Re: CPU Profiling with (j)visualvm
Marc van Dongen wrote:
Thanks. My main application --- the one which made me start this
thread --- is built using several user-defined packages. Since it
didn't make sense to me to write just one package for one main
application, I didn't bother putting the application in a package. Boy
was I punished for that. It has taken me weeks between starting with
(command-line) visualvm [sic] and getting some profiling results.
Interestingly, profiling with netbeans [sic] _without_ packages has always
worked for me (but I don't want to use netbeans [sic]).
No less than the JLS itself tells us, right up in the first part of chapter
7: "For small programs and casual development, a package can be unnamed
(??7.4.2) or have a simple name, but if code is to be widely distributed,
unique package names should be chosen (??7.7)."
If the JLS calls something out like that, one knows /a priori/ that it's
perilous to ignore.
The section John linked goes into even more detail.
One learns this particular advice early on in Java training. The basic Java
tutorial, "Learning the Java Language", in its chapter on packages tells us,
"Generally speaking, an unnamed package is only for small or temporary
applications or when you are just beginning the development process.
Otherwise, classes and interfaces belong in named packages."
http://download.oracle.com/javase/tutorial/java/package/createpkgs.html
--
Lew
Honi soit qui mal y pense.