Package parquet.schema

Examples of parquet.schema.MessageType

24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
import parquet.schema.Type;
import parquet.schema.Type.Repetition;

public class ParquetSchemaMerge {
  public static void main(String[] args) {
    MessageType message1;
    MessageType message2;

    PrimitiveType c = new PrimitiveType(Repetition.OPTIONAL, PrimitiveTypeName.INT32, "c");
    GroupType b = new GroupType(Repetition.REQUIRED, "b");
    GroupType a = new GroupType(Repetition.OPTIONAL, "a", b);
    message1 = new MessageType("root", a);

    PrimitiveType c2 = new PrimitiveType(Repetition.OPTIONAL, PrimitiveTypeName.INT32, "d");
    GroupType b2 = new GroupType(Repetition.OPTIONAL, "b", c2);
    GroupType a2 = new GroupType(Repetition.OPTIONAL, "a", b2);
    message2 = new MessageType("root", a2);

    MessageType message3 = message1.union(message2);

    StringBuilder builder = new StringBuilder();
    message3.writeToStringBuilder(builder, "");
    System.out.println(builder);
  }
View Full Code Here
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
    this.columns = columns;
    this.entry = entry;
  }

  public static MessageType getProjection(MessageType schema, List<SchemaPath> columns) {
    MessageType projection = null;
    for (SchemaPath path : columns) {
      List<String> segments = Lists.newArrayList();
      PathSegment rootSegment = path.getRootSegment();
      PathSegment seg = rootSegment;
      String messageName = schema.getName();
      while(seg != null){
        if(seg.isNamed()) {
          segments.add(seg.getNameSegment().getPath());
        }
        seg = seg.getChild();
      }
      String[] pathSegments = new String[segments.size()];
      segments.toArray(pathSegments);
      Type type = null;
      try {
        type = schema.getType(pathSegments);
      } catch (InvalidRecordException e) {
        logger.warn("Invalid record" , e);
      }
      if (type != null) {
        Type t = getType(pathSegments, 0, schema);
        if (projection == null) {
          projection = new MessageType(messageName, t);
        } else {
          projection = projection.union(new MessageType(messageName, t));
        }
      }
    }
    return projection;
  }
View Full Code Here
108
109
110
111
112
113
114
115
116
117
118
  @Override
  public void setup(OutputMutator output) throws ExecutionSetupException {

    try {
      schema = footer.getFileMetaData().getSchema();
      MessageType projection = null;

      if (columns == null || columns.size() == 0) {
        projection = schema;
      } else {
        projection = getProjection(schema, columns);
View Full Code Here
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
    // remove the last semicolon, java really needs a join method for strings...
    // TODO - nvm apparently it requires a semicolon after every field decl, might want to file a bug
    //messageSchema = messageSchema.substring(schemaType, messageSchema.length() - 1);
    messageSchema += "}";

    MessageType schema = MessageTypeParser.parseMessageType(messageSchema);

    CompressionCodecName codec = CompressionCodecName.UNCOMPRESSED;
    ParquetFileWriter w = new ParquetFileWriter(configuration, schema, path);
    w.start();
    HashMap<String, Integer> columnValuesWritten = new HashMap();
    int valsWritten;
    for (int k = 0; k < props.numberRowGroups; k++){
      w.startBlock(props.recordsPerRowGroup);
      currentBooleanByte = 0;
      booleanBitCounter.reset();

      for (FieldInfo fieldInfo : props.fields.values()) {

        if ( ! columnValuesWritten.containsKey(fieldInfo.name)){
          columnValuesWritten.put((String) fieldInfo.name, 0);
          valsWritten = 0;
        } else {
          valsWritten = columnValuesWritten.get(fieldInfo.name);
        }

        String[] path1 = {(String) fieldInfo.name};
        ColumnDescriptor c1 = schema.getColumnDescription(path1);

        w.startColumn(c1, props.recordsPerRowGroup, codec);
        int valsPerPage = (int) Math.ceil(props.recordsPerRowGroup / (float) fieldInfo.numberOfPages);
        byte[] bytes;
        RunLengthBitPackingHybridValuesWriter defLevels = new RunLengthBitPackingHybridValuesWriter(MAX_EXPECTED_BIT_WIDTH_FOR_DEFINITION_LEVELS, valsPerPage);
View Full Code Here
115
116
117
118
119
120
121
122
123
124
125
      DecimalMetadata decimalMetadata = ParquetTypeHelper.getDecimalMetadataForField(field);
      int length = ParquetTypeHelper.getLengthForMinorType(minorType);
      parquet.schema.Type type = new parquet.schema.PrimitiveType(repetition, primitiveTypeName, length, name, originalType, decimalMetadata);
      types.add(type);
    }
    schema = new MessageType("root", types);

