: chararray)))}

是否可以按如下所示创建结构，这意味着我需要从上一个示例中删除元组b.

Is it possible to create structure as below, which means I need to remove the tuple b from the previous example.

grunt> describe sourceData;
sourceData: {t: (s: (name: chararray,customerId: chararray,VIN: chararray,birth_date: chararray,fuel_mileage: chararray,fuel_consumption: chararray),n: (name: chararray,customerId: chararray,VIN: chararray,birth_date: chararray,fuel_mileage: chararray,fuel_consumption: chararray))}

以下代码无法正常工作.

The below code doesn't work as expected.

  public Schema outputSchema(Schema input) {
    Schema sensTuple = new Schema();
    sensTuple.add(new Schema.FieldSchema("name", DataType.CHARARRAY));
    sensTuple.add(new Schema.FieldSchema("customerId", DataType.CHARARRAY));
    sensTuple.add(new Schema.FieldSchema("VIN", DataType.CHARARRAY));
    sensTuple.add(new Schema.FieldSchema("birth_date", DataType.CHARARRAY));
    sensTuple.add(new Schema.FieldSchema("fuel_mileage", DataType.CHARARRAY));
    sensTuple.add(new Schema.FieldSchema("fuel_consumption", DataType.CHARARRAY));

    Schema nonSensTuple = new Schema();
    nonSensTuple.add(new Schema.FieldSchema("name", DataType.CHARARRAY));
    nonSensTuple.add(new Schema.FieldSchema("customerId", DataType.CHARARRAY));
    nonSensTuple.add(new Schema.FieldSchema("VIN", DataType.CHARARRAY));
    nonSensTuple.add(new Schema.FieldSchema("birth_date", DataType.CHARARRAY));
    nonSensTuple.add(new Schema.FieldSchema("fuel_mileage", DataType.CHARARRAY));
    nonSensTuple.add(new Schema.FieldSchema("fuel_consumption", DataType.CHARARRAY));


    Schema parentTuple = new Schema();
    parentTuple.add(new Schema.FieldSchema(null, sensTuple, DataType.TUPLE));
    parentTuple.add(new Schema.FieldSchema(null, nonSensTuple, DataType.TUPLE));


    Schema outputSchema = new Schema();
    outputSchema.add(new Schema.FieldSchema("stats", parentTuple, DataType.TUPLE));

    return new Schema(new Schema.FieldSchema(getSchemaName(this.getClass().getName().toLowerCase(), input), outputSchema, DataType.TUPLE));

UDF的exec方法返回:

The UDF's exec method returns:

public Tuple exec(Tuple tuple) throws IOException {    
  Tuple parentTuple  = mTupleFactory.newTuple();
  parentTuple.append(tuple1);
  parentTuple.append(tuple2);

EDIT2(固定)

...
Schema outputSchema = new Schema();
outputSchema.add(new Schema.FieldSchema("stats", parentTuple, DataType.TUPLE));

返回新的Schema(新的Schema.FieldSchema(getSchemaName(this.getClass().getName().toLowerCase()，输入)，outputSchema，DataType.TUPLE));

return new Schema (new Schema.FieldSchema (getSchemaName (this.getClass ().getName ().toLowerCase (), input), outputSchema, DataType.TUPLE);

return outputSchema;

现在，我从UDF返回正确的架构，其中所有项目均为chararray，但是当我尝试将这些项目存储为类型为string的avro文件时，出现了相同的错误:

Now I return proper schema from UDF where all items are chararray but when I try to store those items into avro file as type: string I got the same error:

java.lang.Exception: org.apache.avro.file.DataFileWriter$AppendWriteException: java.lang.ClassCastException: org.apache.pig.data.DataByteArray cannot be cast to java.lang.CharSequence
        at org.apache.hadoop.mapred.LocalJobRunner$Job.runTasks(LocalJobRunner.java:462)
        at org.apache.hadoop.mapred.LocalJobRunner$Job.run(LocalJobRunner.java:522)

已解决: 好的，问题是数据没有被转换为UDF主体内的正确类型-exec()方法.看起来现在可以工作了！

SOLVED: Ok, the issue was that data wasnt casted to the proper type inside the UDF body - exec () method. Looks like it works now!

推荐答案

通常，这意味着您使用的UDF不能保留架构，或者会丢失.我相信当不知道实型时，DataByteArray是回退类型.您可能需要添加种姓来解决此问题，但是更好的解决方案是修复所有UDF删除架构的问题.

Usually this means you are using a UDF that isn't preserving the schema, or somewhere it is getting lost. I believe DataByteArray is the fallback type when the real type isn't known. You may need to add a caste to workaround this, however a better solution is to fix whatever UDF is dropping the schema.

这篇关于猪铸件/数据类型的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！