Hi Pavitar Singh,<br><br>I thank you very much about your explanations but actually I opened the issue myself<br>because I have read the same specs and am aware of that.<br>in H.Search (and several other hibernate code) there's this quite common pattern for starting<br>
"replaceable" objects (something like user-written plugins, you can provide your own implementation<br>to do some stuff) but this same pattern is also used to start the built-in default strategies.<br><br>It looks like this:<br>
- an empty constructor, to use class.newInstance();<br>- an initialize() to set configuration options<br>- a start() method (eventually) used to start background tasks<br>- some doStuff() and/or getXX() which need to be fast & threadsafe<br>
<br>As you can see in Concurrecy in Practice at page 50, this is BAD, as for<br>example in the FSSlaveDirectoryProvider nobody takes care of locking<br>or visibility, and nobody is doing anywhere where I see this pattern used<br>
(several times in the project).<br>I'm not saying it is all broken, because usually the threads consuming<br>these unsafely-initialized objects are started after the initialization, so<br>that's ok. In this specific case the state will be used to communicate<br>
between threads, so some visibility fix is needed.<br><br>I know I could use it only for final fields, but this is exactly what I want:<br>there are currently 10 instance variables, of these<br>4 have no concurrent use<br>
4 are configuration constants and could use the "final" (they're not safely published)<br>2 would need some lock/volatile anyway, bot only one is used frequently, so IMHO 1 volatile is ok.<br><br>I was thinking in using the same Piggyback technique you mention to<br>
safely publish the initialization constants,<br>but I'm afraid the code will become more difficult to maintain and more<br>"unreadable", possibly breaking at the first next patch:<br>IMHO using some unchanging fields "final" is the most clean and<br>
readable solution (and best performing), but I need a different<br>constructor for that.<br><br>your opinion is really welcome as I couldn't find other feedback,<br>if you would like to take a look at the sources download the Search trunk<br>
and look at:<br>org.hibernate.search.store.FSSlaveDirectoryProvider<br>or the FIXME in <br>org.hibernate.search.reader.SharingBufferReaderProvider<br><br>Sanne<br><br><div class="gmail_quote">2008/7/4 Pavitar Singh <<a href="mailto:pavitar.singh@gmail.com">pavitar.singh@gmail.com</a>>:<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">Hi Sanne,<br><br>I don't think moving everything in constructor can guarantee safe publication.<br>
<br>From the JMM Specification Section 3.5<br><br>"An object is considered to be completely initialized when its constructor finishes. A thread that can only see a reference to an object after that object has been completely initialized is guaranteed to see the correctly initialized values for that object's final fields."<br>
<br>That means there are no guarantees for other fields which are non-final.<br><br>But once things are moved in constructor then by using Safe Publication one can enforce visibility guarantees.<br><br>Other thing i was wondering was why do one need to make every field as volatile. As just by using a single volatile variable one can enforce memory barriers. A technique documented in JCIP as Piggyback Synchronizations and used by Doug Lea in implementing ConcurrentHashMap.(You will find get method in ConcurrentHashMap is without any locking and visibility is enforced by use of a single volatile variable.)<br>
<br>Also if you can elaborate more on how you are facing the visibility issue then may be i can also spend time on it on figuring performant solution.<br><br>Regards<br>Pavitar Singh<br><a href="http://pitfalls.wordpress.com" target="_blank">http://pitfalls.wordpress.com</a><br>
<br><div class="gmail_quote"><div><div></div><div class="Wj3C7c">On Fri, Jul 4, 2008 at 5:13 AM, Sanne Grinovero <<a href="mailto:sanne.grinovero@gmail.com" target="_blank">sanne.grinovero@gmail.com</a>> wrote:<br>
</div></div><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><div><div></div><div class="Wj3C7c">
Hello all,<br>I'm sorry I've been very busy but as promised I would like to fix HSEARCH-189<br>(and others) very soon;<br><br>I would like to propose a API extension (backwards compatible) that would simplify the patch a lot:<br>
keeping it as is it is very tricky to fix the visibility issues in FSSlaveDirectoryProvider<br>and FSMasterDirectoryProvider without introducing a performance penalty.<br><br>I have these options to close the issue:<br>1) add a "volatile" to more than six fields per class (ugly and not good for performance)<br>
2) use some Locks/synch (more readable, still performance hits)<br>3) move the "initialize" arguments to a constructor.<br><br>As Emmanuel knows I would really love the third option, but he's worried about<br>
the fact we can't force a constructor in an interface*1, so my proposal is:<br><br>if we find there exists a constructor having the same arguments as the initialize method,<br>we use that, otherwise we use a no-arg constructor and then call the initialize.<br>
<br>Reflection is used anyway to instantiate these components,<br>the code in DirectoryProviderFactory doesn't get much more complicated<br>and much more cleanup is made possible in all DPs because of this<br>(as the equals/hashcode comments already ask for).<br>
<br>I actually think this same pattern is needed for other components,<br>such as all ReaderProvider, so I hope you'll want to give it a try<br>and let me apply it on other components too if you like the resulting code.<br>
<font color="#888888">
<br>Sanne<br>
</font><br></div></div>_______________________________________________<br>
hibernate-dev mailing list<br>
<a href="mailto:hibernate-dev@lists.jboss.org" target="_blank">hibernate-dev@lists.jboss.org</a><br>
<a href="https://lists.jboss.org/mailman/listinfo/hibernate-dev" target="_blank">https://lists.jboss.org/mailman/listinfo/hibernate-dev</a><br>
<br></blockquote></div><font color="#888888"><br><br clear="all"><br>-- <br>Pavitar Singh<br>Blog: <a href="http://pitfalls.wordpress.com" target="_blank">http://pitfalls.wordpress.com</a>
</font></blockquote></div><br>