private static List<Map.Entry<String, Object>> orderMapEntries(
RecordDataSchema schema, DataMap map) {
List<Map.Entry<String, Object>> output = new ArrayList<Map.Entry<String, Object>>(map.size());
List<RecordDataSchema.Field> fields = schema.getFields();
// collect fields in the record schema in the order the fields are declared
for (RecordDataSchema.Field field : fields) {
String fieldName = field.getName();
Object found = map.get(fieldName);
if (found != null) {
output.add(new AbstractMap.SimpleImmutableEntry<String, Object>(fieldName, found));
}
}
// collect fields that are in the DataMap that is not in the record schema.
List<Map.Entry<String, Object>> uncollected =
new ArrayList<Map.Entry<String, Object>>(map.size() - output.size());
for (Map.Entry<String, Object> e : map.entrySet()) {
if (schema.contains(e.getKey()) == false) {
uncollected.add(e);
}
}
Collections.sort(
uncollected,
new Comparator<Map.Entry<String, Object>>() {
@Override
public int compare(Map.Entry<String, Object> o1, Map.Entry<String, Object> o2) {
return o1.getKey().compareTo(o2.getKey());
}
});
output.addAll(uncollected);
return output;
}
private Object translate(Object value, DataSchema dataSchema, Schema avroSchema) {
AvroOverride avroOverride = getAvroOverride(dataSchema);
if (avroOverride != null) {
return avroOverride
.getCustomDataTranslator()
.dataToAvroGeneric(this, value, dataSchema, avroSchema);
}
DataSchema dereferencedDataSchema = dataSchema.getDereferencedDataSchema();
DataSchema.Type type = dereferencedDataSchema.getType();
Object result;
switch (type) {
case NULL:
if (value != Data.NULL) {
appendMessage("value must be null for null schema");
result = BAD_RESULT;
break;
}
result = null;
break;
case BOOLEAN:
result = ((Boolean) value).booleanValue();
break;
case INT:
result = ((Number) value).intValue();
break;
case LONG:
result = ((Number) value).longValue();
break;
case FLOAT:
result = ((Number) value).floatValue();
break;
case DOUBLE:
result = ((Number) value).doubleValue();
break;
case STRING:
result = new Utf8((String) value);
break;
case BYTES:
result = ByteBuffer.wrap(translateBytes(value));
break;
case ENUM:
String enumValue = value.toString();
EnumDataSchema enumDataSchema = (EnumDataSchema) dereferencedDataSchema;
if (enumDataSchema.getSymbols().contains(enumValue) == false) {
appendMessage(
"enum value %1$s not one of %2$s", enumValue, enumDataSchema.getSymbols());
result = BAD_RESULT;
break;
}
result = _avroAdapter.createEnumSymbol(avroSchema, enumValue);
break;
case FIXED:
byte[] bytes = translateBytes(value);
FixedDataSchema fixedDataSchema = (FixedDataSchema) dereferencedDataSchema;
if (fixedDataSchema.getSize() != bytes.length) {
appendMessage(
"ByteString size %1$d != FixedDataSchema size %2$d",
bytes.length, fixedDataSchema.getSize());
result = null;
break;
}
GenericData.Fixed fixed = new GenericData.Fixed(avroSchema);
fixed.bytes(bytes);
result = fixed;
break;
case MAP:
DataMap map = (DataMap) value;
DataSchema valueDataSchema = ((MapDataSchema) dereferencedDataSchema).getValues();
Schema valueAvroSchema = avroSchema.getValueType();
Map<String, Object> avroMap = new HashMap<String, Object>(map.size());
for (Map.Entry<String, Object> entry : map.entrySet()) {
String key = entry.getKey();
_path.addLast(key);
Object entryAvroValue = translate(entry.getValue(), valueDataSchema, valueAvroSchema);
_path.removeLast();
avroMap.put(key, entryAvroValue);
}
result = avroMap;
break;
case ARRAY:
DataList list = (DataList) value;
DataSchema elementDataSchema = ((ArrayDataSchema) dereferencedDataSchema).getItems();
Schema elementAvroSchema = avroSchema.getElementType();
GenericData.Array<Object> avroList =
new GenericData.Array<Object>(list.size(), avroSchema);
for (int i = 0; i < list.size(); i++) {
_path.addLast(i);
Object entryAvroValue = translate(list.get(i), elementDataSchema, elementAvroSchema);
_path.removeLast();
avroList.add(entryAvroValue);
}
result = avroList;
break;
case RECORD:
map = (DataMap) value;
RecordDataSchema recordDataSchema = (RecordDataSchema) dereferencedDataSchema;
GenericData.Record avroRecord = new GenericData.Record(avroSchema);
for (RecordDataSchema.Field field : recordDataSchema.getFields()) {
String fieldName = field.getName();
DataSchema fieldDataSchema = field.getType();
Schema.Field avroField = avroSchema.getField(fieldName);
if (avroField == null) {
// field present in input but there is no field for it in Avro schema.
