Re: Proposed new Java feature
"Eric Sosman" <esosman@ieee-dot-org.invalid> wrote in message
news:jptvdf$1s5$1@dont-email.me...
On 5/27/2012 3:04 PM, Mike Schilling wrote:
"Daniel Pitts"<newsgroup.nospam@virtualinfinity.net> wrote in message
news:8Euwr.47425$On2.20024@newsfe16.iad...
On 5/27/12 11:00 AM, Mike Schilling wrote:
"markspace"<-@.> wrote in message news:jptkmp$vbg$1@dont-email.me...
On 5/26/2012 4:11 PM, Mike Schilling wrote:
Proposed feature: a static method on Thread that clears all
ThreadLocals
for
the current thread.
I can see your points. However, I don't have any real experience with
ThreadLocal, and when a neophyte agrees with your argument, that's a
red
flag.
Here's a blog where someone seems to have the same issue as you.
<http://weblogs.java.net/blog/jjviana/archive/2010/06/10/threadlocal-thread-pool-bad-idea-or-dealing-apparent-glassfish-memor>
At the end of the comments, there's a suggestion to use
ThreadLocal::remove(), with the implication that it allows the thread
local variable to be garbage collection. Is there a reason that
doesn't
work for you?
That acts on an individual ThreadLocal (and works quite well), but it
doens't allow removing all ThreadLocals that might have been
accumlated.
You're basically saying "This type of resource can leak if not cleared
appropriately, so there should be a 'Release all resources' method."
When paraphrased that way, does that make it clearer why it isn't a good
idea? It would be about the same as a "File.closeAll()" or a
"Socket.closeAll()" call. Extremely dangerous and only a crutch for not
doing the right thing to begin with.
Or a "collect all unused memory" call . Clearly, that's a crutch for not
keeping track of memory allocation properly in the first place. And the
fact that files and sockets are closed when a process exits is yet
another
crutch.
I'm with Daniel. Your code uses classes that you wrote, and
classes you got from somewhere else -- Sioux Unusuals, perhaps.
And if those classes use ThreadLocals for their own purposes, and
you blithely destroy them all, what happens then? Or what about
the class that invoked your code, passing an InheritableThreadLocal?
Is it a good idea to change parts of your invoker's state that you
don't understand, that you aren't even aware of?
Consider the use case again. I've written a container. It procures a
thread and lets third-party code run in it. Currently I have two choices:
1. Create a fresh thread each time. This kills all the ThreadLocals that
might be lying around, but doesn't perform very well.
2. Use a ThreadPool. This improves performance at the cost of letting
ThreadLocals packrat (which, in the worst case, holds entire ClassLoaders in
memory).
I'm greedy; I want the performance without the memory issues. Or, to say it
another way, when getting a Thread from a ThreadPool, it should be
completely indistinguishable whether it's a recycled or brand-new thread.