private BlockSchema generateOutSchema(BlockSchema inputSchema) {
List<ColumnType> outputColumnTypes = new ArrayList<ColumnType>();
for (ColumnType ct : inputSchema.getColumnTypes()) {
String colName = ct.getName();
int colIndex = inputSchema.getIndex(colName);
if (!flattenColumnNameSet.contains(colName)) {
outputColumnTypes.add(ct);
} else {
BlockSchema inputNestedColumnSchema = ct.getColumnSchema();
ColumnType[] ctypes = inputNestedColumnSchema.getColumnTypes();
if (ctypes.length == 1 && ctypes[0].getType() == DataType.TUPLE)
inputNestedColumnSchema = ctypes[0].getColumnSchema();
List<ColumnType> flattedOutputColumnTypes = inputColumnIndexToOutputTypes.get(colIndex);
if (flattedOutputColumnTypes != null && !flattedOutputColumnTypes.isEmpty()) {
// output schema published in json.
// TODO: assert output schema in json matches nested input schema for the column
if (inputNestedColumnSchema == null || inputNestedColumnSchema.getColumnTypes() == null)
throw new RuntimeException(
"Invalid schema for columnn: "
+ colName
+ " column schema: "
+ inputNestedColumnSchema);
if (flattedOutputColumnTypes.size() != inputNestedColumnSchema.getColumnTypes().length)
throw new RuntimeException(
"Output column specification does not match number of input fields for " + colName);
} else {
// output schema not published in json. Extract from nested input column schema
if (inputNestedColumnSchema == null) {
throw new RuntimeException("Schema is unknown for column: " + colName);
} else {
List<ColumnType> subColTypes = Arrays.asList(inputNestedColumnSchema.getColumnTypes());
flattedOutputColumnTypes = new ArrayList<ColumnType>();
flattedOutputColumnTypes.addAll(subColTypes);
}
inputColumnIndexToOutputTypes.put(colIndex, flattedOutputColumnTypes);
}
outputColumnTypes.addAll(flattedOutputColumnTypes);
}
}
return new BlockSchema(outputColumnTypes.toArray(new ColumnType[0]));
}
private BlockSchema generateSchema(
Map<String, CodeDictionary> dictionaryMap, BlockSchema originalSchema) {
numColumns = originalSchema.getNumColumns();
ColumnType[] columnTypes = new ColumnType[numColumns];
// create dictionary array
dictionaries = new CodeDictionary[numColumns];
decodedTuple = TupleFactory.getInstance().newTuple(numColumns);
for (int i = 0; i < columnTypes.length; i++) {
ColumnType type = new ColumnType();
columnTypes[i] = type;
type.setName(originalSchema.getName(i));
if (dictionaryMap.containsKey(type.getName())) {
// this column is decoded. Transform the schema
type.setType(DataType.STRING);
dictionaries[i] = dictionaryMap.get(type.getName());
} else {
// this column is not decoded. Keep the schema intact
type.setType(originalSchema.getType(i));
dictionaries[i] = null;
}
}
return new BlockSchema(columnTypes);
}
