Examples with Reader org.apache.hudi.common.table.log.HoodieLogFormat.Reader used on opensource projects

Example 11 with Reader

use of org.apache.hudi.common.table.log.HoodieLogFormat.Reader in project hudi by apache.

the class TestHoodieLogFormat method testDataBlockFormatAppendAndReadWithProjectedSchema.

@ParameterizedTest
@EnumSource(names = { "AVRO_DATA_BLOCK", "HFILE_DATA_BLOCK", "PARQUET_DATA_BLOCK" })
public void testDataBlockFormatAppendAndReadWithProjectedSchema(HoodieLogBlockType dataBlockType) throws IOException, URISyntaxException, InterruptedException {
    Writer writer = HoodieLogFormat.newWriterBuilder().onParentPath(partitionPath).withFileExtension(HoodieLogFile.DELTA_EXTENSION).withFileId("test-fileid1").overBaseCommit("100").withFs(fs).build();
    List<GenericRecord> records = SchemaTestUtil.generateTestGenericRecords(0, 1000);
    Schema schema = getSimpleSchema();
    Map<HoodieLogBlock.HeaderMetadataType, String> header = new HashMap<HoodieLogBlock.HeaderMetadataType, String>() {

        {
            put(HoodieLogBlock.HeaderMetadataType.INSTANT_TIME, "100");
            put(HoodieLogBlock.HeaderMetadataType.SCHEMA, schema.toString());
        }
    };
    // Init Benchmark to report number of bytes actually read from the Block
    BenchmarkCounter.initCounterFromReporter(HadoopMapRedUtils.createTestReporter(), fs.getConf());
    // NOTE: Have to use this ugly hack since List generic is not covariant in its type param
    HoodieDataBlock dataBlock = getDataBlock(dataBlockType, (List<IndexedRecord>) (List) records, header);
    writer.appendBlock(dataBlock);
    writer.close();
    Schema projectedSchema = HoodieAvroUtils.generateProjectionSchema(schema, Collections.singletonList("name"));
    List<GenericRecord> projectedRecords = HoodieAvroUtils.rewriteRecords(records, projectedSchema);
    try (Reader reader = HoodieLogFormat.newReader(fs, writer.getLogFile(), projectedSchema, true, false)) {
        assertTrue(reader.hasNext(), "First block should be available");
        HoodieLogBlock nextBlock = reader.next();
        HoodieDataBlock dataBlockRead = (HoodieDataBlock) nextBlock;
        Map<HoodieLogBlockType, Integer> expectedReadBytes = new HashMap<HoodieLogBlockType, Integer>() {

            {
                // not supported
                put(HoodieLogBlockType.AVRO_DATA_BLOCK, 0);
                // not supported
                put(HoodieLogBlockType.HFILE_DATA_BLOCK, 0);
                put(HoodieLogBlockType.PARQUET_DATA_BLOCK, 2605);
            }
        };
        List<IndexedRecord> recordsRead = getRecords(dataBlockRead);
        assertEquals(projectedRecords.size(), recordsRead.size(), "Read records size should be equal to the written records size");
        assertEquals(projectedRecords, recordsRead, "Both records lists should be the same. (ordering guaranteed)");
        assertEquals(dataBlockRead.getSchema(), projectedSchema);
        int bytesRead = (int) BenchmarkCounter.getBytesRead();
        assertEquals(expectedReadBytes.get(dataBlockType), bytesRead, "Read bytes have to match");
    }
}

Example 12 with Reader

use of org.apache.hudi.common.table.log.HoodieLogFormat.Reader in project hudi by apache.

the class TestHoodieLogFormat method testBasicWriteAndScan.

@Test
public void testBasicWriteAndScan() throws IOException, URISyntaxException, InterruptedException {
    Writer writer = HoodieLogFormat.newWriterBuilder().onParentPath(partitionPath).withFileExtension(HoodieLogFile.DELTA_EXTENSION).withFileId("test-fileid1").overBaseCommit("100").withFs(fs).build();
    Schema schema = getSimpleSchema();
    List<IndexedRecord> records = SchemaTestUtil.generateTestRecords(0, 100);
    List<IndexedRecord> copyOfRecords = records.stream().map(record -> HoodieAvroUtils.rewriteRecord((GenericRecord) record, schema)).collect(Collectors.toList());
    Map<HoodieLogBlock.HeaderMetadataType, String> header = new HashMap<>();
    header.put(HoodieLogBlock.HeaderMetadataType.INSTANT_TIME, "100");
    header.put(HoodieLogBlock.HeaderMetadataType.SCHEMA, getSimpleSchema().toString());
    HoodieDataBlock dataBlock = getDataBlock(DEFAULT_DATA_BLOCK_TYPE, records, header);
    writer.appendBlock(dataBlock);
    writer.close();
    Reader reader = HoodieLogFormat.newReader(fs, writer.getLogFile(), SchemaTestUtil.getSimpleSchema());
    assertTrue(reader.hasNext(), "We wrote a block, we should be able to read it");
    HoodieLogBlock nextBlock = reader.next();
    assertEquals(DEFAULT_DATA_BLOCK_TYPE, nextBlock.getBlockType(), "The next block should be a data block");
    HoodieDataBlock dataBlockRead = (HoodieDataBlock) nextBlock;
    List<IndexedRecord> recordsRead = getRecords(dataBlockRead);
    assertEquals(copyOfRecords.size(), recordsRead.size(), "Read records size should be equal to the written records size");
    assertEquals(copyOfRecords, recordsRead, "Both records lists should be the same. (ordering guaranteed)");
    reader.close();
}

