r.enforceMaxLength(getCharacterMaxLength(type));
return r;
}
case VARCHAR: {
HiveVarcharWritable r =
reuse == null ? new HiveVarcharWritable() : (HiveVarcharWritable) reuse;
// Use HiveVarchar's internal Text member to read the value.
deserializeText(buffer, invert, r.getTextValue());
// If we cache helper data for deserialization we could avoid having
// to call getVarcharMaxLength() on every deserialize call.
r.enforceMaxLength(getCharacterMaxLength(type));
return r;
}
case BINARY: {
BytesWritable bw = new BytesWritable() ;
// Get the actual length first
int start = buffer.tell();
int length = 0;
do {
byte b = buffer.read(invert);
if (b == 0) {
// end of string
break;
}
if (b == 1) {
// the last char is an escape char. read the actual char
buffer.read(invert);
}
length++;
} while (true);
if (length == buffer.tell() - start) {
// No escaping happened, so we are already done.
bw.set(buffer.getData(), start, length);
} else {
// Escaping happened, we need to copy byte-by-byte.
// 1. Set the length first.
bw.set(buffer.getData(), start, length);
// 2. Reset the pointer.
buffer.seek(start);
// 3. Copy the data.
byte[] rdata = bw.getBytes();
for (int i = 0; i < length; i++) {
byte b = buffer.read(invert);
if (b == 1) {
// The last char is an escape char, read the actual char.
// The serialization format escape \0 to \1, and \1 to \2,
// to make sure the string is null-terminated.
b = (byte) (buffer.read(invert) - 1);
}
rdata[i] = b;
}
// 4. Read the null terminator.
byte b = buffer.read(invert);
assert (b == 0);
}
return bw;
}
case DATE: {
DateWritable d = reuse == null ? new DateWritable()
: (DateWritable) reuse;
d.set(deserializeInt(buffer, invert));
return d;
}
case TIMESTAMP:
TimestampWritable t = (reuse == null ? new TimestampWritable() :
(TimestampWritable) reuse);
byte[] bytes = new byte[TimestampWritable.BINARY_SORTABLE_LENGTH];
for (int i = 0; i < bytes.length; i++) {
bytes[i] = buffer.read(invert);
}
t.setBinarySortable(bytes, 0);
return t;
case DECIMAL: {
// See serialization of decimal for explanation (below)
HiveDecimalWritable bdw = (reuse == null ? new HiveDecimalWritable() :
(HiveDecimalWritable) reuse);
int b = buffer.read(invert) - 1;
assert (b == 1 || b == -1 || b == 0);
boolean positive = b != -1;
int factor = buffer.read(invert) ^ 0x80;
for (int i = 0; i < 3; i++) {
factor = (factor << 8) + (buffer.read(invert) & 0xff);
}
if (!positive) {
factor = -factor;
}
int start = buffer.tell();
int length = 0;
do {
b = buffer.read(positive ? invert : !invert);
assert(b != 1);
if (b == 0) {
// end of digits
break;
}
length++;
} while (true);
if(decimalBuffer == null || decimalBuffer.length < length) {
decimalBuffer = new byte[length];
}
buffer.seek(start);
for (int i = 0; i < length; ++i) {
decimalBuffer[i] = buffer.read(positive ? invert : !invert);
}
// read the null byte again
buffer.read(positive ? invert : !invert);
String digits = new String(decimalBuffer, 0, length, decimalCharSet);
BigInteger bi = new BigInteger(digits);
HiveDecimal bd = HiveDecimal.create(bi).scaleByPowerOfTen(factor-length);
if (!positive) {
bd = bd.negate();
}
bdw.set(bd);
return bdw;
}
default: {
throw new RuntimeException("Unrecognized type: "
+ ptype.getPrimitiveCategory());
}
}
}
case LIST: {
ListTypeInfo ltype = (ListTypeInfo) type;
TypeInfo etype = ltype.getListElementTypeInfo();
// Create the list if needed
ArrayList<Object> r = reuse == null ? new ArrayList<Object>()
: (ArrayList<Object>) reuse;
// Read the list
int size = 0;
while (true) {
int more = buffer.read(invert);
if (more == 0) {
// \0 to terminate
break;
}
// \1 followed by each element
assert (more == 1);
if (size == r.size()) {
r.add(null);
}
r.set(size, deserialize(buffer, etype, invert, r.get(size)));
size++;
}
// Remove additional elements if the list is reused
while (r.size() > size) {
r.remove(r.size() - 1);
}
return r;
}
case MAP: {
MapTypeInfo mtype = (MapTypeInfo) type;
TypeInfo ktype = mtype.getMapKeyTypeInfo();
TypeInfo vtype = mtype.getMapValueTypeInfo();
// Create the map if needed
Map<Object, Object> r;
if (reuse == null) {
r = new HashMap<Object, Object>();
} else {
r = (HashMap<Object, Object>) reuse;
r.clear();
}
while (true) {
int more = buffer.read(invert);
if (more == 0) {
// \0 to terminate
break;
}
// \1 followed by each key and then each value
assert (more == 1);
Object k = deserialize(buffer, ktype, invert, null);
Object v = deserialize(buffer, vtype, invert, null);
r.put(k, v);
}
return r;
}
case STRUCT: {
StructTypeInfo stype = (StructTypeInfo) type;
List<TypeInfo> fieldTypes = stype.getAllStructFieldTypeInfos();
int size = fieldTypes.size();
// Create the struct if needed
ArrayList<Object> r = reuse == null ? new ArrayList<Object>(size)
: (ArrayList<Object>) reuse;
assert (r.size() <= size);
// Set the size of the struct
while (r.size() < size) {
r.add(null);
}
// Read one field by one field
for (int eid = 0; eid < size; eid++) {
r
.set(eid, deserialize(buffer, fieldTypes.get(eid), invert, r
.get(eid)));
}
return r;
}
case UNION: {
UnionTypeInfo utype = (UnionTypeInfo) type;
StandardUnion r = reuse == null ? new StandardUnion()
: (StandardUnion) reuse;
// Read the tag
byte tag = buffer.read(invert);
r.setTag(tag);
r.setObject(deserialize(buffer, utype.getAllUnionObjectTypeInfos().get(tag),
invert, null));
return r;
}
default: {
throw new RuntimeException("Unrecognized type: " + type.getCategory());