Examples with ColumnDesc io.cdap.cdap.proto.ColumnDesc used on opensource projects

Example 6 with ColumnDesc

use of io.cdap.cdap.proto.ColumnDesc in project cdap by caskdata.

the class HiveExploreServiceFileSetTestRun method testPartitionedFileSet.

private void testPartitionedFileSet(@Nullable String dbName, @Nullable String tableName) throws Exception {
    DatasetId datasetInstanceId = NAMESPACE_ID.dataset("parted");
    String hiveTableName = getDatasetHiveName(datasetInstanceId);
    String showTablesCommand = "show tables";
    FileSetProperties.Builder props = PartitionedFileSetProperties.builder().setPartitioning(Partitioning.builder().addStringField("str").addIntField("num").build()).setBasePath("parted").setEnableExploreOnCreate(true).setSerDe("org.apache.hadoop.hive.serde2.avro.AvroSerDe").setExploreInputFormat("org.apache.hadoop.hive.ql.io.avro.AvroContainerInputFormat").setExploreOutputFormat("org.apache.hadoop.hive.ql.io.avro.AvroContainerOutputFormat").setTableProperty("avro.schema.literal", SCHEMA.toString());
    if (tableName != null) {
        props.setExploreTableName(tableName);
        hiveTableName = tableName;
    }
    String queryTableName = hiveTableName;
    if (dbName != null) {
        props.setExploreDatabaseName(dbName);
        runCommand(NAMESPACE_ID, "create database " + dbName, false, null, null);
        showTablesCommand += " in " + dbName;
        queryTableName = dbName + "." + queryTableName;
    }
    // create a time partitioned file set
    datasetFramework.addInstance("partitionedFileSet", datasetInstanceId, props.build());
    // verify that the hive table was created for this file set
    runCommand(NAMESPACE_ID, showTablesCommand, true, null, Lists.newArrayList(new QueryResult(Lists.<Object>newArrayList(hiveTableName))));
    // Accessing dataset instance to perform data operations
    final PartitionedFileSet partitioned = datasetFramework.getDataset(datasetInstanceId, DatasetDefinition.NO_ARGUMENTS, null);
    Assert.assertNotNull(partitioned);
    FileSet fileSet = partitioned.getEmbeddedFileSet();
    // add some partitions. Beware that Hive expects a partition to be a directory, so we create dirs with one file
    Location locationX1 = fileSet.getLocation("fileX1/nn");
    Location locationY1 = fileSet.getLocation("fileY1/nn");
    Location locationX2 = fileSet.getLocation("fileX2/nn");
    Location locationY2 = fileSet.getLocation("fileY2/nn");
    FileWriterHelper.generateAvroFile(locationX1.getOutputStream(), "x", 1, 2);
    FileWriterHelper.generateAvroFile(locationY1.getOutputStream(), "y", 1, 2);
    FileWriterHelper.generateAvroFile(locationX2.getOutputStream(), "x", 2, 3);
    FileWriterHelper.generateAvroFile(locationY2.getOutputStream(), "y", 2, 3);
    final PartitionKey keyX1 = PartitionKey.builder().addStringField("str", "x").addIntField("num", 1).build();
    PartitionKey keyY1 = PartitionKey.builder().addStringField("str", "y").addIntField("num", 1).build();
    final PartitionKey keyX2 = PartitionKey.builder().addStringField("str", "x").addIntField("num", 2).build();
    PartitionKey keyY2 = PartitionKey.builder().addStringField("str", "y").addIntField("num", 2).build();
    addPartition(partitioned, keyX1, "fileX1");
    addPartition(partitioned, keyY1, "fileY1");
    addPartition(partitioned, keyX2, "fileX2");
    addPartition(partitioned, keyY2, "fileY2");
    // verify that the partitions were added to Hive
    validatePartitions(queryTableName, partitioned, ImmutableList.of(keyX1, keyX2, keyY1, keyY2));
    // verify that count() and where... work in Hive
    runCommand(NAMESPACE_ID, "SELECT count(*) AS count FROM " + queryTableName, true, Lists.newArrayList(new ColumnDesc("count", "BIGINT", 1, null)), Lists.newArrayList(new QueryResult(Lists.<Object>newArrayList(4L))));
    runCommand(NAMESPACE_ID, "SELECT * FROM " + queryTableName + " WHERE num = 2 ORDER BY key, value", true, Lists.newArrayList(new ColumnDesc(hiveTableName + ".key", "STRING", 1, null), new ColumnDesc(hiveTableName + ".value", "STRING", 2, null), new ColumnDesc(hiveTableName + ".str", "STRING", 3, null), new ColumnDesc(hiveTableName + ".num", "INT", 4, null)), Lists.newArrayList(new QueryResult(Lists.<Object>newArrayList("x2", "#2", "x", 2)), new QueryResult(Lists.<Object>newArrayList("y2", "#2", "y", 2))));
    // drop a partition and query again
    dropPartition(partitioned, keyX2);
    validatePartitions(queryTableName, partitioned, ImmutableSet.of(keyX1, keyY1, keyY2));
    // attempt a transaction that drops one partition, adds another, and then fails
    try {
        doTransaction(partitioned, new Runnable() {

