Examples with InputPartition org.apache.spark.sql.connector.read.InputPartition used on opensource projects

Example 1 with InputPartition

use of org.apache.spark.sql.connector.read.InputPartition in project iceberg by apache.

the class TestFilteredScan method testUnpartitionedCaseInsensitiveIDFilters.

@Test
public void testUnpartitionedCaseInsensitiveIDFilters() {
    CaseInsensitiveStringMap options = new CaseInsensitiveStringMap(ImmutableMap.of("path", unpartitioned.toString()));
    // set spark.sql.caseSensitive to false
    String caseSensitivityBeforeTest = TestFilteredScan.spark.conf().get("spark.sql.caseSensitive");
    TestFilteredScan.spark.conf().set("spark.sql.caseSensitive", "false");
    try {
        for (int i = 0; i < 10; i += 1) {
            SparkScanBuilder builder = new SparkScanBuilder(spark, TABLES.load(options.get("path")), options).caseSensitive(false);
            // note lower(ID) == lower(id), so there must be a match
            pushFilters(builder, EqualTo.apply("ID", i));
            Batch scan = builder.build().toBatch();
            InputPartition[] tasks = scan.planInputPartitions();
            Assert.assertEquals("Should only create one task for a small file", 1, tasks.length);
            // validate row filtering
            assertEqualsSafe(SCHEMA.asStruct(), expected(i), read(unpartitioned.toString(), vectorized, "id = " + i));
        }
    } finally {
        // return global conf to previous state
        TestFilteredScan.spark.conf().set("spark.sql.caseSensitive", caseSensitivityBeforeTest);
    }
}

Example 2 with InputPartition

use of org.apache.spark.sql.connector.read.InputPartition in project iceberg by apache.

the class TestFilteredScan method testUnpartitionedTimestampFilter.

@Test
public void testUnpartitionedTimestampFilter() {
    CaseInsensitiveStringMap options = new CaseInsensitiveStringMap(ImmutableMap.of("path", unpartitioned.toString()));
    SparkScanBuilder builder = new SparkScanBuilder(spark, TABLES.load(options.get("path")), options);
    pushFilters(builder, LessThan.apply("ts", "2017-12-22T00:00:00+00:00"));
    Batch scan = builder.build().toBatch();
    InputPartition[] tasks = scan.planInputPartitions();
    Assert.assertEquals("Should only create one task for a small file", 1, tasks.length);
    assertEqualsSafe(SCHEMA.asStruct(), expected(5, 6, 7, 8, 9), read(unpartitioned.toString(), vectorized, "ts < cast('2017-12-22 00:00:00+00:00' as timestamp)"));
}

Example 3 with InputPartition

use of org.apache.spark.sql.connector.read.InputPartition in project iceberg by apache.

the class TestFilteredScan method testBucketPartitionedIDFilters.

@Test
public void testBucketPartitionedIDFilters() {
    Table table = buildPartitionedTable("bucketed_by_id", BUCKET_BY_ID, "bucket4", "id");
    CaseInsensitiveStringMap options = new CaseInsensitiveStringMap(ImmutableMap.of("path", table.location()));
    Batch unfiltered = new SparkScanBuilder(spark, TABLES.load(options.get("path")), options).build().toBatch();
    Assert.assertEquals("Unfiltered table should created 4 read tasks", 4, unfiltered.planInputPartitions().length);
    for (int i = 0; i < 10; i += 1) {
        SparkScanBuilder builder = new SparkScanBuilder(spark, TABLES.load(options.get("path")), options);
        pushFilters(builder, EqualTo.apply("id", i));
        Batch scan = builder.build().toBatch();
        InputPartition[] tasks = scan.planInputPartitions();
        // validate predicate push-down
        Assert.assertEquals("Should create one task for a single bucket", 1, tasks.length);
        // validate row filtering
        assertEqualsSafe(SCHEMA.asStruct(), expected(i), read(table.location(), vectorized, "id = " + i));
    }
}

Example 4 with InputPartition

use of org.apache.spark.sql.connector.read.InputPartition in project iceberg by apache.

the class TestFilteredScan method testUnpartitionedIDFilters.