// TODO: Whether and how to indicate this condition to clients.
continue;
}
_path.addLast(fieldName);
Schema fieldAvroSchema = avroField.schema();
Object fieldValue = map.get(fieldName);
boolean isOptional = field.getOptional();
if (isOptional) {
if (fieldDataSchema.getDereferencedType() != DataSchema.Type.UNION) {
if (fieldValue == null) {
fieldValue = Data.NULL;
fieldDataSchema = DataSchemaConstants.NULL_DATA_SCHEMA;
}
Map.Entry<String, Schema> fieldAvroEntry =
findUnionMember(fieldDataSchema, fieldAvroSchema);
if (fieldAvroEntry == null) {
_path.removeLast();
continue;
}
fieldAvroSchema = fieldAvroEntry.getValue();
} else {
// already a union
if (fieldValue == null) {
// field is not present
fieldValue = Data.NULL;
fieldDataSchema = DataSchemaConstants.NULL_DATA_SCHEMA;
}
}
} else {
if (fieldValue == null) {
appendMessage("required field is absent");
_path.removeLast();
continue;
}
}
Object fieldAvroValue = translate(fieldValue, fieldDataSchema, fieldAvroSchema);
avroRecord.put(fieldName, fieldAvroValue);
_path.removeLast();
}
result = avroRecord;
break;
case UNION:
UnionDataSchema unionDataSchema = (UnionDataSchema) dereferencedDataSchema;
String key;
Object memberValue;
if (value == Data.NULL) {
key = DataSchemaConstants.NULL_TYPE;
memberValue = Data.NULL;
} else {
map = (DataMap) value;
Map.Entry<String, Object> entry = map.entrySet().iterator().next();
key = entry.getKey();
memberValue = entry.getValue();
}
DataSchema memberDataSchema = unionDataSchema.getType(key);
Map.Entry<String, Schema> memberAvroEntry = findUnionMember(memberDataSchema, avroSchema);
if (memberAvroEntry == null) {
result = BAD_RESULT;
break;
}
Schema memberAvroSchema = memberAvroEntry.getValue();
_path.addLast(memberAvroEntry.getKey());
Object memberAvroValue = translate(memberValue, memberDataSchema, memberAvroSchema);
_path.removeLast();
result = memberAvroValue;
break;
default:
appendMessage("schema type unknown %1$s", dereferencedDataSchema.getType());
result = BAD_RESULT;
break;
}
return result;
}
private Object translate(Object value, DataSchema dataSchema, Schema avroSchema) {
AvroOverride avroOverride = getAvroOverride(dataSchema);
if (avroOverride != null) {
return avroOverride
.getCustomDataTranslator()
.avroGenericToData(this, value, avroSchema, dataSchema);
}
DataSchema dereferencedDataSchema = dataSchema.getDereferencedDataSchema();
DataSchema.Type type = dereferencedDataSchema.getType();
Object result;
switch (type) {
case NULL:
if (value != null) {
appendMessage("value must be null for null schema");
result = BAD_RESULT;
break;
}
result = Data.NULL;
break;
case BOOLEAN:
result = ((Boolean) value).booleanValue();
break;
case INT:
result = ((Number) value).intValue();
break;
case LONG:
result = ((Number) value).longValue();
break;
case FLOAT:
result = ((Number) value).floatValue();
break;
case DOUBLE:
result = ((Number) value).doubleValue();
break;
case STRING:
result = value.toString();
break;
case BYTES:
ByteBuffer byteBuffer = (ByteBuffer) value;