            @Override
            public void run() {
                partitioned.dropPartition(keyX1);
                partitioned.addPartition(keyX2, "fileX2");
                Assert.fail("fail tx");
            }
        });
    } catch (TransactionFailureException e) {
    // expected
    }
    // validate that both the drop and addPartition were undone
    validatePartitions(queryTableName, partitioned, ImmutableSet.of(keyX1, keyY1, keyY2));
    // attempt a transaction that attempts to add an existing partition, hence fails
    try {
        doTransaction(partitioned, new Runnable() {

            @Override
            public void run() {
                partitioned.addPartition(keyX1, "fileX1");
                throw new RuntimeException("on purpose");
            }
        });
    } catch (TransactionFailureException e) {
        // expected if the cause is not "on purpose"
        Assert.assertTrue(e.getCause() instanceof DataSetException);
    }
    // validate that both the drop and addPartition were undone
    validatePartitions(queryTableName, partitioned, ImmutableSet.of(keyX1, keyY1, keyY2));
    // drop a partition directly from hive
    runCommand(NAMESPACE_ID, "ALTER TABLE " + queryTableName + " DROP PARTITION (str='y', num=2)", false, null, null);
    // verify that one more value is gone now, namely y2, in Hive, but the PFS still has it
    validatePartitionsInHive(queryTableName, ImmutableSet.of(keyX1, keyY1));
    validatePartitionsInPFS(partitioned, ImmutableSet.of(keyX1, keyY1, keyY2));
    // make sure the partition can still be dropped from the PFS dataset
    dropPartition(partitioned, keyY2);
    validatePartitions(queryTableName, partitioned, ImmutableSet.of(keyX1, keyY1));
    // change the explore schema by updating the props
    datasetFramework.updateInstance(datasetInstanceId, props.setTableProperty("avro.schema.literal", K_SCHEMA.toString()).build());
    // valudate the schema was updated
    validatePartitions(queryTableName, partitioned, ImmutableSet.of(keyX1, keyY1), true);
    // disable explore by updating the props
    datasetFramework.updateInstance(datasetInstanceId, props.setEnableExploreOnCreate(false).build());
    // verify the Hive table is gone
    runCommand(NAMESPACE_ID, showTablesCommand, false, null, Collections.<QueryResult>emptyList());
    // re-enable explore by updating the props
    datasetFramework.updateInstance(datasetInstanceId, props.setEnableExploreOnCreate(true).build());
    // verify the Hive table is back
    runCommand(NAMESPACE_ID, showTablesCommand, true, null, Lists.newArrayList(new QueryResult(Lists.<Object>newArrayList(hiveTableName))));
    // drop the dataset
    datasetFramework.deleteInstance(datasetInstanceId);
    // verify the Hive table is gone
    runCommand(NAMESPACE_ID, "show tables", false, null, Collections.<QueryResult>emptyList());
}

Example 7 with ColumnDesc

use of io.cdap.cdap.proto.ColumnDesc in project cdap by caskdata.

the class HiveExploreServiceFileSetTestRun method testPartitionedAvroSchemaUpdate.

