Your thoughts on Java API and Archives (zip,jar etc..)

[christian.bongiorno]

So, I have been working extensively with jar files of late and now we
need to migrate to tar files, and the nagging problem continues to be

LACK OF INTERFACE!

I mean, the use of a jar and a manifest using the API's is really
whacked -- you can't create a Manifest and add entries to it, you have
to create a properly formatted UTF-8 byte sequence. But of course, you
can READ manifest entries all day.

Now we have the issue of the Streams and files. The streams have a
handfull of methods in common, but do things like:

ZipEntry getNextEntry()

JarEntry getNextJarEntry()

Now, the implementation I have found for tar (which the API's do not
have) is that they too have similar method names (close(),
putNextEntry()) etc, but because there isn't a common interface for
archives, I cannot seemlessly process the above mentioned types.

What I have done, to facilitate thise generic behavior is to implement
a combination of adapter interfaces and delegation -- and then I only
have to switch my current code to the new interface.

basically, I have done is

public interface ArchiveInputStream {

public ArchiveEntry nextEntry() throws IOException;

public int read(byte[] outBucket,int start, int amount) throws
IOException;

public void close() throws IOException;
}

public class MyJarInputStream extends JarInputStream implements
ArchiveInputStream {
private static final Log logger =
LogFactory.getLog(DaveJarInputStream.class);

public MyJarInputStream (InputStream in) throws IOException {
super(in);
}

public ArchiveEntry nextEntry() throws IOException {
return new JarArchiveEntry(getNextJarEntry());
}

}

I have done the same trick with the entry types for the archives and
OutputStreams

This seems to me to be exceptionally poor design in the Java libraries
-- a rarity.
Thoughts? Am I just spiking my java with peyote or what?

Christian

http://christian.bongiorno.org