Examples with HoodieFlinkWriteableTestTable org.apache.hudi.testutils.HoodieFlinkWriteableTestTable used on opensource projects

Example 1 with HoodieFlinkWriteableTestTable

use of org.apache.hudi.testutils.HoodieFlinkWriteableTestTable in project hudi by apache.

the class TestFlinkHoodieBloomIndex method testCheckUUIDsAgainstOneFile.

@Test
public void testCheckUUIDsAgainstOneFile() throws Exception {
    final String partition = "2016/01/31";
    // Create some records to use
    String recordStr1 = "{\"_row_key\":\"1eb5b87a-1feh-4edd-87b4-6ec96dc405a0\"," + "\"time\":\"2016-01-31T03:16:41.415Z\",\"number\":12}";
    String recordStr2 = "{\"_row_key\":\"2eb5b87b-1feu-4edd-87b4-6ec96dc405a0\"," + "\"time\":\"2016-01-31T03:20:41.415Z\",\"number\":100}";
    String recordStr3 = "{\"_row_key\":\"3eb5b87c-1fej-4edd-87b4-6ec96dc405a0\"," + "\"time\":\"2016-01-31T03:16:41.415Z\",\"number\":15}";
    String recordStr4 = "{\"_row_key\":\"4eb5b87c-1fej-4edd-87b4-6ec96dc405a0\"," + "\"time\":\"2016-01-31T03:16:41.415Z\",\"number\":32}";
    RawTripTestPayload rowChange1 = new RawTripTestPayload(recordStr1);
    HoodieRecord record1 = new HoodieAvroRecord(new HoodieKey(rowChange1.getRowKey(), rowChange1.getPartitionPath()), rowChange1);
    RawTripTestPayload rowChange2 = new RawTripTestPayload(recordStr2);
    HoodieRecord record2 = new HoodieAvroRecord(new HoodieKey(rowChange2.getRowKey(), rowChange2.getPartitionPath()), rowChange2);
    RawTripTestPayload rowChange3 = new RawTripTestPayload(recordStr3);
    HoodieRecord record3 = new HoodieAvroRecord(new HoodieKey(rowChange3.getRowKey(), rowChange3.getPartitionPath()), rowChange3);
    RawTripTestPayload rowChange4 = new RawTripTestPayload(recordStr4);
    HoodieRecord record4 = new HoodieAvroRecord(new HoodieKey(rowChange4.getRowKey(), rowChange4.getPartitionPath()), rowChange4);
    // We write record1, record2 to a base file, but the bloom filter contains (record1,
    // record2, record3).
    BloomFilter filter = BloomFilterFactory.createBloomFilter(10000, 0.0000001, -1, BloomFilterTypeCode.SIMPLE.name());
    filter.add(record3.getRecordKey());
    HoodieFlinkWriteableTestTable testTable = HoodieFlinkWriteableTestTable.of(metaClient, SCHEMA, filter);
    String fileId = testTable.addCommit("000").getFileIdWithInserts(partition, record1, record2);
    String filename = testTable.getBaseFileNameById(fileId);
    // The bloom filter contains 3 records
    assertTrue(filter.mightContain(record1.getRecordKey()));
    assertTrue(filter.mightContain(record2.getRecordKey()));
    assertTrue(filter.mightContain(record3.getRecordKey()));
    assertFalse(filter.mightContain(record4.getRecordKey()));
    // Compare with file
    List<String> uuids = asList(record1.getRecordKey(), record2.getRecordKey(), record3.getRecordKey(), record4.getRecordKey());
    HoodieWriteConfig config = HoodieWriteConfig.newBuilder().withPath(basePath).build();
    HoodieFlinkTable table = HoodieFlinkTable.create(config, context, metaClient);
    List<String> results = HoodieIndexUtils.filterKeysFromFile(new Path(java.nio.file.Paths.get(basePath, partition, filename).toString()), uuids, hadoopConf);
    assertEquals(results.size(), 2);
    assertTrue(results.get(0).equals("1eb5b87a-1feh-4edd-87b4-6ec96dc405a0") || results.get(1).equals("1eb5b87a-1feh-4edd-87b4-6ec96dc405a0"));
    assertTrue(results.get(0).equals("2eb5b87b-1feu-4edd-87b4-6ec96dc405a0") || results.get(1).equals("2eb5b87b-1feu-4edd-87b4-6ec96dc405a0"));
// TODO(vc): Need more coverage on actual filenames
// assertTrue(results.get(0)._2().equals(filename));
// assertTrue(results.get(1)._2().equals(filename));
}

Example 2 with HoodieFlinkWriteableTestTable

use of org.apache.hudi.testutils.HoodieFlinkWriteableTestTable in project hudi by apache.

the class TestFlinkHoodieBloomIndex method testCheckExists.

