/**
* This method trims doc (1st arg) according to projected (2nd arg), which may have been modified
* by projection. This method works in-place, meaning that the doc may be mutated.
*/
private static void trim(DataMap doc, DataMap projected) {
DataMap toAddDoc = new DataMap();
Set<String> fields = doc.keySet();
List<String> toRemoveDoc = new ArrayList<String>(fields.size());
for (String f : fields) {
Object v = doc.get(f);
if (f.equals(PatchConstants.DELETE_COMMAND)) {
DataList deletedFields = (DataList) v;
DataList filteredDeleteFields = new DataList();
for (Object patchDeleteField : deletedFields) {
if (projected.containsKey(patchDeleteField)) {
filteredDeleteFields.add(patchDeleteField);
}
}
toRemoveDoc.add(f);
if (!filteredDeleteFields.isEmpty()) {
toAddDoc.put(PatchConstants.DELETE_COMMAND, filteredDeleteFields);
}
} else if (f.equals(PatchConstants.SET_COMMAND)) {
DataMap setFields = (DataMap) v;
Set<String> setFieldNames = setFields.keySet();
List<String> toRemove = new LinkedList<String>();
DataMap filteredSetFields = new DataMap();
for (String setFieldName : setFieldNames) {
if (projected.containsKey(setFieldName)) {
filteredSetFields.put(setFieldName, projected.get(setFieldName));
}
toRemove.add(setFieldName);
}
for (String fieldToRemove : toRemove) {
setFields.remove(fieldToRemove);
if (filteredSetFields.containsKey(fieldToRemove)) {
setFields.put(fieldToRemove, filteredSetFields.get(fieldToRemove));
}
}
if (setFields.isEmpty()) {
toRemoveDoc.add(f);
}
} else if (v instanceof DataMap) {
if (projected.containsKey(f)) {
trim((DataMap) v, (DataMap) projected.get(f));
} else {
toRemoveDoc.add(f);
}
}
}
// apply changes to doc
for (String f : toRemoveDoc) {
doc.remove(f);
}
for (String f : toAddDoc.keySet()) {
doc.put(f, toAddDoc.get(f));
}
}
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;
}
/**
* This method 'exposes' changes conveyed in the patch's meta commands to the main document.
* Contents of $set commands are moved to the node which contains $set command. Names of removed
* fields from $delete commands are moved to the nod which contains $delete command. The effect is
* that patch will resemble structurally document which it is supposed to modify. This allows
* application of projection to such patch to discover which changes relate to fields specified by
* that projection. Examples: $delete: ['x', 'y', 'z'] => x: true, y: true, z: true $set: {x: 10,
* y: {z: 'yeey'}, t: [10]} => x: 10, y: {z: 'yeey'}, t: [10]
*
* <p>This method works in-place, meaning that the doc may be mutated.
*/
private static void expose(DataMap doc) {
Set<String> fields = doc.keySet();
DataMap toAdd = new DataMap();
for (String f : fields) {
Object v = doc.get(f);
if (f.equals(PatchConstants.DELETE_COMMAND)) {
for (Object removedFields : (DataList) v) {
toAdd.put((String) removedFields, true);
}
} else if (f.equals(PatchConstants.SET_COMMAND)) {
toAdd.putAll((DataMap) v);
} else if (v instanceof DataMap) {
expose((DataMap) v);
}
}
doc.putAll(toAdd);
}
private Object translate(Object value, DataSchema dataSchema, Schema avroSchema) {
AvroOverride avroOverride = getAvroOverride(dataSchema);
if (avroOverride != null) {
return avroOverride
.getCustomDataTranslator()
.dataToAvroGeneric(this, value, dataSchema, avroSchema);
}
DataSchema dereferencedDataSchema = dataSchema.getDereferencedDataSchema();
DataSchema.Type type = dereferencedDataSchema.getType();
Object result;
switch (type) {
case NULL:
if (value != Data.NULL) {
appendMessage("value must be null for null schema");
result = BAD_RESULT;
break;
}
result = null;
break;
case BOOLEAN:
result = ((Boolean) value).booleanValue();
break;
case INT:
result = ((Number) value).intValue();
break;
case LONG:
result = ((Number) value).longValue();
break;
case FLOAT:
result = ((Number) value).floatValue();
break;
case DOUBLE:
result = ((Number) value).doubleValue();
break;
case STRING:
result = new Utf8((String) value);
break;
case BYTES:
result = ByteBuffer.wrap(translateBytes(value));
break;
case ENUM:
String enumValue = value.toString();
EnumDataSchema enumDataSchema = (EnumDataSchema) dereferencedDataSchema;
if (enumDataSchema.getSymbols().contains(enumValue) == false) {
appendMessage(
"enum value %1$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;
}
