hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Naama Kraus" <naamakr...@gmail.com>
Subject Re: Map Reduce over HBase - sample code
Date Thu, 26 Jun 2008 05:27:09 GMT
Here is an updated code

Naama

/**
 * <pre>
 * 'Toy tables' for experiencing with MapReduce over HBase
 *
 * grades table - a HBase table of the form -
 * raw id is a student name
 * column name is Course:course_name
 * cell value is the student's grade in the course 'course_name'
 *
 * Exmaple:
 *
 *         Course:Math  |  Course:Art  |  Course:Sports
 *         ----------------------------------------------------------------
 * Dan          87                97              99
 * Dana       100               100             80
 *
 * =======================================
 *
 * courses table - a HBase table of the form -
 * raw id is a course name
 * column name is Stats:Average
 * cell value is the average grade in that course, computed by a map reduce
job
 *
 * Exmaple:
 *
 *               Stats:Average
 *               --------------
 *  Art             86
 *  Match        77
 * </pre>
 * @see GradesTableMapReduce
 *
 *
 */
public class GradesTable {
  // Table operation type
  enum OP { CREATE, DELETE, DUMP };

  public static final String GRADES_TABLE_NAME = "grades";
  public static final String COURSE_TABLE_NAME = "courses";
  public static final String COURSE_FAMILY = "Course:";
  // A column family holding grades statistics
  public static final String STATS_FAMILY = "Stats:";
  // A column member holding average grade in course
  public static final String AVG = "Average";

  private static final String [] STUDENT_NAMES = {
    "Dan", "Dana", "Sara", "David"
  };

  private static final String [] COURSE_NAMES = {
    "Math", "Art", "Sports"
  };

  private HBaseConfiguration conf;
  private HBaseAdmin admin;
  private HTableDescriptor grades_desc;
  private HTableDescriptor courses_desc;
  // Randomly generate a grade
  private Random rand;

  public GradesTable() throws IOException {
    conf = new HBaseConfiguration();
    admin = new HBaseAdmin(conf);
    grades_desc = new HTableDescriptor(GRADES_TABLE_NAME);
    courses_desc = new HTableDescriptor(COURSE_TABLE_NAME);
    rand = new Random();
  }

  /**
   * Create tables and populate with content
   */
  public void create() throws IOException {
    grades_desc.addFamily(new HColumnDescriptor(COURSE_FAMILY));
    courses_desc.addFamily(new HColumnDescriptor(STATS_FAMILY));
    admin.createTable(grades_desc);
    admin.createTable(courses_desc);
    System.out.println("Tables created");

    // Populate grades table with students and their grades in courses
    HTable table = new HTable(conf, new Text(GRADES_TABLE_NAME));

    // Start an update transaction, student name is row id
    for (int i = 0; i < STUDENT_NAMES.length; i++) {
      System.out.println("<<< Row " + i + ", student: " + STUDENT_NAMES[i] +
" >>>");
      Text stuName = new Text(STUDENT_NAMES[i]);
      long writeid = table.startUpdate(stuName);
      for (int j = 0; j < COURSE_NAMES.length; j++) {
        Text courseColumn = new Text(COURSE_FAMILY + COURSE_NAMES[j]);
        // Put a cell with a student's grade in this course
        int grade = Math.abs(rand.nextInt()) % 101;
        table.put(writeid, courseColumn, new IntWritable(grade));
        System.out.println("Course: " + COURSE_NAMES[j] + ", grade: " +
grade);
      }
      table.commit(writeid);
    }
    System.out.println("Grades Table populated");
  }
}


====================================================

/**
 * A map reduce job over {@link GradesTable}
 * The job produces for each course the average grade in that course.
 * It puts the average in a separate table which holds course statistics.
 *
 */
public class GradesTableMapReduce  extends Configured implements Tool {

  /**
   * Map a row to {key, value} pairs.
   * Emit a {course, grade} pair for each course grade appearing in the
student row.
   * E.g. Sara {Math:62, Art:45, Sports:87} -> {Math, 62}, {Art, 45},
{Sports, 87}
   *
   */
  public static class GradesTableMap extends TableMap<Text, IntWritable> {

    @Override
    public void map(HStoreKey key, MapWritable value,
        OutputCollector<Text, IntWritable> output, Reporter reporter) throws
IOException {