@ParameterizedTest(name = TEST_NAME_WITH_PARAMS)
@MethodSource("configParams")
public void testCheckExists(boolean rangePruning, boolean treeFiltering, boolean bucketizedChecking) throws Exception {
    // We have some records to be tagged (two different partitions)
    String recordStr1 = "{\"_row_key\":\"1eb5b87a-1feh-4edd-87b4-6ec96dc405a0\"," + "\"time\":\"2016-01-31T03:16:41.415Z\",\"number\":12}";
    String recordStr2 = "{\"_row_key\":\"2eb5b87b-1feu-4edd-87b4-6ec96dc405a0\"," + "\"time\":\"2016-01-31T03:20:41.415Z\",\"number\":100}";
    String recordStr3 = "{\"_row_key\":\"3eb5b87c-1fej-4edd-87b4-6ec96dc405a0\"," + "\"time\":\"2016-01-31T03:16:41.415Z\",\"number\":15}";
    // record key same as recordStr2
    String recordStr4 = "{\"_row_key\":\"2eb5b87b-1feu-4edd-87b4-6ec96dc405a0\"," + "\"time\":\"2015-01-31T03:16:41.415Z\",\"number\":32}";
    RawTripTestPayload rowChange1 = new RawTripTestPayload(recordStr1);
    HoodieKey key1 = new HoodieKey(rowChange1.getRowKey(), rowChange1.getPartitionPath());
    HoodieRecord record1 = new HoodieAvroRecord(key1, rowChange1);
    RawTripTestPayload rowChange2 = new RawTripTestPayload(recordStr2);
    HoodieKey key2 = new HoodieKey(rowChange2.getRowKey(), rowChange2.getPartitionPath());
    HoodieRecord record2 = new HoodieAvroRecord(key2, rowChange2);
    RawTripTestPayload rowChange3 = new RawTripTestPayload(recordStr3);
    HoodieKey key3 = new HoodieKey(rowChange3.getRowKey(), rowChange3.getPartitionPath());
    RawTripTestPayload rowChange4 = new RawTripTestPayload(recordStr4);
    HoodieKey key4 = new HoodieKey(rowChange4.getRowKey(), rowChange4.getPartitionPath());
    HoodieRecord record4 = new HoodieAvroRecord(key4, rowChange4);
    List<HoodieKey> keys = asList(key1, key2, key3, key4);
    // Also create the metadata and config
    HoodieWriteConfig config = makeConfig(rangePruning, treeFiltering, bucketizedChecking);
    HoodieTable hoodieTable = HoodieFlinkTable.create(config, context, metaClient);
    HoodieFlinkWriteableTestTable testTable = HoodieFlinkWriteableTestTable.of(hoodieTable, SCHEMA);
    // Let's tag
    HoodieBloomIndex bloomIndex = new HoodieBloomIndex(config, ListBasedHoodieBloomIndexHelper.getInstance());
    List<HoodieRecord> toTagRecords = new ArrayList<>();
    toTagRecords.add(new HoodieAvroRecord(record4.getKey(), null));
    List<HoodieRecord> taggedRecords = tagLocation(bloomIndex, toTagRecords, hoodieTable);
    Map<HoodieKey, Option<Pair<String, String>>> recordLocations = new HashMap<>();
    for (HoodieRecord taggedRecord : taggedRecords) {
        recordLocations.put(taggedRecord.getKey(), taggedRecord.isCurrentLocationKnown() ? Option.of(Pair.of(taggedRecord.getPartitionPath(), taggedRecord.getCurrentLocation().getFileId())) : Option.empty());
    }
    // Should not find any files
    for (Option<Pair<String, String>> record : recordLocations.values()) {
        assertTrue(!record.isPresent());
    }
    // We create three base file, each having one record. (two different partitions)
    String fileId1 = testTable.addCommit("001").getFileIdWithInserts("2016/01/31", record1);
    String fileId2 = testTable.addCommit("002").getFileIdWithInserts("2016/01/31", record2);
    String fileId3 = testTable.addCommit("003").getFileIdWithInserts("2015/01/31", record4);
    // We do the tag again
    metaClient = HoodieTableMetaClient.reload(metaClient);
    hoodieTable = HoodieFlinkTable.create(config, context, metaClient);
    List<HoodieRecord> toTagRecords1 = new ArrayList<>();
    for (HoodieKey key : keys) {
        taggedRecords.add(new HoodieAvroRecord(key, null));
    }
    taggedRecords = tagLocation(bloomIndex, toTagRecords1, hoodieTable);
    recordLocations.clear();
    for (HoodieRecord taggedRecord : taggedRecords) {
        recordLocations.put(taggedRecord.getKey(), taggedRecord.isCurrentLocationKnown() ? Option.of(Pair.of(taggedRecord.getPartitionPath(), taggedRecord.getCurrentLocation().getFileId())) : Option.empty());
    }
    // Check results
    for (Map.Entry<HoodieKey, Option<Pair<String, String>>> record : recordLocations.entrySet()) {
        if (record.getKey().getRecordKey().equals("1eb5b87a-1feh-4edd-87b4-6ec96dc405a0")) {
            assertTrue(record.getValue().isPresent());
            assertEquals(fileId1, record.getValue().get().getRight());
        } else if (record.getKey().getRecordKey().equals("2eb5b87b-1feu-4edd-87b4-6ec96dc405a0")) {
            assertTrue(record.getValue().isPresent());
            if (record.getKey().getPartitionPath().equals("2015/01/31")) {
                assertEquals(fileId3, record.getValue().get().getRight());
            } else {
                assertEquals(fileId2, record.getValue().get().getRight());
            }
        } else if (record.getKey().getRecordKey().equals("3eb5b87c-1fej-4edd-87b4-6ec96dc405a0")) {
            assertFalse(record.getValue().isPresent());
        }
    }
}

