Examples with ColumnValueGenerator org.apache.druid.segment.generator.ColumnValueGenerator used on opensource projects

Example 1 with ColumnValueGenerator

use of org.apache.druid.segment.generator.ColumnValueGenerator in project druid by druid-io.

the class FloatCompressionBenchmarkFileGenerator method main.

public static void main(String[] args) throws IOException {
    if (args.length >= 1) {
        dirPath = args[0];
    }
    GeneratorColumnSchema enumeratedSchema = GeneratorColumnSchema.makeEnumerated("", ValueType.FLOAT, true, 1, 0d, ImmutableList.of(0f, 1.1f, 2.2f, 3.3f, 4.4f), ImmutableList.of(0.95, 0.001, 0.0189, 0.03, 0.0001));
    GeneratorColumnSchema zipfLowSchema = GeneratorColumnSchema.makeZipf("", ValueType.FLOAT, true, 1, 0d, -1, 1000, 1d);
    GeneratorColumnSchema zipfHighSchema = GeneratorColumnSchema.makeZipf("", ValueType.FLOAT, true, 1, 0d, -1, 1000, 3d);
    GeneratorColumnSchema sequentialSchema = GeneratorColumnSchema.makeSequential("", ValueType.FLOAT, true, 1, 0d, 1470187671, 2000000000);
    GeneratorColumnSchema uniformSchema = GeneratorColumnSchema.makeContinuousUniform("", ValueType.FLOAT, true, 1, 0d, 0, 1000);
    Map<String, ColumnValueGenerator> generators = new HashMap<>();
    generators.put("enumerate", new ColumnValueGenerator(enumeratedSchema, 1));
    generators.put("zipfLow", new ColumnValueGenerator(zipfLowSchema, 1));
    generators.put("zipfHigh", new ColumnValueGenerator(zipfHighSchema, 1));
    generators.put("sequential", new ColumnValueGenerator(sequentialSchema, 1));
    generators.put("uniform", new ColumnValueGenerator(uniformSchema, 1));
    File dir = new File(dirPath);
    dir.mkdir();
    // create data files using BenchmarkColunValueGenerator
    for (Map.Entry<String, ColumnValueGenerator> entry : generators.entrySet()) {
        final File dataFile = new File(dir, entry.getKey());
        dataFile.delete();
        try (Writer writer = Files.newBufferedWriter(dataFile.toPath(), StandardCharsets.UTF_8)) {
            for (int i = 0; i < ROW_NUM; i++) {
                writer.write((Float) entry.getValue().generateRowValue() + "\n");
            }
        }
    }
    // create compressed files using all combinations of CompressionStrategy and FloatEncoding provided
    for (Map.Entry<String, ColumnValueGenerator> entry : generators.entrySet()) {
        for (CompressionStrategy compression : COMPRESSIONS) {
            String name = entry.getKey() + "-" + compression;
            log.info("%s: ", name);
            File compFile = new File(dir, name);
            compFile.delete();
            File dataFile = new File(dir, entry.getKey());
            ColumnarFloatsSerializer writer = CompressionFactory.getFloatSerializer("float-benchmark", new OffHeapMemorySegmentWriteOutMedium(), "float", ByteOrder.nativeOrder(), compression);
            try (BufferedReader br = Files.newBufferedReader(dataFile.toPath(), StandardCharsets.UTF_8);
                FileChannel output = FileChannel.open(compFile.toPath(), StandardOpenOption.CREATE_NEW, StandardOpenOption.WRITE)) {
                writer.open();
                String line;
                while ((line = br.readLine()) != null) {
                    writer.add(Float.parseFloat(line));
                }
                writer.writeTo(output, null);
            }
            log.info("%d", compFile.length() / 1024);
        }
    }
}

Example 2 with ColumnValueGenerator

use of org.apache.druid.segment.generator.ColumnValueGenerator in project druid by druid-io.

the class LongCompressionBenchmarkFileGenerator method main.