    Path fileName = new Path(location, prefix + "_" + index + ".parquet");
    w = new ParquetFileWriter(conf, schema, fileName);
    w.start();
View Full Code Here
143
144
145
146
147
148
149
150
151
152
153
154
    return new ScanBatch(rowGroupScan, context, readers.iterator(), partitionColumns, selectedPartitionColumns);
  }

  private static boolean isComplex(ParquetMetadata footer) {
    MessageType schema = footer.getFileMetaData().getSchema();
    for (Type type : schema.getFields()) {
      if (!type.isPrimitive()) {
        return true;
      }
    }
    return false;
View Full Code Here
55
56
57
58
59
60
61
62
63
  }

  private void testConversion(final String columnNamesStr, final String columnsTypeStr, final String expectedSchema) throws Exception {
    final List<String> columnNames = createHiveColumnsFrom(columnNamesStr);
    final List<TypeInfo> columnTypes = createHiveTypeInfoFrom(columnsTypeStr);
    final MessageType messageTypeFound = HiveSchemaConverter.convert(columnNames, columnTypes);
    final MessageType expectedMT = MessageTypeParser.parseMessageType(expectedSchema);
    assertEquals("converting " + columnNamesStr + ": " + columnsTypeStr + " to " + expectedSchema, expectedMT, messageTypeFound);
  }
View Full Code Here
118
119
120
121
122
123
124
125
126
127
128
129
130
131
  public void testMapOriginalType() throws Exception {
    final String hiveColumnTypes = "map<string,string>";
    final String hiveColumnNames = "mapCol";
    final List<String> columnNames = createHiveColumnsFrom(hiveColumnNames);
    final List<TypeInfo> columnTypes = createHiveTypeInfoFrom(hiveColumnTypes);
    final MessageType messageTypeFound = HiveSchemaConverter.convert(columnNames, columnTypes);
    // this messageType only has one optional field, whose name is mapCol, original Type is MAP
    assertEquals(1, messageTypeFound.getFieldCount());
    parquet.schema.Type topLevel = messageTypeFound.getFields().get(0);
    assertEquals("mapCol",topLevel.getName());
    assertEquals(OriginalType.MAP, topLevel.getOriginalType());
    assertEquals(Repetition.OPTIONAL, topLevel.getRepetition());

    assertEquals(1, topLevel.asGroupType().getFieldCount());
View Full Code Here
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
        } else {
          // below allows schema evolution
          typeListTable.add(new PrimitiveType(Repetition.OPTIONAL, PrimitiveTypeName.BINARY, col));
        }
      }
      MessageType tableSchema = new MessageType(TABLE_SCHEMA, typeListTable);
      contextMetadata.put(HIVE_SCHEMA_KEY, tableSchema.toString());

      MessageType requestedSchemaByUser = tableSchema;
      final List<Integer> indexColumnsWanted = ColumnProjectionUtils.getReadColumnIDs(configuration);

      final List<Type> typeListWanted = new ArrayList<Type>();
      for (final Integer idx : indexColumnsWanted) {
        typeListWanted.add(tableSchema.getType(listColumns.get(idx)));
      }
      requestedSchemaByUser = new MessageType(fileSchema.getName(), typeListWanted);

      return new ReadContext(requestedSchemaByUser, contextMetadata);
    } else {
      contextMetadata.put(HIVE_SCHEMA_KEY, fileSchema.toString());
      return new ReadContext(fileSchema, contextMetadata);
View Full Code Here
121
122
123
124
125
126
127
128
129
    final Map<String, String> metadata = readContext.getReadSupportMetadata();
    if (metadata == null) {
      throw new IllegalStateException("ReadContext not initialized properly. " +
        "Don't know the Hive Schema.");
    }
    final MessageType tableSchema = MessageTypeParser.
        parseMessageType(metadata.get(HIVE_SCHEMA_KEY));
    return new DataWritableRecordConverter(readContext.getRequestedSchema(), tableSchema);
  }
View Full Code Here
TOP

Related Classes of parquet.schema.MessageType

  • com.asakusafw.directio.hive.parquet.DataModelMaterializer
  • com.asakusafw.directio.hive.parquet.DataModelWriteSupport
  • com.facebook.presto.hive.ParquetHiveRecordCursor$PrestoReadSupport
  • org.apache.drill.exec.store.parquet.ParquetRecordWriter
  • org.apache.drill.exec.store.parquet.ParquetScanBatchCreator
  • org.apache.drill.exec.store.parquet.TestFileGenerator
  • org.apache.drill.exec.store.parquet2.DrillParquetReader
  • org.apache.drill.exec.store.ParquetRecordReaderTest
  • org.apache.drill.ParquetSchemaMerge
  • org.apache.hadoop.hive.ql.io.parquet.convert.HiveSchemaConverter
    • Copyright © 2018 www.massapicom. All rights reserved.
    All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.