コード例 #1
0
ファイル: SplitTest.java プロジェクト: scriptreiter/myria
  @Test
  public void testGeneratedSplitsSingleColumn() throws DbException {
    final String[][] expectedResults = {
      {"foo:bar:baz", "foo"}, {"foo:bar:baz", "bar"}, {"foo:bar:baz", "baz"}
    };
    final Schema schema = Schema.ofFields("string", Type.STRING_TYPE);
    final Schema expectedResultSchema =
        Schema.appendColumn(schema, Type.STRING_TYPE, "string_splits");
    final TupleBatchBuffer input = new TupleBatchBuffer(schema);
    input.putString(0, "foo:bar:baz");
    Split splitOp = new Split(new TupleSource(input), 0, ":");

    splitOp.open(TestEnvVars.get());
    int rowIdx = 0;
    while (!splitOp.eos()) {
      TupleBatch result = splitOp.nextReady();
      if (result != null) {
        assertEquals(expectedResultSchema, result.getSchema());

        for (int batchIdx = 0; batchIdx < result.numTuples(); ++batchIdx, ++rowIdx) {
          assertEquals(expectedResults[rowIdx][0], result.getString(0, batchIdx));
          assertEquals(expectedResults[rowIdx][1], result.getString(1, batchIdx));
        }
      }
    }
    assertEquals(expectedResults.length, rowIdx);
    splitOp.close();
  }
コード例 #2
0
ファイル: SplitTest.java プロジェクト: scriptreiter/myria
  @Test
  public void testGeneratedSplits() throws DbException {
    final Object[][] expectedResults = {
      {true, "foo:bar:baz", 1L, 0.1, "foo"},
      {true, "foo:bar:baz", 1L, 0.1, "bar"},
      {true, "foo:bar:baz", 1L, 0.1, "baz"},
      {false, ":qux::", 2L, 0.2, ""},
      {false, ":qux::", 2L, 0.2, "qux"},
      {false, ":qux::", 2L, 0.2, ""},
      {false, ":qux::", 2L, 0.2, ""}
    };
    final Schema schema =
        Schema.ofFields(
            "bool",
            Type.BOOLEAN_TYPE,
            "string",
            Type.STRING_TYPE,
            "long",
            Type.LONG_TYPE,
            "double",
            Type.DOUBLE_TYPE);
    final Schema expectedResultSchema =
        Schema.appendColumn(schema, Type.STRING_TYPE, "string_splits");
    final TupleBatchBuffer input = new TupleBatchBuffer(schema);
    // First row to explode
    input.putBoolean(0, true);
    input.putString(1, "foo:bar:baz");
    input.putLong(2, 1L);
    input.putDouble(3, 0.1);
    // Second row to explode
    input.putBoolean(0, false);
    input.putString(1, ":qux::");
    input.putLong(2, 2L);
    input.putDouble(3, 0.2);
    Split splitOp = new Split(new TupleSource(input), 1, ":");

    splitOp.open(TestEnvVars.get());
    int rowIdx = 0;
    while (!splitOp.eos()) {
      TupleBatch result = splitOp.nextReady();
      if (result != null) {
        assertEquals(expectedResultSchema, result.getSchema());

        for (int batchIdx = 0; batchIdx < result.numTuples(); ++batchIdx, ++rowIdx) {
          assertEquals(
              ((Boolean) expectedResults[rowIdx][0]).booleanValue(),
              result.getBoolean(0, batchIdx));
          assertEquals((expectedResults[rowIdx][1]).toString(), result.getString(1, batchIdx));
          assertEquals(
              ((Long) expectedResults[rowIdx][2]).longValue(), result.getLong(2, batchIdx));
          assertEquals(
              Double.doubleToLongBits(((Double) expectedResults[rowIdx][3]).doubleValue()),
              Double.doubleToLongBits(result.getDouble(3, batchIdx)));
          assertEquals((expectedResults[rowIdx][4]).toString(), result.getString(4, batchIdx));
        }
      }
    }
    assertEquals(expectedResults.length, rowIdx);
    splitOp.close();
  }
コード例 #3
0
ファイル: SplitTest.java プロジェクト: scriptreiter/myria
  /**
   * Test output spanning multiple batches. All integers from 0 to 2 * TupleBatch.BATCH_SIZE are
   * concatenated as a single comma-separated string. Result should contain each integer from the
   * input in its own row.
   *
   * @throws DbException
   */
  @Test
  public void testAllBatchesReturned() throws DbException {
    final Schema schema = Schema.ofFields("joined_ints", Type.STRING_TYPE);
    final Schema expectedResultSchema =
        Schema.appendColumn(schema, Type.STRING_TYPE, "joined_ints_splits");
    final TupleBatchBuffer input = new TupleBatchBuffer(schema);
    final long expectedResults = 2 * TupleBatch.BATCH_SIZE + 1;
    StringBuilder sb = new StringBuilder();
    for (int i = 0; i < expectedResults; ++i) {
      sb.append(i);
      if (i < expectedResults - 1) {
        sb.append(",");
      }
    }
    input.putString(0, sb.toString());