Example 3 with HoodieFlinkWriteableTestTable

use of org.apache.hudi.testutils.HoodieFlinkWriteableTestTable in project hudi by apache.

the class TestFlinkHoodieBloomIndex method testLoadInvolvedFiles.

@ParameterizedTest(name = TEST_NAME_WITH_PARAMS)
@MethodSource("configParams")
public void testLoadInvolvedFiles(boolean rangePruning, boolean treeFiltering, boolean bucketizedChecking) throws Exception {
    HoodieWriteConfig config = makeConfig(rangePruning, treeFiltering, bucketizedChecking);
    HoodieBloomIndex index = new HoodieBloomIndex(config, ListBasedHoodieBloomIndexHelper.getInstance());
    HoodieTable hoodieTable = HoodieFlinkTable.create(config, context, metaClient, false);
    HoodieFlinkWriteableTestTable testTable = HoodieFlinkWriteableTestTable.of(hoodieTable, SCHEMA);
    // Create some partitions, and put some files
    // "2016/01/21": 0 file
    // "2016/04/01": 1 file (2_0_20160401010101.parquet)
    // "2015/03/12": 3 files (1_0_20150312101010.parquet, 3_0_20150312101010.parquet, 4_0_20150312101010.parquet)
    testTable.withPartitionMetaFiles("2016/01/21", "2016/04/01", "2015/03/12");
    RawTripTestPayload rowChange1 = new RawTripTestPayload("{\"_row_key\":\"000\",\"time\":\"2016-01-31T03:16:41.415Z\",\"number\":12}");
    HoodieRecord record1 = new HoodieAvroRecord(new HoodieKey(rowChange1.getRowKey(), rowChange1.getPartitionPath()), rowChange1);
    RawTripTestPayload rowChange2 = new RawTripTestPayload("{\"_row_key\":\"001\",\"time\":\"2016-01-31T03:16:41.415Z\",\"number\":12}");
    HoodieRecord record2 = new HoodieAvroRecord(new HoodieKey(rowChange2.getRowKey(), rowChange2.getPartitionPath()), rowChange2);
    RawTripTestPayload rowChange3 = new RawTripTestPayload("{\"_row_key\":\"002\",\"time\":\"2016-01-31T03:16:41.415Z\",\"number\":12}");
    HoodieRecord record3 = new HoodieAvroRecord(new HoodieKey(rowChange3.getRowKey(), rowChange3.getPartitionPath()), rowChange3);
    RawTripTestPayload rowChange4 = new RawTripTestPayload("{\"_row_key\":\"003\",\"time\":\"2016-01-31T03:16:41.415Z\",\"number\":12}");
    HoodieRecord record4 = new HoodieAvroRecord(new HoodieKey(rowChange4.getRowKey(), rowChange4.getPartitionPath()), rowChange4);
    List<String> partitions = asList("2016/01/21", "2016/04/01", "2015/03/12");
    List<Pair<String, BloomIndexFileInfo>> filesList = index.loadColumnRangesFromFiles(partitions, context, hoodieTable);
    // Still 0, as no valid commit
    assertEquals(0, filesList.size());
    testTable.addCommit("20160401010101").withInserts("2016/04/01", "2");
    testTable.addCommit("20150312101010").withInserts("2015/03/12", "1").withInserts("2015/03/12", "3", record1).withInserts("2015/03/12", "4", record2, record3, record4);
    metaClient.reloadActiveTimeline();
    filesList = index.loadColumnRangesFromFiles(partitions, context, hoodieTable);
    assertEquals(4, filesList.size());
    if (rangePruning) {
        // these files will not have the key ranges
        assertNull(filesList.get(0).getRight().getMaxRecordKey());
        assertNull(filesList.get(0).getRight().getMinRecordKey());
        assertFalse(filesList.get(1).getRight().hasKeyRanges());
        assertNotNull(filesList.get(2).getRight().getMaxRecordKey());
        assertNotNull(filesList.get(2).getRight().getMinRecordKey());
        assertTrue(filesList.get(3).getRight().hasKeyRanges());
        // no longer sorted, but should have same files.
        List<Pair<String, BloomIndexFileInfo>> expected = asList(Pair.of("2016/04/01", new BloomIndexFileInfo("2")), Pair.of("2015/03/12", new BloomIndexFileInfo("1")), Pair.of("2015/03/12", new BloomIndexFileInfo("3", "000", "000")), Pair.of("2015/03/12", new BloomIndexFileInfo("4", "001", "003")));
        assertEquals(expected, filesList);
    }
}