@Test
public void testPartitionedAvroSchemaUpdate() throws Exception {
    final DatasetId datasetId = NAMESPACE_ID.dataset("avroupd");
    final String tableName = getDatasetHiveName(datasetId);
    // create a time partitioned file set
    datasetFramework.addInstance(PartitionedFileSet.class.getName(), datasetId, PartitionedFileSetProperties.builder().setPartitioning(Partitioning.builder().addIntField("number").build()).setEnableExploreOnCreate(true).setSerDe("org.apache.hadoop.hive.serde2.avro.AvroSerDe").setExploreInputFormat("org.apache.hadoop.hive.ql.io.avro.AvroContainerInputFormat").setExploreOutputFormat("org.apache.hadoop.hive.ql.io.avro.AvroContainerOutputFormat").setTableProperty("avro.schema.literal", SCHEMA.toString()).build());
    // Accessing dataset instance to perform data operations
    PartitionedFileSet partitioned = datasetFramework.getDataset(datasetId, DatasetDefinition.NO_ARGUMENTS, null);
    Assert.assertNotNull(partitioned);
    FileSet fileSet = partitioned.getEmbeddedFileSet();
    // add a partition
    Location location4 = fileSet.getLocation("file4/nn");
    FileWriterHelper.generateAvroFile(location4.getOutputStream(), "x", 4, 5);
    addPartition(partitioned, PartitionKey.builder().addIntField("number", 4).build(), "file4");
    // new partition should have new format, validate with query
    List<ColumnDesc> expectedColumns = Lists.newArrayList(new ColumnDesc(tableName + ".key", "STRING", 1, null), new ColumnDesc(tableName + ".value", "STRING", 2, null), new ColumnDesc(tableName + ".number", "INT", 3, null));
    runCommand(NAMESPACE_ID, "SELECT * FROM " + tableName + " WHERE number=4", true, expectedColumns, Lists.newArrayList(// avro file has key=x4, value=#4
    new QueryResult(Lists.<Object>newArrayList("x4", "#4", 4))));
    // update the partitioned file set
    datasetFramework.updateInstance(datasetId, PartitionedFileSetProperties.builder().setPartitioning(Partitioning.builder().addIntField("number").build()).setEnableExploreOnCreate(true).setSerDe("org.apache.hadoop.hive.serde2.avro.AvroSerDe").setExploreInputFormat("org.apache.hadoop.hive.ql.io.avro.AvroContainerInputFormat").setExploreOutputFormat("org.apache.hadoop.hive.ql.io.avro.AvroContainerOutputFormat").setTableProperty("avro.schema.literal", K_SCHEMA.toString()).build());
    expectedColumns = Lists.newArrayList(new ColumnDesc(tableName + ".key", "STRING", 1, null), new ColumnDesc(tableName + ".number", "INT", 2, null));
    runCommand(NAMESPACE_ID, "SELECT * FROM " + tableName + " WHERE number=4", true, expectedColumns, Lists.newArrayList(// avro file has key=x4, value=#4
    new QueryResult(Lists.<Object>newArrayList("x4", 4))));
}

Example 8 with ColumnDesc

use of io.cdap.cdap.proto.ColumnDesc in project cdap by caskdata.

the class HiveExploreServiceFileSetTestRun method testTPFSWithDateTimestamp.

