使用教程

数据准备

在表格存储中准备一张数据表pet，name是唯一的一列主键，数据示例如下。

Hive访问示例

前提条件

按照准备工作准备好Hadoop、Hive、JDK环境以及表格存储Java SDK和EMR SDK依赖包。

示例

# HADOOP_HOME及HADOOP_CLASSPATH可以添加到/etc/profile中。
$ export HADOOP_HOME=${您的Hadoop安装目录}
$ export HADOOP_CLASSPATH=emr-tablestore-1.4.2.jar:tablestore-4.3.1-jar-with-dependencies.jar:joda-time-2.9.4.jar
$ bin/hive
hive> CREATE EXTERNAL TABLE pet
  (name STRING, owner STRING, species STRING, sex STRING, birth STRING, death STRING)
  STORED BY 'com.aliyun.openservices.tablestore.hive.TableStoreStorageHandler'
  WITH SERDEPROPERTIES(
    "tablestore.columns.mapping"="name,owner,species,sex,birth,death")
  TBLPROPERTIES (
    "tablestore.endpoint"="YourEndpoint",
    "tablestore.access_key_id"="YourAccessKeyId",
    "tablestore.access_key_secret"="YourAccessKeySecret",
    "tablestore.table.name"="pet");
hive> SELECT * FROM pet;
Bowser  Diane   dog     m       1979-08-31      1995-07-29
Buffy   Harold  dog     f       1989-05-13      NULL
Chirpy  Gwen    bird    f       1998-09-11      NULL
Claws   Gwen    cat     m       1994-03-17      NULL
Fang    Benny   dog     m       1990-08-27      NULL
Fluffy  Harold  cat     f       1993-02-04      NULL
Puffball        Diane   hamster f       1999-03-30      NULL
Slim    Benny   snake   m       1996-04-29      NULL
Whistler        Gwen    bird    NULL    1997-12-09      NULL
Time taken: 5.045 seconds, Fetched 9 row(s)
hive> SELECT * FROM pet WHERE birth > "1995-01-01";
Chirpy  Gwen    bird    f       1998-09-11      NULL
Puffball        Diane   hamster f       1999-03-30      NULL
Slim    Benny   snake   m       1996-04-29      NULL
Whistler        Gwen    bird    NULL    1997-12-09      NULL
Time taken: 1.41 seconds, Fetched 4 row(s)
            

HadoopMR访问示例

以下示例介绍如何使用HadoopMR程序统计数据表pet的行数。

示例

• 构建Mappers和Reducers

  public class RowCounter {
  public static class RowCounterMapper
  extends Mapper<PrimaryKeyWritable, RowWritable, Text, LongWritable> {
      private final static Text agg = new Text("TOTAL");
      private final static LongWritable one = new LongWritable(1);
  
      @Override
      public void map(
          PrimaryKeyWritable key, RowWritable value, Context context)
          throws IOException, InterruptedException {
          context.write(agg, one);
      }
  }
  
  public static class IntSumReducer
  extends Reducer<Text,LongWritable,Text,LongWritable> {
  
      @Override
      public void reduce(
          Text key, Iterable<LongWritable> values, Context context)
          throws IOException, InterruptedException {
          long sum = 0;
          for (LongWritable val : values) {
              sum += val.get();
          }
          context.write(key, new LongWritable(sum));
      }
  }
  }
                      

  数据源每从表格存储上读出一行，都会调用一次mapper的map()。前两个参数PrimaryKeyWritable和RowWritable分别对应这行的主键以及这行的内容。可以通过调用PrimaryKeyWritable.getPrimaryKey()和RowWritable.getRow()取得表格存储Java SDK定义的主键对象及行对象。

• 配置表格存储作为mapper的数据源。

  private static RangeRowQueryCriteria fetchCriteria() {
      RangeRowQueryCriteria res = new     RangeRowQueryCriteria("YourTableName");
      res.setMaxVersions(1);
      List<PrimaryKeyColumn> lower = new ArrayList<PrimaryKeyColumn>();
      List<PrimaryKeyColumn> upper = new ArrayList<PrimaryKeyColumn>();
      lower.add(new PrimaryKeyColumn("YourPkeyName", PrimaryKeyValue.INF_MIN));
      upper.add(new PrimaryKeyColumn("YourPkeyName", PrimaryKeyValue.INF_MAX));
      res.setInclusiveStartPrimaryKey(new PrimaryKey(lower));
      res.setExclusiveEndPrimaryKey(new PrimaryKey(upper));
      return res;
  }
  
  public static void main(String[] args) throws Exception {
      Configuration conf = new Configuration();
      Job job = Job.getInstance(conf, "row count");
      job.addFileToClassPath(new Path("hadoop-connector.jar"));
      job.setJarByClass(RowCounter.class);
      job.setMapperClass(RowCounterMapper.class);
      job.setCombinerClass(IntSumReducer.class);
      job.setReducerClass(IntSumReducer.class);
      job.setOutputKeyClass(Text.class);
      job.setOutputValueClass(LongWritable.class);
      job.setInputFormatClass(TableStoreInputFormat.class);
      TableStoreInputFormat.setEndpoint(job, "https://YourInstance.Region.ots.aliyuncs.com/");
      TableStoreInputFormat.setCredential(job, "YourAccessKeyId", "YourAccessKeySecret");
      TableStoreInputFormat.addCriteria(job, fetchCriteria());
      FileOutputFormat.setOutputPath(job, new Path("output"));
      System.exit(job.waitForCompletion(true) ? 0 : 1);
  }                    

  示例中使用job.setInputFormatClass(TableStoreInputFormat.class)把表格存储设为数据源，除此之外，还需要：

  • 把hadoop-connector.jar部署到集群上并添加到classpath里面。路径为addFileToClassPath()指定hadoop-connector.jar的本地路径。代码中假定hadoop-connector.jar在当前路径。
  • 访问表格存储需要指定入口和身份。通过TableStoreInputFormat.setEndpoint()和TableStoreInputFormat.setCredential()设置访问表格存储需要指定的Endpoint和Access Key信息。

  • 指定一张表用来计数。

    说明

    • 每调用一次addCriteria()可以在数据源里添加一个Java SDK定义的RangeRowQueryCriteria对象。可以多次调用addCriteria()。RangeRowQueryCriteria对象与表格存储Java SDK GetRange接口所用的RangeRowQueryCriteria对象具有相同的限制条件。
    • 使用RangeRowQueryCriteria的setFilter()和addColumnsToGet()可以在表格存储的服务器端过滤掉不必要的行和列，减少访问数据的大小，降低成本，提高性能。
    • 通过添加对应多张表的多个RangeRowQueryCriteria，可以实现多表的union。
    • 通过添加同一张表的多个RangeRowQueryCriteria，可以做到更均匀的切分。TableStore-Hadoop Connector会根据一些策略将用户传入的范围切细。

程序运行示例

$ HADOOP_CLASSPATH=hadoop-connector.jar bin/hadoop jar row-counter.jar
...
$ find output -type f
output/_SUCCESS
output/part-r-00000
output/._SUCCESS.crc
output/.part-r-00000.crc
$ cat out/part-r-00000
TOTAL   9           

类型转换说明

表格存储支持的数据类型和Hive/Spark支持的数据类型不完全相同。

下表列出了从表格存储的数据类型（行）转换到Hive/Spark数据类型（列）的支持情况。