Example 13 with Reader

use of org.apache.hudi.common.table.log.HoodieLogFormat.Reader in project hudi by apache.

the class ExportCommand method copyArchivedInstants.

private int copyArchivedInstants(List<FileStatus> statuses, Set<String> actionSet, int limit, String localFolder) throws Exception {
    int copyCount = 0;
    for (FileStatus fs : statuses) {
        // read the archived file
        Reader reader = HoodieLogFormat.newReader(FSUtils.getFs(HoodieCLI.getTableMetaClient().getBasePath(), HoodieCLI.conf), new HoodieLogFile(fs.getPath()), HoodieArchivedMetaEntry.getClassSchema());
        // read the avro blocks
        while (reader.hasNext() && copyCount < limit) {
            HoodieAvroDataBlock blk = (HoodieAvroDataBlock) reader.next();
            try (ClosableIterator<IndexedRecord> recordItr = blk.getRecordItr()) {
                while (recordItr.hasNext()) {
                    IndexedRecord ir = recordItr.next();
                    // Archived instants are saved as arvo encoded HoodieArchivedMetaEntry records. We need to get the
                    // metadata record from the entry and convert it to json.
                    HoodieArchivedMetaEntry archiveEntryRecord = (HoodieArchivedMetaEntry) SpecificData.get().deepCopy(HoodieArchivedMetaEntry.SCHEMA$, ir);
                    final String action = archiveEntryRecord.get("actionType").toString();
                    if (!actionSet.contains(action)) {
                        continue;
                    }
                    GenericRecord metadata = null;
                    switch(action) {
                        case HoodieTimeline.CLEAN_ACTION:
                            metadata = archiveEntryRecord.getHoodieCleanMetadata();
                            break;
                        case HoodieTimeline.COMMIT_ACTION:
                        case HoodieTimeline.DELTA_COMMIT_ACTION:
                            metadata = archiveEntryRecord.getHoodieCommitMetadata();
                            break;
                        case HoodieTimeline.ROLLBACK_ACTION:
                            metadata = archiveEntryRecord.getHoodieRollbackMetadata();
                            break;
                        case HoodieTimeline.SAVEPOINT_ACTION:
                            metadata = archiveEntryRecord.getHoodieSavePointMetadata();
                            break;
                        case HoodieTimeline.COMPACTION_ACTION:
                            metadata = archiveEntryRecord.getHoodieCompactionMetadata();
                            break;
                        default:
                            throw new HoodieException("Unknown type of action " + action);
                    }
                    final String instantTime = archiveEntryRecord.get("commitTime").toString();
                    final String outPath = localFolder + Path.SEPARATOR + instantTime + "." + action;
                    writeToFile(outPath, HoodieAvroUtils.avroToJson(metadata, true));
                    if (++copyCount == limit) {
                        break;
                    }
                }
            }
        }
        reader.close();
    }
    return copyCount;
}

Example 14 with Reader

use of org.apache.hudi.common.table.log.HoodieLogFormat.Reader in project hudi by apache.

the class HoodieLogFileCommand method showLogFileRecords.