public static void main(String[] args) throws IOException {
    if (args.length >= 1) {
        dirPath = args[0];
    }
    GeneratorColumnSchema enumeratedSchema = GeneratorColumnSchema.makeEnumerated("", ValueType.LONG, true, 1, 0d, ImmutableList.of(0, 1, 2, 3, 4), ImmutableList.of(0.95, 0.001, 0.0189, 0.03, 0.0001));
    GeneratorColumnSchema zipfLowSchema = GeneratorColumnSchema.makeZipf("", ValueType.LONG, true, 1, 0d, -1, 1000, 1d);
    GeneratorColumnSchema zipfHighSchema = GeneratorColumnSchema.makeZipf("", ValueType.LONG, true, 1, 0d, -1, 1000, 3d);
    GeneratorColumnSchema sequentialSchema = GeneratorColumnSchema.makeSequential("", ValueType.LONG, true, 1, 0d, 1470187671, 2000000000);
    GeneratorColumnSchema uniformSchema = GeneratorColumnSchema.makeDiscreteUniform("", ValueType.LONG, true, 1, 0d, 0, 1000);
    Map<String, ColumnValueGenerator> generators = new HashMap<>();
    generators.put("enumerate", new ColumnValueGenerator(enumeratedSchema, 1));
    generators.put("zipfLow", new ColumnValueGenerator(zipfLowSchema, 1));
    generators.put("zipfHigh", new ColumnValueGenerator(zipfHighSchema, 1));
    generators.put("sequential", new ColumnValueGenerator(sequentialSchema, 1));
    generators.put("uniform", new ColumnValueGenerator(uniformSchema, 1));
    File dir = new File(dirPath);
    dir.mkdir();
    // create data files using BenchmarkColunValueGenerator
    for (Map.Entry<String, ColumnValueGenerator> entry : generators.entrySet()) {
        final File dataFile = new File(dir, entry.getKey());
        dataFile.delete();
        try (Writer writer = Files.newBufferedWriter(dataFile.toPath(), StandardCharsets.UTF_8)) {
            for (int i = 0; i < ROW_NUM; i++) {
                writer.write((long) entry.getValue().generateRowValue() + "\n");
            }
        }
    }
    // create compressed files using all combinations of CompressionStrategy and LongEncoding provided
    for (Map.Entry<String, ColumnValueGenerator> entry : generators.entrySet()) {
        for (CompressionStrategy compression : COMPRESSIONS) {
            for (CompressionFactory.LongEncodingStrategy encoding : ENCODINGS) {
                String name = entry.getKey() + "-" + compression + "-" + encoding;
                log.info("%s: ", name);
                File compFile = new File(dir, name);
                compFile.delete();
                File dataFile = new File(dir, entry.getKey());
                ColumnarLongsSerializer writer = CompressionFactory.getLongSerializer("long-benchmark", new OffHeapMemorySegmentWriteOutMedium(), "long", ByteOrder.nativeOrder(), encoding, compression);
                try (BufferedReader br = Files.newBufferedReader(dataFile.toPath(), StandardCharsets.UTF_8);
                    FileChannel output = FileChannel.open(compFile.toPath(), StandardOpenOption.CREATE_NEW, StandardOpenOption.WRITE)) {
                    writer.open();
                    String line;
                    while ((line = br.readLine()) != null) {
                        writer.add(Long.parseLong(line));
                    }
                    writer.writeTo(output, null);
                }
                log.info("%d", compFile.length() / 1024);
            }
        }
    }
}

Example 3 with ColumnValueGenerator

use of org.apache.druid.segment.generator.ColumnValueGenerator in project druid by druid-io.

the class ColumnarLongsEncodeDataFromGeneratorBenchmark method setup.

@Setup
public void setup() throws Exception {
    vals = new long[rows];
    final String filename = getGeneratorValueFilename(distribution, rows, zeroProbability);
    File dir = getTmpDir();
    File dataFile = new File(dir, filename);
    if (dataFile.exists()) {
        System.out.println("Data files already exist, re-using");
        try (BufferedReader br = Files.newBufferedReader(dataFile.toPath(), StandardCharsets.UTF_8)) {
            int lineNum = 0;
            String line;
            while ((line = br.readLine()) != null) {
                vals[lineNum] = Long.parseLong(line);
                if (vals[lineNum] < minValue) {
                    minValue = vals[lineNum];
                }
                if (vals[lineNum] > maxValue) {
                    maxValue = vals[lineNum];
                }
                lineNum++;
            }
        }
    } else {
        try (Writer writer = Files.newBufferedWriter(dataFile.toPath(), StandardCharsets.UTF_8)) {
            ColumnValueGenerator valueGenerator = makeGenerator(distribution, rows, zeroProbability);
            for (int i = 0; i < rows; i++) {
                long value;
                Object rowValue = valueGenerator.generateRowValue();
                value = rowValue != null ? (long) rowValue : 0;
                vals[i] = value;
                if (vals[i] < minValue) {
                    minValue = vals[i];
                }
                if (vals[i] > maxValue) {
                    maxValue = vals[i];
                }
                writer.write(vals[i] + "\n");
            }
        }
    }
}