Example 4 with HoodieFlinkWriteableTestTable

use of org.apache.hudi.testutils.HoodieFlinkWriteableTestTable in project hudi by apache.

the class TestFlinkHoodieBloomIndex method testBloomFilterFalseError.

@ParameterizedTest(name = TEST_NAME_WITH_PARAMS)
@MethodSource("configParams")
public void testBloomFilterFalseError(boolean rangePruning, boolean treeFiltering, boolean bucketizedChecking) throws Exception {
    // We have two hoodie records
    String recordStr1 = "{\"_row_key\":\"1eb5b87a-1feh-4edd-87b4-6ec96dc405a0\"," + "\"time\":\"2016-01-31T03:16:41.415Z\",\"number\":12}";
    String recordStr2 = "{\"_row_key\":\"2eb5b87b-1feu-4edd-87b4-6ec96dc405a0\"," + "\"time\":\"2016-01-31T03:20:41.415Z\",\"number\":100}";
    // We write record1 to a base file, using a bloom filter having both records
    RawTripTestPayload rowChange1 = new RawTripTestPayload(recordStr1);
    HoodieRecord record1 = new HoodieAvroRecord(new HoodieKey(rowChange1.getRowKey(), rowChange1.getPartitionPath()), rowChange1);
    RawTripTestPayload rowChange2 = new RawTripTestPayload(recordStr2);
    HoodieRecord record2 = new HoodieAvroRecord(new HoodieKey(rowChange2.getRowKey(), rowChange2.getPartitionPath()), rowChange2);
    BloomFilter filter = BloomFilterFactory.createBloomFilter(10000, 0.0000001, -1, BloomFilterTypeCode.SIMPLE.name());
    filter.add(record2.getRecordKey());
    HoodieFlinkWriteableTestTable testTable = HoodieFlinkWriteableTestTable.of(metaClient, SCHEMA, filter);
    String fileId = testTable.addCommit("000").getFileIdWithInserts("2016/01/31", record1);
    assertTrue(filter.mightContain(record1.getRecordKey()));
    assertTrue(filter.mightContain(record2.getRecordKey()));
    // We do the tag
    List<HoodieRecord> records = asList(record1, record2);
    HoodieWriteConfig config = makeConfig(rangePruning, treeFiltering, bucketizedChecking);
    metaClient = HoodieTableMetaClient.reload(metaClient);
    HoodieTable table = HoodieFlinkTable.create(config, context, metaClient);
    HoodieBloomIndex bloomIndex = new HoodieBloomIndex(config, ListBasedHoodieBloomIndexHelper.getInstance());
    List<HoodieRecord> taggedRecords = tagLocation(bloomIndex, records, table);
    // Check results
    for (HoodieRecord record : taggedRecords) {
        if (record.getKey().equals("1eb5b87a-1feh-4edd-87b4-6ec96dc405a0")) {
            assertEquals(record.getCurrentLocation().getFileId(), fileId);
        } else if (record.getRecordKey().equals("2eb5b87b-1feu-4edd-87b4-6ec96dc405a0")) {
            assertFalse(record.isCurrentLocationKnown());
        }
    }
}

Example 5 with HoodieFlinkWriteableTestTable

use of org.apache.hudi.testutils.HoodieFlinkWriteableTestTable in project hudi by apache.

the class TestFlinkHoodieBloomIndex method testTagLocation.