@CliCommand(value = "show logfile records", help = "Read records from log files")
public String showLogFileRecords(@CliOption(key = { "limit" }, help = "Limit commits", unspecifiedDefaultValue = "10") final Integer limit, @CliOption(key = "logFilePathPattern", mandatory = true, help = "Fully qualified paths for the log files") final String logFilePathPattern, @CliOption(key = "mergeRecords", help = "If the records in the log files should be merged", unspecifiedDefaultValue = "false") final Boolean shouldMerge) throws IOException {
    System.out.println("===============> Showing only " + limit + " records <===============");
    HoodieTableMetaClient client = HoodieCLI.getTableMetaClient();
    FileSystem fs = client.getFs();
    List<String> logFilePaths = FSUtils.getGlobStatusExcludingMetaFolder(fs, new Path(logFilePathPattern)).stream().map(status -> status.getPath().toString()).sorted(Comparator.reverseOrder()).collect(Collectors.toList());
    // logFilePaths size must > 1
    assert logFilePaths.size() > 0 : "There is no log file";
    // TODO : readerSchema can change across blocks/log files, fix this inside Scanner
    AvroSchemaConverter converter = new AvroSchemaConverter();
    // get schema from last log file
    Schema readerSchema = converter.convert(Objects.requireNonNull(TableSchemaResolver.readSchemaFromLogFile(fs, new Path(logFilePaths.get(logFilePaths.size() - 1)))));
    List<IndexedRecord> allRecords = new ArrayList<>();
    if (shouldMerge) {
        System.out.println("===========================> MERGING RECORDS <===================");
        HoodieMergedLogRecordScanner scanner = HoodieMergedLogRecordScanner.newBuilder().withFileSystem(fs).withBasePath(client.getBasePath()).withLogFilePaths(logFilePaths).withReaderSchema(readerSchema).withLatestInstantTime(client.getActiveTimeline().getCommitTimeline().lastInstant().get().getTimestamp()).withReadBlocksLazily(Boolean.parseBoolean(HoodieCompactionConfig.COMPACTION_LAZY_BLOCK_READ_ENABLE.defaultValue())).withReverseReader(Boolean.parseBoolean(HoodieCompactionConfig.COMPACTION_REVERSE_LOG_READ_ENABLE.defaultValue())).withBufferSize(HoodieMemoryConfig.MAX_DFS_STREAM_BUFFER_SIZE.defaultValue()).withMaxMemorySizeInBytes(HoodieMemoryConfig.DEFAULT_MAX_MEMORY_FOR_SPILLABLE_MAP_IN_BYTES).withSpillableMapBasePath(HoodieMemoryConfig.SPILLABLE_MAP_BASE_PATH.defaultValue()).withDiskMapType(HoodieCommonConfig.SPILLABLE_DISK_MAP_TYPE.defaultValue()).withBitCaskDiskMapCompressionEnabled(HoodieCommonConfig.DISK_MAP_BITCASK_COMPRESSION_ENABLED.defaultValue()).build();
        for (HoodieRecord<? extends HoodieRecordPayload> hoodieRecord : scanner) {
            Option<IndexedRecord> record = hoodieRecord.getData().getInsertValue(readerSchema);
            if (allRecords.size() < limit) {
                allRecords.add(record.get());
            }
        }
    } else {
        for (String logFile : logFilePaths) {
            Schema writerSchema = new AvroSchemaConverter().convert(Objects.requireNonNull(TableSchemaResolver.readSchemaFromLogFile(client.getFs(), new Path(logFile))));
            HoodieLogFormat.Reader reader = HoodieLogFormat.newReader(fs, new HoodieLogFile(new Path(logFile)), writerSchema);
            // read the avro blocks
            while (reader.hasNext()) {
                HoodieLogBlock n = reader.next();
                if (n instanceof HoodieDataBlock) {
                    HoodieDataBlock blk = (HoodieDataBlock) n;
                    try (ClosableIterator<IndexedRecord> recordItr = blk.getRecordItr()) {
                        recordItr.forEachRemaining(record -> {
                            if (allRecords.size() < limit) {
                                allRecords.add(record);
                            }
                        });
                    }
                }
            }
            reader.close();
            if (allRecords.size() >= limit) {
                break;
            }
        }
    }
    String[][] rows = new String[allRecords.size()][];
    int i = 0;
    for (IndexedRecord record : allRecords) {
        String[] data = new String[1];
        data[0] = record.toString();
        rows[i] = data;
        i++;
    }
    return HoodiePrintHelper.print(new String[] { HoodieTableHeaderFields.HEADER_RECORDS }, rows);
}

Example 15 with Reader

use of org.apache.hudi.common.table.log.HoodieLogFormat.Reader in project hudi by apache.

the class HoodieLogFileCommand method showLogFileCommits.

