Examples with Encoding org.apache.parquet.column.Encoding used on opensource projects

Example 11 with Encoding

use of org.apache.parquet.column.Encoding in project parquet-mr by apache.

the class DictionaryFilter method hasNonDictionaryPages.

@SuppressWarnings("deprecation")
private static boolean hasNonDictionaryPages(ColumnChunkMetaData meta) {
    EncodingStats stats = meta.getEncodingStats();
    if (stats != null) {
        return stats.hasNonDictionaryEncodedPages();
    }
    // without EncodingStats, fall back to testing the encoding list
    Set<Encoding> encodings = new HashSet<Encoding>(meta.getEncodings());
    if (encodings.remove(Encoding.PLAIN_DICTIONARY)) {
        // if remove returned true, PLAIN_DICTIONARY was present, which means at
        // least one page was dictionary encoded and 1.0 encodings are used
        // RLE and BIT_PACKED are only used for repetition or definition levels
        encodings.remove(Encoding.RLE);
        encodings.remove(Encoding.BIT_PACKED);
        if (encodings.isEmpty()) {
            // no encodings other than dictionary or rep/def levels
            return false;
        }
        return true;
    } else {
        // page encoding stats
        return true;
    }
}

Example 12 with Encoding

use of org.apache.parquet.column.Encoding in project parquet-mr by apache.

the class TestInputFormat method makeBlockFromStats.

public static BlockMetaData makeBlockFromStats(IntStatistics stats, long valueCount) {
    BlockMetaData blockMetaData = new BlockMetaData();
    ColumnChunkMetaData column = ColumnChunkMetaData.get(ColumnPath.get("foo"), PrimitiveTypeName.INT32, CompressionCodecName.GZIP, new HashSet<Encoding>(Arrays.asList(Encoding.PLAIN)), stats, 100l, 100l, valueCount, 100l, 100l);
    blockMetaData.addColumn(column);
    blockMetaData.setTotalByteSize(200l);
    blockMetaData.setRowCount(valueCount);
    return blockMetaData;
}

Example 13 with Encoding

use of org.apache.parquet.column.Encoding in project parquet-mr by apache.

the class TestParquetFileWriter method testAlignmentWithPadding.

