public List<InferredType> materializeWithoutUnions() {
List<InferredType> allOptions = new ArrayList<InferredType>();
for (InferredType branch : unionTypes) {
allOptions.addAll(branch.materializeWithoutUnions());
}
return allOptions;
}
List<String> getBases() {
List<String> tr = new ArrayList<String>();
for (InferredType subelt : structTypes) {
tr.addAll(subelt.getBases());
}
return tr;
}
InferredType duplicate() {
List<InferredType> newBranches = new ArrayList<InferredType>();
for (InferredType branch : unionTypes) {
newBranches.add(branch.duplicate());
}
return new UnionType(newBranches);
}
/** Parse the given string, return resulting data if appropriate. */
ParseResult internalParse(
String s, Map<String, Integer> targetUnionDecisions, boolean mustConsumeStr) {
boolean hasData = false;
GenericData.Record gdr = new GenericData.Record(getAvroSchema());
String currentStr = s;
for (InferredType subelt : structTypes) {
if (currentStr.length() == 0) {
return null;
}
ParseResult pr = subelt.internalParse(currentStr, targetUnionDecisions, false);
if (pr == null) {
return null;
}
if (pr.hasData()) {
hasData = true;
gdr.put(subelt.getName(), pr.getData());
}
currentStr = pr.getRemainingString();
}
if (mustConsumeStr && currentStr.trim().length() != 0) {
return null;
}
return new ParseResult(gdr, hasData, currentStr);
}
Map<String, Set<Integer>> findCandidateUnionDecisions() {
Map<String, Set<Integer>> candidateUnionDecisions = new HashMap<String, Set<Integer>>();
for (InferredType subelt : structTypes) {
candidateUnionDecisions.putAll(subelt.findCandidateUnionDecisions());
}
return candidateUnionDecisions;
}
/** Parse the given string, return resulting data if appropriate. */
ParseResult internalParse(
String s, Map<String, Integer> targetUnionDecisions, boolean mustConsumeStr) {
//
// If there's no target decision, then go ahead and try all branches.
//
if (targetUnionDecisions == null || targetUnionDecisions.get(name) == null) {
for (InferredType subelt : unionTypes) {
ParseResult pr = subelt.internalParse(s, targetUnionDecisions, false);
if (pr != null
&& (!mustConsumeStr
|| (mustConsumeStr && pr.getRemainingString().trim().length() == 0))) {
return new ParseResult(pr.getData(), pr.hasData(), pr.getRemainingString());
}
}
return null;
}
//
// If there is a target decision, then carry it out.
//
InferredType subelt = unionTypes.get(targetUnionDecisions.get(name));
ParseResult pr = subelt.internalParse(s, targetUnionDecisions, false);
if (pr != null
&& (!mustConsumeStr || (mustConsumeStr && pr.getRemainingString().trim().length() == 0))) {
return new ParseResult(pr.getData(), pr.hasData(), pr.getRemainingString());
}
return null;
}
public double getDescriptionCost() {
double dc = CARD_COST;
for (InferredType it : unionTypes) {
dc += it.getDescriptionCost();
}
return dc;
}
InferredType duplicate() {
List<InferredType> newElts = new ArrayList<InferredType>();
for (InferredType elt : structTypes) {
newElts.add(elt.duplicate());
}
return new StructType(newElts);
}
public InferredType hoistUnions() {
List<InferredType> newStructTypes = new ArrayList<InferredType>();
for (InferredType it : structTypes) {
newStructTypes.add(it.hoistUnions());
}
return new StructType(newStructTypes);
}
public void write(DataOutput out) throws IOException {
out.write(STRUCT_TYPE);
out.writeUTF(name);
out.writeInt(structTypes.size());
for (InferredType it : structTypes) {
it.write(out);
}
}
public void write(DataOutput out) throws IOException {
out.write(UNION_TYPE);
out.writeUTF(name);
out.writeInt(unionTypes.size());
for (InferredType it : unionTypes) {
it.write(out);
}
}
public String toString() {
StringBuffer buf = new StringBuffer();
buf.append("(Union (" + unionTypes.size() + "): ");
for (InferredType it : unionTypes) {
buf.append(it.toString() + ", ");
}
buf.append(") ");
return buf.toString();
}
public String toString() {
StringBuffer buf = new StringBuffer();
buf.append("(Struct: ");
for (InferredType it : structTypes) {
buf.append(it.toString() + ", ");
}
buf.append(") ");
return buf.toString();
}
public InferredType hoistUnions() {
List<InferredType> newUnionTypes = new ArrayList<InferredType>();
for (InferredType it : unionTypes) {
if (it instanceof UnionType) {
UnionType subUnion = (UnionType) it;