@Test
public void testUnpartitionedIDFilters() {
    CaseInsensitiveStringMap options = new CaseInsensitiveStringMap(ImmutableMap.of("path", unpartitioned.toString()));
    SparkScanBuilder builder = new SparkScanBuilder(spark, TABLES.load(options.get("path")), options);
    for (int i = 0; i < 10; i += 1) {
        pushFilters(builder, EqualTo.apply("id", i));
        Batch scan = builder.build().toBatch();
        InputPartition[] partitions = scan.planInputPartitions();
        Assert.assertEquals("Should only create one task for a small file", 1, partitions.length);
        // validate row filtering
        assertEqualsSafe(SCHEMA.asStruct(), expected(i), read(unpartitioned.toString(), vectorized, "id = " + i));
    }
}

Example 5 with InputPartition

use of org.apache.spark.sql.connector.read.InputPartition in project phoenix-connectors by apache.

the class PhoenixScan method planInputPartitions.

@Override
public InputPartition[] planInputPartitions() {
    populateOverriddenProperties();
    try (Connection conn = DriverManager.getConnection(JDBC_PROTOCOL + JDBC_PROTOCOL_SEPARATOR + zkUrl, overriddenProps)) {
        List<ColumnInfo> columnInfos = PhoenixRuntime.generateColumnInfo(conn, tableName, new ArrayList<>(Arrays.asList(schema.names())));
        final Statement statement = conn.createStatement();
        final String selectStatement = QueryUtil.constructSelectStatement(tableName, columnInfos, whereClause);
        if (selectStatement == null) {
            throw new NullPointerException();
        }
        final PhoenixStatement pstmt = statement.unwrap(PhoenixStatement.class);
        // Optimize the query plan so that we potentially use secondary indexes
        final QueryPlan queryPlan = pstmt.optimizeQuery(selectStatement);
        final org.apache.hadoop.hbase.client.Scan scan = queryPlan.getContext().getScan();
        // Initialize the query plan so it sets up the parallel scans
        queryPlan.iterator(MapReduceParallelScanGrouper.getInstance());
        List<KeyRange> allSplits = queryPlan.getSplits();
        // Get the RegionSizeCalculator
        PhoenixConnection phxConn = conn.unwrap(PhoenixConnection.class);
        org.apache.hadoop.hbase.client.Connection connection = phxConn.getQueryServices().getAdmin().getConnection();
        RegionLocator regionLocator = connection.getRegionLocator(TableName.valueOf(queryPlan.getTableRef().getTable().getPhysicalName().toString()));
        final InputPartition[] partitions = new PhoenixInputPartition[allSplits.size()];
        int partitionCount = 0;
        for (List<org.apache.hadoop.hbase.client.Scan> scans : queryPlan.getScans()) {
            // Get the region location
            HRegionLocation location = regionLocator.getRegionLocation(scans.get(0).getStartRow(), false);
            String regionLocation = location.getHostname();
            // Get the region size
            long regionSize = CompatUtil.getSize(regionLocator, connection.getAdmin(), location);
            phoenixDataSourceOptions = new PhoenixDataSourceReadOptions(zkUrl, currentScnValue, tenantId, selectStatement, overriddenProps);
            if (splitByStats) {
                for (org.apache.hadoop.hbase.client.Scan aScan : scans) {
                    partitions[partitionCount++] = new PhoenixInputPartition(new PhoenixInputSplit(Collections.singletonList(aScan), regionSize, regionLocation));
                }
            } else {
                partitions[partitionCount++] = new PhoenixInputPartition(new PhoenixInputSplit(scans, regionSize, regionLocation));
            }
        }
        return partitions;
    } catch (Exception e) {
        throw new RuntimeException("Unable to plan query", e);
    }
}