Hi,
I think there is a bug CompressedWritable, more exactly in write()
method, which makes subclasses of CompressedWritable not "reusable".
More exactly: the data that is written to the DataOutput is always the
same, because is "cached" in the /compressed/ field.
public final void write(DataOutput out) throws IOException {
/if (compressed == null) {/
ByteArrayOutputStream deflated = new ByteArrayOutputStream();
Deflater deflater = new Deflater(Deflater.BEST_SPEED);
DataOutputStream dout =
new DataOutputStream(new DeflaterOutputStream(deflated, deflater));
writeCompressed(dout);
dout.close();
compressed = deflated.toByteArray();
}
out.writeInt(compressed.length);
out.write(compressed);
}
(code from hadoop 0.12.1)
This makes subclasses not "reusable", no matter if the fields change,
data written will be the same.
John