@ParameterizedTest(name = TEST_NAME_WITH_PARAMS)
@MethodSource("configParams")
public void testTagLocation(boolean rangePruning, boolean treeFiltering, boolean bucketizedChecking) throws Exception {
    // We have some records to be tagged (two different partitions)
    String rowKey1 = randomUUID().toString();
    String rowKey2 = randomUUID().toString();
    String rowKey3 = randomUUID().toString();
    String recordStr1 = "{\"_row_key\":\"" + rowKey1 + "\",\"time\":\"2016-01-31T03:16:41.415Z\",\"number\":12}";
    String recordStr2 = "{\"_row_key\":\"" + rowKey2 + "\",\"time\":\"2016-01-31T03:20:41.415Z\",\"number\":100}";
    String recordStr3 = "{\"_row_key\":\"" + rowKey3 + "\",\"time\":\"2016-01-31T03:16:41.415Z\",\"number\":15}";
    // place same row key under a different partition.
    String recordStr4 = "{\"_row_key\":\"" + rowKey1 + "\",\"time\":\"2015-01-31T03:16:41.415Z\",\"number\":32}";
    RawTripTestPayload rowChange1 = new RawTripTestPayload(recordStr1);
    HoodieRecord record1 = new HoodieAvroRecord(new HoodieKey(rowChange1.getRowKey(), rowChange1.getPartitionPath()), rowChange1);
    RawTripTestPayload rowChange2 = new RawTripTestPayload(recordStr2);
    HoodieRecord record2 = new HoodieAvroRecord(new HoodieKey(rowChange2.getRowKey(), rowChange2.getPartitionPath()), rowChange2);
    RawTripTestPayload rowChange3 = new RawTripTestPayload(recordStr3);
    HoodieRecord record3 = new HoodieAvroRecord(new HoodieKey(rowChange3.getRowKey(), rowChange3.getPartitionPath()), rowChange3);
    RawTripTestPayload rowChange4 = new RawTripTestPayload(recordStr4);
    HoodieRecord record4 = new HoodieAvroRecord(new HoodieKey(rowChange4.getRowKey(), rowChange4.getPartitionPath()), rowChange4);
    List<HoodieRecord> records = asList(record1, record2, record3, record4);
    // Also create the metadata and config
    HoodieWriteConfig config = makeConfig(rangePruning, treeFiltering, bucketizedChecking);
    HoodieFlinkTable hoodieTable = HoodieFlinkTable.create(config, context, metaClient);
    HoodieFlinkWriteableTestTable testTable = HoodieFlinkWriteableTestTable.of(hoodieTable, SCHEMA);
    // Let's tag
    HoodieBloomIndex bloomIndex = new HoodieBloomIndex(config, ListBasedHoodieBloomIndexHelper.getInstance());
    List<HoodieRecord> taggedRecords = tagLocation(bloomIndex, records, hoodieTable);
    // Should not find any files
    for (HoodieRecord record : taggedRecords) {
        assertFalse(record.isCurrentLocationKnown());
    }
    // We create three base file, each having one record. (two different partitions)
    String fileId1 = testTable.addCommit("001").getFileIdWithInserts("2016/01/31", record1);
    String fileId2 = testTable.addCommit("002").getFileIdWithInserts("2016/01/31", record2);
    String fileId3 = testTable.addCommit("003").getFileIdWithInserts("2015/01/31", record4);
    metaClient.reloadActiveTimeline();
    // We do the tag again
    taggedRecords = tagLocation(bloomIndex, records, HoodieFlinkTable.create(config, context, metaClient));
    // Check results
    for (HoodieRecord record : taggedRecords) {
        if (record.getRecordKey().equals(rowKey1)) {
            if (record.getPartitionPath().equals("2015/01/31")) {
                assertEquals(record.getCurrentLocation().getFileId(), fileId3);
            } else {
                assertEquals(record.getCurrentLocation().getFileId(), fileId1);
            }
        } else if (record.getRecordKey().equals(rowKey2)) {
            assertEquals(record.getCurrentLocation().getFileId(), fileId2);
        } else if (record.getRecordKey().equals(rowKey3)) {
            assertFalse(record.isCurrentLocationKnown());
        }
    }
}