ByteString byteString = ByteString.copy(byteBuffer);
byteBuffer.rewind();
result = byteString;
break;
case ENUM:
String enumValue = value.toString();
EnumDataSchema enumDataSchema = (EnumDataSchema) dereferencedDataSchema;
if (enumDataSchema.getSymbols().contains(enumValue) == false) {
appendMessage(
"enum value %1$s not one of %2$s", enumValue, enumDataSchema.getSymbols());
result = BAD_RESULT;
break;
}
result = enumValue;
break;
case FIXED:
GenericFixed fixed = (GenericFixed) value;
byte[] fixedBytes = fixed.bytes();
FixedDataSchema fixedDataSchema = (FixedDataSchema) dereferencedDataSchema;
if (fixedDataSchema.getSize() != fixedBytes.length) {
appendMessage(
"GenericFixed size %1$d != FixedDataSchema size %2$d",
fixedBytes.length, fixedDataSchema.getSize());
result = BAD_RESULT;
break;
}
byteString = ByteString.copy(fixedBytes);
result = byteString;
break;
case MAP:
@SuppressWarnings("unchecked")
Map<?, Object> map = (Map<?, Object>) value;
DataSchema valueDataSchema = ((MapDataSchema) dereferencedDataSchema).getValues();
Schema valueAvroSchema = avroSchema.getValueType();
DataMap dataMap = new DataMap(map.size());
for (Map.Entry<?, Object> entry : map.entrySet()) {
String key = entry.getKey().toString();
_path.addLast(key);
Object entryValue = translate(entry.getValue(), valueDataSchema, valueAvroSchema);
_path.removeLast();
dataMap.put(key, entryValue);
}
result = dataMap;
break;
case ARRAY:
GenericArray<?> list = (GenericArray<?>) value;
DataSchema elementDataSchema = ((ArrayDataSchema) dereferencedDataSchema).getItems();
Schema elementAvroSchema = avroSchema.getElementType();
DataList dataList = new DataList(list.size());
for (int i = 0; i < list.size(); i++) {
_path.addLast(i);
Object entryValue = translate(list.get(i), elementDataSchema, elementAvroSchema);
_path.removeLast();
dataList.add(entryValue);
}
result = dataList;
break;
case RECORD:
GenericRecord record = (GenericRecord) value;
RecordDataSchema recordDataSchema = (RecordDataSchema) dereferencedDataSchema;
dataMap = new DataMap(avroSchema.getFields().size());
for (RecordDataSchema.Field field : recordDataSchema.getFields()) {
String fieldName = field.getName();
Object fieldValue = record.get(fieldName);
// fieldValue could be null if the Avro schema does not contain the named field or
// the field is present with a null value. In either case we do not add a value
// to the translated DataMap. We do not consider optional/required/default here
// either (i.e. it is not an error if a required field is missing); the user can
// later call ValidateDataAgainstSchema with various
// settings for RequiredMode to obtain the desired behaviour.
if (fieldValue == null) {
continue;
}
boolean isOptional = field.getOptional();
DataSchema fieldDataSchema = field.getType();
Schema fieldAvroSchema = avroSchema.getField(fieldName).schema();
if (isOptional && (fieldDataSchema.getDereferencedType() != DataSchema.Type.UNION)) {
// Avro schema should be union with 2 types: null and the field's type.