      // Walk through the columns
      for (Map.Entry<Writable, Writable> e: value.entrySet()) {
        // Column name is course name
        Text course = (Text) e.getKey();
        // Remove the family prefix
        String courseStr = course.toString();
        courseStr =
          courseStr.substring(courseStr.indexOf(':') + 1);
        course = new Text(courseStr);
        byte [] gradeInBytes = ((ImmutableBytesWritable)
e.getValue()).get();
        DataInputStream in = new DataInputStream(new
ByteArrayInputStream(gradeInBytes));
        IntWritable grade = new IntWritable();
        grade.readFields(in);
        // Emit course name and a grade
        output.collect(course, grade);
      }
    }
  }

  /**
   * Reduce - compute an average of key's values which is actually the
average grade in each course.
   * E.g. {Math, {62, 45, 87}} -> {Math, 65.6}
   *
   */
  public static class GradesTableReduce extends TableReduce<Text,
IntWritable> {

    @Override
    // key is course name, values are grades in the course
    public void reduce(Text key, Iterator<IntWritable> values,
        OutputCollector<Text, MapWritable> output, Reporter reporter)
    throws IOException {
      // Compute grades average
      int total = 0;
      int sum = 0;
      while (values.hasNext()) {
        total++;
        sum += values.next().get();
      }
      float average = sum / total;

      // We put the average as a separate column in the courses table
      ByteArrayOutputStream baos = new ByteArrayOutputStream();
      DataOutputStream out = new DataOutputStream(baos);
      FloatWritable avgWritable = new FloatWritable(average);
      avgWritable.write(out);
      MapWritable map = new MapWritable();
      map.put(new Text(GradesTable.STATS_FAMILY + GradesTable.AVG),
              new ImmutableBytesWritable(baos.toByteArray()));
      output.collect(key, map);
    }
  }

  /**
   * Run
   */
  public int run(String[] args) throws Exception {
    JobConf jobConf = new JobConf();
    jobConf.setJobName("compute average grades");
    jobConf.setNumReduceTasks(1);

    // All columns in the course family (i.e. all grades) get into the map
    TableMap.initJob(GradesTable.GRADES_TABLE_NAME,
GradesTable.COURSE_FAMILY,
        GradesTableMap.class, jobConf);

    // Reduce output (course average grade) is put in the courses table
    TableReduce.initJob(GradesTable.COURSE_TABLE_NAME,
        GradesTableReduce.class, jobConf);

    // Map produces a value which is an IntWritable
    jobConf.setMapOutputValueClass(IntWritable.class);

    JobClient.runJob(jobConf);
    return 0;
  }

  public static void main(String [] args) throws Exception {
    ToolRunner.run(new Configuration(), new GradesTableMapReduce(), args);
  }
}

On Tue, Jun 24, 2008 at 5:57 PM, stack <stack@duboce.net> wrote:

> Naama Kraus wrote:
>
>> ..
>> What if the mission was the following - for each course in the table,
>> calculate the average grade in that course. In that case both map and
>> reduce
>> are required, is that correct ? Map will emit for each row a {course_name,
>> grade} pair. Reduce will emit the average grades for each course
>> (course_name, avg_grade}. Output can be put in a separate table (probably
>> one holding courses information). Does this make sense ?
>>
>>
>>
>>
> That'll work.
>
>  * At a higher level, I'd suggest a refactoring.  Do all of your work in
>>> the map phase.  Have no reduce phase.  I suggest this because as is, all
>>> rows emitted by the map are being sorted by the MR framework.  But hbase
>>> will also do a sort on insert.   Avoid paying the prices of the MR sort.
>>>  Do
>>> your calculation in the map and then insert the result at map time.
>>> Either
>>> emit nothing or, emit a '1' for every row processed so the MR counters
>>> tell
>>> a story about your MR job.*
>>>
>>>
>>>
>>
>> That's an interesting point. So if both map and reduce are a required,
>> then
>> two sorts must take place. Is that correct ?
>>
>>
> Yes but with your new example, they are orthogonal toward different ends;
> the first does collecting together all course data and the second orders
> courses in hbase lexicographically (presuming course is primary key).
>
> St.Ack
>



-- 
oo 00 oo 00 oo 00 oo 00 oo 00 oo 00 oo 00 oo 00 oo 00 oo 00 oo 00 oo 00 oo
00 oo 00 oo
"If you want your children to be intelligent, read them fairy tales. If you
want them to be more intelligent, read them more fairy tales." (Albert
Einstein)

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message