XOM API for Processing XML with Java

[Sandy Mustard <Sandy.Mustard AT ugsplm.com>]

After looking at the test code, I think some analysis might be needed 
before we can rely on these numbers.  Enlighten me if I am wrong but 
doesn't loading a class that references other classes also loads the 
referenced classes?  If so, the loading of Node also loads ParentNode. 
Later in the test code, ParentNode is loaded, doesn't the JVM already 
have this in memory?  Wouldn't an analysis of the class heirarchies and 
references be appropriate in developing the test case?
Would also a test case that approximates the normal pattern of class 
reference/usage from a user's program be appropriate?

Sandy Mustard

Janek Bogucki wrote:

> I performed a benchmark on the class loading times for a compressed jar 
> versus an uncompressed jar
> and my conclusion was that there is not much justification for distributing 
> xom-1.0d25.jar in
> uncompressed form. (I am assuming class loading time improvement was the 
> motivation for generating
> an uncompressed jar. If that was not the sole reason then this may be 
> irrelevant.)
>
> Jar sizes:
>
>     99590 xom-1.0d25-compressed.jar
>    248217 xom-1.0d25.jar
>
> The benchmark class is attached. This is the result of five runs. Times in 
> milliseconds.
>
>    uncompressed    compressed
>    417             439
>    428             440
>    414             434
>    415             441
>    416             445
>
> Given that a class is loaded once, on demand, in the lifetime of the JVM this 
> benchmark shows a
> marginal gain in a usage pattern designed to maximise the difference. 
> Shipping a jar which is ~2.5
> times as large as the compressed version is the wrong size of 80/20. Changing 
> build.xml to
> generate a compressed jar would definitely improve XOM's golf score.
>
> -Janek
>
>
>         
>         
>                 
> ___________________________________________________________
> Yahoo! Messenger - Communicate instantly..."Ping" 
> your friends today! Download Messenger Now 
> http://uk.messenger.yahoo.com/download/index.html
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> XOM-interest mailing list
> XOM-interest AT lists.ibiblio.org
> http://lists.ibiblio.org/mailman/listinfo/xom-interest