Map.Entry<String, Schema> fieldAvroEntry =
findUnionMember(fieldDataSchema, fieldAvroSchema);
if (fieldAvroEntry == null) {
continue;
}
fieldAvroSchema = fieldAvroEntry.getValue();
}
_path.addLast(fieldName);
dataMap.put(fieldName, translate(fieldValue, fieldDataSchema, fieldAvroSchema));
_path.removeLast();
}
result = dataMap;
break;
case UNION:
UnionDataSchema unionDataSchema = (UnionDataSchema) dereferencedDataSchema;
Map.Entry<DataSchema, Schema> memberSchemas =
findUnionMemberSchema(value, unionDataSchema, avroSchema);
if (memberSchemas == null) {
result = BAD_RESULT;
break;
}
if (value == null) {
// schema must be "null" schema
result = Data.NULL;
} else {
DataSchema memberDataSchema = memberSchemas.getKey();
Schema memberAvroSchema = memberSchemas.getValue();
String key = memberDataSchema.getUnionMemberKey();
dataMap = new DataMap(1);
_path.addLast(key);
dataMap.put(key, translate(value, memberDataSchema, memberAvroSchema));
_path.removeLast();
result = dataMap;
}
break;
default:
appendMessage("schema type unknown %1$s", dereferencedDataSchema.getType());
result = BAD_RESULT;
break;
}
return result;
}
private void checkRecord(RecordDataSchema older, RecordDataSchema newer) {
checkName(older, newer);
List<RecordDataSchema.Field> commonFields =
new ArrayList<RecordDataSchema.Field>(newer.getFields().size());
List<String> newerRequiredAdded = new CheckerArrayList<String>();
List<String> newerOptionalAdded = new CheckerArrayList<String>();
List<String> requiredToOptional = new CheckerArrayList<String>();
List<String> optionalToRequired = new CheckerArrayList<String>();
List<String> newerRequiredRemoved = new CheckerArrayList<String>();
List<String> newerOptionalRemoved = new CheckerArrayList<String>();
for (RecordDataSchema.Field newerField : newer.getFields()) {
String fieldName = newerField.getName();
RecordDataSchema.Field olderField = older.getField(fieldName);
if (olderField == null) {
(newerField.getOptional() ? newerOptionalAdded : newerRequiredAdded).add(fieldName);
} else {
commonFields.add(newerField);
boolean newerFieldOptional = newerField.getOptional();
if (newerFieldOptional != olderField.getOptional()) {
(newerFieldOptional ? requiredToOptional : optionalToRequired).add(fieldName);
}
}
}
for (RecordDataSchema.Field olderField : older.getFields()) {
String fieldName = olderField.getName();
RecordDataSchema.Field newerField = newer.getField(fieldName);
if (newerField == null) {
(olderField.getOptional() ? newerOptionalRemoved : newerRequiredRemoved).add(fieldName);
}
}
if (newerRequiredAdded.isEmpty() == false) {
appendMessage(
CompatibilityMessage.Impact.BREAKS_NEW_READER,
"new record added required fields %s",
newerRequiredAdded);
}
if (newerRequiredRemoved.isEmpty() == false) {
appendMessage(
CompatibilityMessage.Impact.BREAKS_OLD_READER,
"new record removed required fields %s",
newerRequiredRemoved);
}
if (optionalToRequired.isEmpty() == false) {
appendMessage(
CompatibilityMessage.Impact.BREAKS_NEW_READER,
"new record changed optional fields to required fields %s",
optionalToRequired);
}
if (requiredToOptional.isEmpty() == false) {
appendMessage(
CompatibilityMessage.Impact.BREAKS_OLD_READER,
"new record changed required fields to optional fields %s",
requiredToOptional);
}
if (newerOptionalAdded.isEmpty() == false) {
appendMessage(
CompatibilityMessage.Impact.OLD_READER_IGNORES_DATA,
"new record added optional fields %s",
newerOptionalAdded);
}
if (newerOptionalRemoved.isEmpty() == false) {
appendMessage(
CompatibilityMessage.Impact.NEW_READER_IGNORES_DATA,
"new record removed optional fields %s",
newerOptionalRemoved);
}
for (RecordDataSchema.Field newerField : commonFields) {
String fieldName = newerField.getName();
_path.addLast(fieldName);
RecordDataSchema.Field olderField = older.getField(fieldName);
assert (olderField != null);
check(olderField.getType(), newerField.getType());
_path.removeLast();
}
}