@Test
public void testTPFSWithDateTimestamp() throws Exception {
    TimeZone.setDefault(TimeZone.getTimeZone("UTC"));
    final DatasetId datasetInstanceId = NAMESPACE_ID.dataset("dtfs");
    final String tableName = getDatasetHiveName(datasetInstanceId);
    final Schema dtSchema = Schema.recordOf("dt", Schema.Field.of("id", Schema.of(Schema.Type.INT)), Schema.Field.of("name", Schema.nullableOf(Schema.of(Schema.Type.STRING))), Schema.Field.of("dt", Schema.of(Schema.LogicalType.DATE)), Schema.Field.of("ts", Schema.nullableOf(Schema.of(Schema.LogicalType.TIMESTAMP_MILLIS))));
    // create a file set
    datasetFramework.addInstance("timePartitionedFileSet", datasetInstanceId, FileSetProperties.builder().setBasePath("somePath").setEnableExploreOnCreate(true).setSerDe("org.apache.hadoop.hive.serde2.avro.AvroSerDe").setExploreInputFormat("org.apache.hadoop.hive.ql.io.avro.AvroContainerInputFormat").setExploreOutputFormat("org.apache.hadoop.hive.ql.io.avro.AvroContainerOutputFormat").setTableProperty("avro.schema.literal", dtSchema.toString()).build());
    // verify that the hive table was created for this file set
    runCommand(NAMESPACE_ID, "show tables", true, Lists.newArrayList(new ColumnDesc("tab_name", "STRING", 1, "from deserializer")), Lists.newArrayList(new QueryResult(Lists.newArrayList(tableName))));
    // Accessing dataset instance to perform data operations
    TimePartitionedFileSet tpfs = datasetFramework.getDataset(datasetInstanceId, DatasetDefinition.NO_ARGUMENTS, null);
    Assert.assertNotNull(tpfs);
    Location location1 = tpfs.getEmbeddedFileSet().getLocation("file1/nn");
    generateAvroFile(location1.getOutputStream(), dtSchema);
    // add some partitions. Beware that Hive expects a partition to be a directory, so we create dirs with one file
    long time1 = DATE_FORMAT.parse("12/10/14 1:00 am").getTime();
    addTimePartition(tpfs, time1, "file1");
    // verify that we can query the date and timestamp in the file with Hive
    runCommand(NAMESPACE_ID, "SELECT id, name, dt, ts FROM " + tableName + " LIMIT 50", true, Lists.newArrayList(new ColumnDesc("id", "INT", 1, null), new ColumnDesc("name", "STRING", 2, null), new ColumnDesc("dt", "DATE", 3, null), new ColumnDesc("ts", "TIMESTAMP", 4, null)), Lists.newArrayList(new QueryResult(Lists.newArrayList(1, "alice", "1970-01-01", "2018-09-07 16:09:50.595"))));
    // drop the dataset
    datasetFramework.deleteInstance(datasetInstanceId);
    // verify the Hive table is gone
    runCommand(NAMESPACE_ID, "show tables", false, Lists.newArrayList(new ColumnDesc("tab_name", "STRING", 1, "from deserializer")), Collections.emptyList());
    // create a file set
    datasetFramework.addInstance("timePartitionedFileSet", datasetInstanceId, FileSetProperties.builder().setBasePath("somePath").setEnableExploreOnCreate(true).setSerDe("org.apache.hadoop.hive.serde2.avro.AvroSerDe").setExploreInputFormat("org.apache.hadoop.hive.ql.io.avro.AvroContainerInputFormat").setExploreOutputFormat("org.apache.hadoop.hive.ql.io.avro.AvroContainerOutputFormat").setTableProperty("avro.schema.literal", dtSchema.toString()).build());
    // verify that the hive table was created for this file set
    runCommand(NAMESPACE_ID, "show tables", true, Lists.newArrayList(new ColumnDesc("tab_name", "STRING", 1, "from deserializer")), Lists.newArrayList(new QueryResult(Lists.newArrayList(tableName))));
}

Example 9 with ColumnDesc

use of io.cdap.cdap.proto.ColumnDesc in project cdap by caskdata.

the class HiveExploreServiceTimeoutTest method testTimeoutNoResults.