for (InferredType it2 : subUnion.unionTypes) {
newUnionTypes.add(it2.hoistUnions());
}
} else {
newUnionTypes.add(it.hoistUnions());
}
}
return new UnionType(newUnionTypes);
}
Schema computeAvroSchema() {
HashSet<String> observedSchemas = new HashSet<String>();
List<Schema> fields = new ArrayList<Schema>();
for (InferredType it : unionTypes) {
Schema itS = it.getAvroSchema();
if (itS == null) {
continue;
}
String schemaDesc = itS.toString();
if (!observedSchemas.contains(schemaDesc)) {
observedSchemas.add(schemaDesc);
fields.add(it.getAvroSchema());
}
}
return Schema.createUnion(fields);
}
public List<InferredType> materializeWithoutUnions() {
List<InferredType> newArrays = new ArrayList<InferredType>();
for (InferredType subtype : bodyType.materializeWithoutUnions()) {
newArrays.add(new ArrayType(subtype));
}
return newArrays;
}
public void readFields(DataInput in) throws IOException {
int numUnionElts = in.readInt();
this.unionTypes = new ArrayList<InferredType>();
for (int i = 0; i < numUnionElts; i++) {
unionTypes.add(InferredType.readType(in));
}
this.schema = computeAvroSchema();
}
/** Deserialize an unknown InferredType from the given input stream */
public static InferredType readType(DataInput in) throws IOException {
InferredType it = null;
byte b = in.readByte();
String name = in.readUTF();
if (b == BASE_TYPE) {
it = new BaseType(name);
} else if (b == STRUCT_TYPE) {
it = new StructType(name);
} else if (b == ARRAY_TYPE) {
it = new ArrayType(name);
} else if (b == UNION_TYPE) {
it = new UnionType(name);
} else {
throw new IOException("No type found: " + b);
}
it.readFields(in);
return it;
}
Map<String, Set<Integer>> findCandidateUnionDecisions() {
HashSet<Integer> curDecisions = new HashSet<Integer>();
for (int i = 0; i < unionTypes.size(); i++) {
for (int j = i + 1; j < unionTypes.size(); j++) {
List<String> iBases = unionTypes.get(i).getBases();
List<String> jBases = unionTypes.get(j).getBases();
if (isPrefixOf(iBases, jBases) || isPrefixOf(jBases, iBases)) {
curDecisions.add(i);
curDecisions.add(j);
}
}
}
Map<String, Set<Integer>> candidateUnionDecisions = new HashMap<String, Set<Integer>>();
for (InferredType subelt : unionTypes) {
candidateUnionDecisions.putAll(subelt.findCandidateUnionDecisions());
}
if (curDecisions.size() > 0) {
candidateUnionDecisions.put(name, curDecisions);
}
return candidateUnionDecisions;
}
Schema computeAvroSchema() {
List<Schema.Field> fields = new ArrayList<Schema.Field>();
for (InferredType it : structTypes) {
Schema itS = it.getAvroSchema();
if (itS == null) {
continue;
}
fields.add(
new Schema.Field(
it.getName(), it.getAvroSchema(), it.getDocString(), it.getDefaultValue()));
}
Schema s = Schema.createRecord(name, "RECORD", "", false);
s.setFields(fields);
return s;
}
/** Parse the given string, return resulting data if appropriate. */
ParseResult internalParse(
String s, Map<String, Integer> targetUnionDecisions, boolean mustConsumeStr) {
boolean hasData = false;
Schema localSchema = getAvroSchema();
GenericData.Array gda = new GenericData.Array(5, localSchema);
Map<String, Integer> curUnionDecisions = new HashMap<String, Integer>();
String currentStr = s;
while (true) {
ParseResult pr = bodyType.internalParse(currentStr, targetUnionDecisions, false);
if (pr == null) {
break;
}
assert (pr.hasData());
gda.add(pr.getData());
currentStr = pr.getRemainingString();
}
if (mustConsumeStr && currentStr.trim().length() != 0) {
return null;
}
return new ParseResult(gda, true, currentStr);
}
public String toString() {
return "(Array: " + bodyType.toString() + ") ";
}
Schema computeAvroSchema() {
return Schema.createArray(bodyType.getAvroSchema());
}
public double getDescriptionCost() {
return CARD_COST + bodyType.getDescriptionCost();
}
Map<String, Set<Integer>> findCandidateUnionDecisions() {
return bodyType.findCandidateUnionDecisions();
}
List<String> getBases() {
return bodyType.getBases();
}
public void readFields(DataInput in) throws IOException {
bodyType = InferredType.readType(in);
this.schema = computeAvroSchema();
}
public void write(DataOutput out) throws IOException {
out.write(ARRAY_TYPE);
out.writeUTF(name);
bodyType.write(out);
}
public InferredType hoistUnions() {
return new ArrayType(bodyType.hoistUnions());
}