@Test
public void testAlignmentWithPadding() throws Exception {
    File testFile = temp.newFile();
    Path path = new Path(testFile.toURI());
    Configuration conf = new Configuration();
    // uses the test constructor
    ParquetFileWriter w = new ParquetFileWriter(conf, SCHEMA, path, 120, 60);
    w.start();
    w.startBlock(3);
    w.startColumn(C1, 5, CODEC);
    long c1Starts = w.getPos();
    w.writeDataPage(2, 4, BytesInput.from(BYTES1), EMPTY_STATS, BIT_PACKED, BIT_PACKED, PLAIN);
    w.writeDataPage(3, 4, BytesInput.from(BYTES1), EMPTY_STATS, BIT_PACKED, BIT_PACKED, PLAIN);
    w.endColumn();
    long c1Ends = w.getPos();
    w.startColumn(C2, 6, CODEC);
    long c2Starts = w.getPos();
    w.writeDataPage(2, 4, BytesInput.from(BYTES2), EMPTY_STATS, BIT_PACKED, BIT_PACKED, PLAIN);
    w.writeDataPage(3, 4, BytesInput.from(BYTES2), EMPTY_STATS, BIT_PACKED, BIT_PACKED, PLAIN);
    w.writeDataPage(1, 4, BytesInput.from(BYTES2), EMPTY_STATS, BIT_PACKED, BIT_PACKED, PLAIN);
    w.endColumn();
    long c2Ends = w.getPos();
    w.endBlock();
    // should be 109
    long firstRowGroupEnds = w.getPos();
    w.startBlock(4);
    w.startColumn(C1, 7, CODEC);
    w.writeDataPage(7, 4, BytesInput.from(BYTES3), EMPTY_STATS, BIT_PACKED, BIT_PACKED, PLAIN);
    w.endColumn();
    w.startColumn(C2, 8, CODEC);
    w.writeDataPage(8, 4, BytesInput.from(BYTES4), EMPTY_STATS, BIT_PACKED, BIT_PACKED, PLAIN);
    w.endColumn();
    w.endBlock();
    long secondRowGroupEnds = w.getPos();
    w.end(new HashMap<String, String>());
    FileSystem fs = path.getFileSystem(conf);
    long fileLen = fs.getFileStatus(path).getLen();
    FSDataInputStream data = fs.open(path);
    // 4-byte offset + "PAR1"
    data.seek(fileLen - 8);
    long footerLen = BytesUtils.readIntLittleEndian(data);
    long startFooter = fileLen - footerLen - 8;
    assertEquals("Footer should start after second row group without padding", secondRowGroupEnds, startFooter);
    ParquetMetadata readFooter = ParquetFileReader.readFooter(conf, path);
    assertEquals("footer: " + readFooter, 2, readFooter.getBlocks().size());
    assertEquals(c1Ends - c1Starts, readFooter.getBlocks().get(0).getColumns().get(0).getTotalSize());
    assertEquals(c2Ends - c2Starts, readFooter.getBlocks().get(0).getColumns().get(1).getTotalSize());
    assertEquals(c2Ends - c1Starts, readFooter.getBlocks().get(0).getTotalByteSize());
    HashSet<Encoding> expectedEncoding = new HashSet<Encoding>();
    expectedEncoding.add(PLAIN);
    expectedEncoding.add(BIT_PACKED);
    assertEquals(expectedEncoding, readFooter.getBlocks().get(0).getColumns().get(0).getEncodings());
    // verify block starting positions with padding
    assertEquals("First row group should start after magic", 4, readFooter.getBlocks().get(0).getStartingPos());
    assertTrue("First row group should end before the block size (120)", firstRowGroupEnds < 120);
    assertEquals("Second row group should start at the block size", 120, readFooter.getBlocks().get(1).getStartingPos());
    {
        // read first block of col #1
        ParquetFileReader r = new ParquetFileReader(conf, readFooter.getFileMetaData(), path, Arrays.asList(readFooter.getBlocks().get(0)), Arrays.asList(SCHEMA.getColumnDescription(PATH1)));
        PageReadStore pages = r.readNextRowGroup();
        assertEquals(3, pages.getRowCount());
        validateContains(SCHEMA, pages, PATH1, 2, BytesInput.from(BYTES1));
        validateContains(SCHEMA, pages, PATH1, 3, BytesInput.from(BYTES1));
        assertNull(r.readNextRowGroup());
    }
    {
        // read all blocks of col #1 and #2
        ParquetFileReader r = new ParquetFileReader(conf, readFooter.getFileMetaData(), path, readFooter.getBlocks(), Arrays.asList(SCHEMA.getColumnDescription(PATH1), SCHEMA.getColumnDescription(PATH2)));
        PageReadStore pages = r.readNextRowGroup();
        assertEquals(3, pages.getRowCount());
        validateContains(SCHEMA, pages, PATH1, 2, BytesInput.from(BYTES1));
        validateContains(SCHEMA, pages, PATH1, 3, BytesInput.from(BYTES1));
        validateContains(SCHEMA, pages, PATH2, 2, BytesInput.from(BYTES2));
        validateContains(SCHEMA, pages, PATH2, 3, BytesInput.from(BYTES2));
        validateContains(SCHEMA, pages, PATH2, 1, BytesInput.from(BYTES2));
        pages = r.readNextRowGroup();
        assertEquals(4, pages.getRowCount());
        validateContains(SCHEMA, pages, PATH1, 7, BytesInput.from(BYTES3));
        validateContains(SCHEMA, pages, PATH2, 8, BytesInput.from(BYTES4));
        assertNull(r.readNextRowGroup());
    }
    PrintFooter.main(new String[] { path.toString() });
}

Example 14 with Encoding

use of org.apache.parquet.column.Encoding in project parquet-mr by apache.

the class TestParquetWriter method test.