@CliCommand(value = "show logfile metadata", help = "Read commit metadata from log files")
public String showLogFileCommits(@CliOption(key = "logFilePathPattern", mandatory = true, help = "Fully qualified path for the log file") final String logFilePathPattern, @CliOption(key = { "limit" }, help = "Limit commits", unspecifiedDefaultValue = "-1") final Integer limit, @CliOption(key = { "sortBy" }, help = "Sorting Field", unspecifiedDefaultValue = "") final String sortByField, @CliOption(key = { "desc" }, help = "Ordering", unspecifiedDefaultValue = "false") final boolean descending, @CliOption(key = { "headeronly" }, help = "Print Header Only", unspecifiedDefaultValue = "false") final boolean headerOnly) throws IOException {
    FileSystem fs = HoodieCLI.getTableMetaClient().getFs();
    List<String> logFilePaths = FSUtils.getGlobStatusExcludingMetaFolder(fs, new Path(logFilePathPattern)).stream().map(status -> status.getPath().toString()).collect(Collectors.toList());
    Map<String, List<Tuple3<HoodieLogBlockType, Tuple2<Map<HeaderMetadataType, String>, Map<HeaderMetadataType, String>>, Integer>>> commitCountAndMetadata = new HashMap<>();
    int numCorruptBlocks = 0;
    int dummyInstantTimeCount = 0;
    for (String logFilePath : logFilePaths) {
        FileStatus[] fsStatus = fs.listStatus(new Path(logFilePath));
        Schema writerSchema = new AvroSchemaConverter().convert(Objects.requireNonNull(TableSchemaResolver.readSchemaFromLogFile(fs, new Path(logFilePath))));
        Reader reader = HoodieLogFormat.newReader(fs, new HoodieLogFile(fsStatus[0].getPath()), writerSchema);
        // read the avro blocks
        while (reader.hasNext()) {
            HoodieLogBlock n = reader.next();
            String instantTime;
            AtomicInteger recordCount = new AtomicInteger(0);
            if (n instanceof HoodieCorruptBlock) {
                try {
                    instantTime = n.getLogBlockHeader().get(HeaderMetadataType.INSTANT_TIME);
                    if (instantTime == null) {
                        throw new Exception("Invalid instant time " + instantTime);
                    }
                } catch (Exception e) {
                    numCorruptBlocks++;
                    instantTime = "corrupt_block_" + numCorruptBlocks;
                // could not read metadata for corrupt block
                }
            } else {
                instantTime = n.getLogBlockHeader().get(HeaderMetadataType.INSTANT_TIME);
                if (instantTime == null) {
                    // This can happen when reading archived commit files since they were written without any instant time
                    dummyInstantTimeCount++;
                    instantTime = "dummy_instant_time_" + dummyInstantTimeCount;
                }
                if (n instanceof HoodieDataBlock) {
                    try (ClosableIterator<IndexedRecord> recordItr = ((HoodieDataBlock) n).getRecordItr()) {
                        recordItr.forEachRemaining(r -> recordCount.incrementAndGet());
                    }
                }
            }
            if (commitCountAndMetadata.containsKey(instantTime)) {
                commitCountAndMetadata.get(instantTime).add(new Tuple3<>(n.getBlockType(), new Tuple2<>(n.getLogBlockHeader(), n.getLogBlockFooter()), recordCount.get()));
            } else {
                List<Tuple3<HoodieLogBlockType, Tuple2<Map<HeaderMetadataType, String>, Map<HeaderMetadataType, String>>, Integer>> list = new ArrayList<>();
                list.add(new Tuple3<>(n.getBlockType(), new Tuple2<>(n.getLogBlockHeader(), n.getLogBlockFooter()), recordCount.get()));
                commitCountAndMetadata.put(instantTime, list);
            }
        }
        reader.close();
    }
    List<Comparable[]> rows = new ArrayList<>();
    ObjectMapper objectMapper = new ObjectMapper();
    for (Map.Entry<String, List<Tuple3<HoodieLogBlockType, Tuple2<Map<HeaderMetadataType, String>, Map<HeaderMetadataType, String>>, Integer>>> entry : commitCountAndMetadata.entrySet()) {
        String instantTime = entry.getKey();
        for (Tuple3<HoodieLogBlockType, Tuple2<Map<HeaderMetadataType, String>, Map<HeaderMetadataType, String>>, Integer> tuple3 : entry.getValue()) {
            Comparable[] output = new Comparable[5];
            output[0] = instantTime;
            output[1] = tuple3._3();
            output[2] = tuple3._1().toString();
            output[3] = objectMapper.writeValueAsString(tuple3._2()._1());
            output[4] = objectMapper.writeValueAsString(tuple3._2()._2());
            rows.add(output);
        }
    }
    TableHeader header = new TableHeader().addTableHeaderField(HoodieTableHeaderFields.HEADER_INSTANT_TIME).addTableHeaderField(HoodieTableHeaderFields.HEADER_RECORD_COUNT).addTableHeaderField(HoodieTableHeaderFields.HEADER_BLOCK_TYPE).addTableHeaderField(HoodieTableHeaderFields.HEADER_HEADER_METADATA).addTableHeaderField(HoodieTableHeaderFields.HEADER_FOOTER_METADATA);
    return HoodiePrintHelper.print(header, new HashMap<>(), sortByField, descending, limit, headerOnly, rows);
}