ObjectIn/OutputStream improvements

[Martin Buchholz]

I had forgotten about the use of Modified UTF-8 in serialization.
Leave it alone.
http://en.wikipedia.org/wiki/UTF-8#Modified_UTF-8


On Thu, Feb 6, 2014 at 2:50 PM, Robert Stupp <snazy at snazy.de> wrote:

> Hi,
>
> I looked into the UTF serialization and deserialization code - and
> compared it a bit with the code behind "new String(byte[], Charset)",
> "String.getBytes(Charset)".
> Just to find something that can be safely reused in Object*Stream classes
> to optimize String handling.
>
> The first thing I noticed is that the class sun.nio.cs.UTF_8 class uses
> one byte for (char)0 - Object*Stream use two 0 bytes to represent (char)0.
> Although this representation is incompatible to the representation that the
> class UTF_8 uses it is not a big issue.
> The other thing is that Object*Stream seem not to be able to (de)serialize
> 21 bit characters. Methods in UTF_8 class are
> sun.nio.cs.UTF_8.Decoder#decode and sun.nio.cs.UTF_8.Encoder#encode. Is it
> ok to add 21-bit UTF-8 representation in Object*Stream?
>
> -
> Robert
>
> Am 06.02.2014 um 13:07 schrieb Chris Hegarty <chris.hegarty at oracle.com>:
>
> > Thanks Peter, this is a nice improvement. I’ll incorporate your changes
> before pushing.
> >
> > -Chris.
> >
> > On 5 Feb 2014, at 16:39, Peter Levart <peter.levart at gmail.com> wrote:
> >
> >> On 02/05/2014 04:11 PM, Chris Hegarty wrote:
> >>> Thanks stuart, Mike, and Paul.
> >>>
> >>>> - Why not have getClassSignature() return an interned string? (that's
> if interning is actually essential. Are we sure it's not just overhead?)
> >>> I didn’t want to change the existing use of interning here, just
> refactor the code a little to make it cleaner.
> >>>
> >>>> I think that would be a better way to spend our performance
> investigation budget. Thus, it might be better to remove the StringBuilder
> field from this patch and submit another one focusing on UTF decoding; the
> other changes in the patch look good.
> >>> Agreed. This could be looked at separately.
> >>>
> >>> Latest webrev:
> >>>  http://cr.openjdk.java.net/~chegar/serial_stupp.01/
> >>>
> >>> Thanks,
> >>> -Chris.
> >>
> >> Hi Chris,
> >>
> >> What about the following even less garbage-producing-and-copying
> ObjectStreamClass.get[Class|Method]Signature combo:
> >>
> >>   /**
> >>    * Returns JVM type signature for given class.
> >>    */
> >>   static String getClassSignature(Class<?> cl) {
> >>       if (cl.isPrimitive())
> >>           return getPrimitiveSignature(cl);
> >>       else
> >>           return appendClassSignature(new StringBuilder(),
> cl).toString();
> >>   }
> >>
> >>   private static StringBuilder appendClassSignature(StringBuilder sbuf,
> Class<?> cl) {
> >>       while (cl.isArray()) {
> >>           sbuf.append('[');
> >>           cl = cl.getComponentType();
> >>       }
> >>
> >>       if (cl.isPrimitive())
> >>           sbuf.append(getPrimitiveSignature(cl));
> >>       else
> >> sbuf.append('L').append(cl.getName().replace('.', '/')).append(';');
> >>
> >>       return sbuf;
> >>   }
> >>
> >>   /**
> >>    * Returns JVM type signature for given list of parameters and return
> type.
> >>    */
> >>   private static String getMethodSignature(Class<?>[] paramTypes,
> >>                                            Class<?> retType)
> >>   {
> >>       StringBuilder sbuf = new StringBuilder();
> >>       sbuf.append('(');
> >>       for (int i = 0; i < paramTypes.length; i++) {
> >>           appendClassSignature(sbuf, paramTypes[i]);
> >>       }
> >>       sbuf.append(')');
> >>       appendClassSignature(sbuf, retType);
> >>       return sbuf.toString();
> >>   }
> >>
> >>
> >> Regards, Peter
> >>
> >
>
>