    Split splitOp = new Split(new TupleSource(input), 0, ",");
    splitOp.open(TestEnvVars.get());
    long rowIdx = 0;
    while (!splitOp.eos()) {
      TupleBatch result = splitOp.nextReady();
      if (result != null) {
        assertEquals(expectedResultSchema, result.getSchema());

        for (int batchIdx = 0; batchIdx < result.numTuples(); ++batchIdx, ++rowIdx) {
          assertEquals(rowIdx, Integer.parseInt(result.getString(1, batchIdx)));
        }
      }
    }
    assertEquals(expectedResults, rowIdx);
    splitOp.close();
  }
コード例 #4
0
ファイル: SplitTest.java プロジェクト: scriptreiter/myria
 @Test(expected = PatternSyntaxException.class)
 public void testInvalidRegex() throws DbException {
   final Schema schema = Schema.ofFields("string", Type.STRING_TYPE);
   final TupleBatchBuffer input = new TupleBatchBuffer(schema);
   input.putString(0, "foo");
   Split splitOp = new Split(new TupleSource(input), 0, "?:(");
   splitOp.open(TestEnvVars.get());
 }
コード例 #5
0
ファイル: SplitTest.java プロジェクト: scriptreiter/myria
 @Test(expected = IllegalStateException.class)
 public void testSplitColumnInvalidType() throws DbException {
   final Schema schema = Schema.ofFields("long", Type.LONG_TYPE);
   final TupleBatchBuffer input = new TupleBatchBuffer(schema);
   input.putLong(0, 1L);
   Split splitOp = new Split(new TupleSource(input), 0, ":");
   splitOp.open(TestEnvVars.get());
 }
コード例 #6
0
ファイル: QueryConstruct.java プロジェクト: jamesmarva/myria
  /**
   * Builds the query plan to update the {@link Server}'s master catalog with the number of tuples
   * in every relation written by a subquery. The query plan is basically "SELECT RelationKey,
   * COUNT(*)" -> Collect at master -> "SELECT RelationKey, SUM(counts)".
   *
   * @param relationsWritten the metadata about which relations were written during the execution of
   *     this subquery.
   * @param server the server on which the catalog will be updated
   * @return the query plan to update the master's catalog with the new number of tuples for all
   *     written relations.
   */
  public static SubQuery getRelationTupleUpdateSubQuery(
      final Map<RelationKey, RelationWriteMetadata> relationsWritten, final Server server) {
    ExchangePairID collectId = ExchangePairID.newID();
    Schema schema =
        Schema.ofFields(
            "userName",
            Type.STRING_TYPE,
            "programName",
            Type.STRING_TYPE,
            "relationName",
            Type.STRING_TYPE,
            "tupleCount",
            Type.LONG_TYPE);

    String dbms = server.getDBMS();
    Preconditions.checkState(
        dbms != null, "Server must have a configured DBMS environment variable");

    /*
     * Worker plans: for each relation, create a {@link DbQueryScan} to get the count, an {@link Apply} to add the
     * {@link RelationKey}, then a {@link CollectProducer} to send the count to the master.
     */
    Map<Integer, SubQueryPlan> workerPlans = Maps.newHashMap();
    for (RelationWriteMetadata meta : relationsWritten.values()) {
      Set<Integer> workers = meta.getWorkers();
      RelationKey relation = meta.getRelationKey();
      for (Integer worker : workers) {
        DbQueryScan localCount =
            new DbQueryScan(
                "SELECT COUNT(*) FROM " + relation.toString(dbms),
                Schema.ofFields("tupleCount", Type.LONG_TYPE));
        List<Expression> expressions =
            ImmutableList.of(
                new Expression(
                    schema.getColumnName(0), new ConstantExpression(relation.getUserName())),
                new Expression(
                    schema.getColumnName(1), new ConstantExpression(relation.getProgramName())),
                new Expression(
                    schema.getColumnName(2), new ConstantExpression(relation.getRelationName())),
                new Expression(schema.getColumnName(3), new VariableExpression(0)));
        Apply addRelationName = new Apply(localCount, expressions);
        CollectProducer producer =
            new CollectProducer(addRelationName, collectId, MyriaConstants.MASTER_ID);
        if (!workerPlans.containsKey(worker)) {
          workerPlans.put(worker, new SubQueryPlan(producer));
        } else {
          workerPlans.get(worker).addRootOp(producer);
        }
      }
    }

    /* Master plan: collect, sum, insert the updates. */
    CollectConsumer consumer = new CollectConsumer(schema, collectId, workerPlans.keySet());
    MultiGroupByAggregate aggCounts =
        new MultiGroupByAggregate(
            consumer, new int[] {0, 1, 2}, new SingleColumnAggregatorFactory(3, AggregationOp.SUM));
    UpdateCatalog catalog = new UpdateCatalog(aggCounts, server);
    SubQueryPlan masterPlan = new SubQueryPlan(catalog);

    return new SubQuery(masterPlan, workerPlans);
  }