@Test
public void testTimeoutNoResults() throws Exception {
    Set<Long> beforeTxns = transactionManager.getCurrentState().getInProgress().keySet();
    QueryHandle handle = exploreService.execute(NAMESPACE_ID, "drop table if exists not_existing_table_name");
    Set<Long> queryTxns = Sets.difference(transactionManager.getCurrentState().getInProgress().keySet(), beforeTxns);
    Assert.assertFalse(queryTxns.isEmpty());
    QueryStatus status = waitForCompletionStatus(handle, 200, TimeUnit.MILLISECONDS, 20);
    Assert.assertEquals(QueryStatus.OpStatus.FINISHED, status.getStatus());
    Assert.assertFalse(status.hasResults());
    List<ColumnDesc> schema = exploreService.getResultSchema(handle);
    // Sleep for some time for txn to get closed
    TimeUnit.SECONDS.sleep(1);
    // Make sure that the transaction got closed
    Assert.assertEquals(ImmutableSet.<Long>of(), Sets.intersection(queryTxns, transactionManager.getCurrentState().getInProgress().keySet()).immutableCopy());
    // Check if calls using inactive handle still work
    Assert.assertEquals(status, exploreService.getStatus(handle));
    Assert.assertEquals(schema, exploreService.getResultSchema(handle));
    exploreService.close(handle);
    // Sleep for timeout to happen
    TimeUnit.SECONDS.sleep(INACTIVE_OPERATION_TIMEOUT_SECS + 3);
    try {
        exploreService.getStatus(handle);
        Assert.fail("Should throw HandleNotFoundException due to operation cleanup");
    } catch (HandleNotFoundException e) {
    // Expected exception due to timeout
    }
}

Example 10 with ColumnDesc

use of io.cdap.cdap.proto.ColumnDesc in project cdap by caskdata.

the class HiveExploreServiceTimeoutTest method testTimeoutFetchAllResults.

@Test
public void testTimeoutFetchAllResults() throws Exception {
    Set<Long> beforeTxns = transactionManager.getCurrentState().getInProgress().keySet();
    QueryHandle handle = exploreService.execute(NAMESPACE_ID, "select key, value from " + MY_TABLE_NAME);
    Set<Long> queryTxns = Sets.difference(transactionManager.getCurrentState().getInProgress().keySet(), beforeTxns);
    Assert.assertFalse(queryTxns.isEmpty());
    QueryStatus status = waitForCompletionStatus(handle, 200, TimeUnit.MILLISECONDS, 20);
    Assert.assertEquals(QueryStatus.OpStatus.FINISHED, status.getStatus());
    Assert.assertTrue(status.hasResults());
    List<ColumnDesc> schema = exploreService.getResultSchema(handle);
    // noinspection StatementWithEmptyBody
    while (!exploreService.nextResults(handle, 100).isEmpty()) {
    // nothing to do
    }
    // Sleep for some time for txn to get closed
    TimeUnit.SECONDS.sleep(1);
    // Make sure that the transaction got closed
    Assert.assertEquals(ImmutableSet.<Long>of(), Sets.intersection(queryTxns, transactionManager.getCurrentState().getInProgress().keySet()).immutableCopy());
    // Check if calls using inactive handle still work
    Assert.assertEquals(status, exploreService.getStatus(handle));
    Assert.assertEquals(schema, exploreService.getResultSchema(handle));
    exploreService.close(handle);
    // Sleep for timeout to happen
    TimeUnit.SECONDS.sleep(INACTIVE_OPERATION_TIMEOUT_SECS + 3);
    try {
        exploreService.getStatus(handle);
        Assert.fail("Should throw HandleNotFoundException due to operation cleanup");
    } catch (HandleNotFoundException e) {
    // Expected exception due to timeout
    }
}