Bug in CompressingUnMarshaller

[dreyk]

You can find in documentation reference on using CompressingUnMarshaller for reduce trafic. For this aim you can define you UnMarshaller by this way CompressingUnMarshaller(new HTTPUnMarshaller()). But it's does't work, see you read method for CompressingUnMarshaller

public Object read(InputStream inputStream, Map metadata) throws IOException, ClassNotFoundException
 {
 GZIPInputStream gzis = new GZIPInputStream(inputStream);
 ObjectInputStream ois = SerializationStreamFactory.getManagerInstance(getSerializationType()).createRegularInput(gzis);

 if(wrappedUnMarshaller != null)
 {
 //if(wrappedUnMarshaller instanceof HTTPUnMarshaller){
 //
 //}
 return wrappedUnMarshaller.read(ois, metadata);
 }
 else
 {
 return super.read(ois, metadata);
 }
 }

wrappedUnMarshaller.read(ois, metadata) will generate error for HTTPUnMarshaller wrapped class, because in following code

 while(amtRead > 0)
 {
 byteOutputStream.write(byteBuffer, pointer, amtRead);
 if(amtRead < bufferSize & byteOutputStream.size() >= contentLength)
 {
 //done reading, so process
 break;
 }
 amtRead = inputStream.read(byteBuffer);
 }

you don't read all content actualy, beacause contentLength equals comressed lenght of content while you read uncomressed stream!!! You must first set in metadata uncomressed length of the content!!!

[ron_sigal]

Hi Alex,

Thanks for the input. I'll look into it.

-Ron

[dreyk]

Hi, thank you for attention.
I solve this problem by following way.
1. First I read compressed stream to some byte array.
2. Second calculate original buffer size and place it into metadata.

[ron_sigal]

CompressingUnMarshaller now sets content-length to Integer.MAX_VALUE if the wrapped unmarshaller is HTTPUnMarshaller. This way, HTTPUnMarshaller will not stop reading until it hits end of file.

See JIRA issue JBREM-927 "Adjust content length when CompressingUnMarshaller wraps HTTPUnMarshaller" (http://jira.jboss.com/jira/browse/JBREM-927).