@Test
public void test() throws Exception {
    Configuration conf = new Configuration();
    Path root = new Path("target/tests/TestParquetWriter/");
    enforceEmptyDir(conf, root);
    MessageType schema = parseMessageType("message test { " + "required binary binary_field; " + "required int32 int32_field; " + "required int64 int64_field; " + "required boolean boolean_field; " + "required float float_field; " + "required double double_field; " + "required fixed_len_byte_array(3) flba_field; " + "required int96 int96_field; " + "} ");
    GroupWriteSupport.setSchema(schema, conf);
    SimpleGroupFactory f = new SimpleGroupFactory(schema);
    Map<String, Encoding> expected = new HashMap<String, Encoding>();
    expected.put("10-" + PARQUET_1_0, PLAIN_DICTIONARY);
    expected.put("1000-" + PARQUET_1_0, PLAIN);
    expected.put("10-" + PARQUET_2_0, RLE_DICTIONARY);
    expected.put("1000-" + PARQUET_2_0, DELTA_BYTE_ARRAY);
    for (int modulo : asList(10, 1000)) {
        for (WriterVersion version : WriterVersion.values()) {
            Path file = new Path(root, version.name() + "_" + modulo);
            ParquetWriter<Group> writer = new ParquetWriter<Group>(file, new GroupWriteSupport(), UNCOMPRESSED, 1024, 1024, 512, true, false, version, conf);
            for (int i = 0; i < 1000; i++) {
                writer.write(f.newGroup().append("binary_field", "test" + (i % modulo)).append("int32_field", 32).append("int64_field", 64l).append("boolean_field", true).append("float_field", 1.0f).append("double_field", 2.0d).append("flba_field", "foo").append("int96_field", Binary.fromConstantByteArray(new byte[12])));
            }
            writer.close();
            ParquetReader<Group> reader = ParquetReader.builder(new GroupReadSupport(), file).withConf(conf).build();
            for (int i = 0; i < 1000; i++) {
                Group group = reader.read();
                assertEquals("test" + (i % modulo), group.getBinary("binary_field", 0).toStringUsingUTF8());
                assertEquals(32, group.getInteger("int32_field", 0));
                assertEquals(64l, group.getLong("int64_field", 0));
                assertEquals(true, group.getBoolean("boolean_field", 0));
                assertEquals(1.0f, group.getFloat("float_field", 0), 0.001);
                assertEquals(2.0d, group.getDouble("double_field", 0), 0.001);
                assertEquals("foo", group.getBinary("flba_field", 0).toStringUsingUTF8());
                assertEquals(Binary.fromConstantByteArray(new byte[12]), group.getInt96("int96_field", 0));
            }
            reader.close();
            ParquetMetadata footer = readFooter(conf, file, NO_FILTER);
            for (BlockMetaData blockMetaData : footer.getBlocks()) {
                for (ColumnChunkMetaData column : blockMetaData.getColumns()) {
                    if (column.getPath().toDotString().equals("binary_field")) {
                        String key = modulo + "-" + version;
                        Encoding expectedEncoding = expected.get(key);
                        assertTrue(key + ":" + column.getEncodings() + " should contain " + expectedEncoding, column.getEncodings().contains(expectedEncoding));
                    }
                }
            }
            assertEquals("Object model property should be example", "example", footer.getFileMetaData().getKeyValueMetaData().get(ParquetWriter.OBJECT_MODEL_NAME_PROP));
        }
    }
}

Example 15 with Encoding

use of org.apache.parquet.column.Encoding in project parquet-mr by apache.

the class TestColumnChunkMetaData method newMD.

private ColumnChunkMetaData newMD(long big) {
    Set<Encoding> e = new HashSet<Encoding>();
    PrimitiveTypeName t = BINARY;
    ColumnPath p = ColumnPath.get("foo");
    CompressionCodecName c = CompressionCodecName.GZIP;
    BinaryStatistics s = new BinaryStatistics();
    ColumnChunkMetaData md = ColumnChunkMetaData.get(p, t, c, e, s, big, 0, 0, 0, 0);
    return md;
}