In my experience, the following are true: 1. Usage of any object has a hash key whose equals() result depends on external or mutable factors should never be used as a hash key for any reason. 2. Any object class for which, when given two instances "a" and "b", a == b is true but a.equals(b) is false is incorrectly implementing .equals(). 3. Any object whose equals() or hashCode() method implementation performs network activity (e.g. a DNS request) is defective by design.
This includes java.net.URL as well as java.net.Proxy and any other JDK classes which have these unfortunate design errors. On 01/12/2011 10:43 AM, Chris Hegarty wrote: > Are you referring to java.net.Proxy? Proxy.equals depends on the > resolution of its address. Address resolution could change over time, > depending on the caching policy. java.net.URL has a similar issue. > > Strangely,I would have expected containsValue and containsKey to behave > in a similar fashion. I think the specification for these methods is > very clear, they should use equals(). I'm not sure if containsKey is > correct if it accepts key equality. > > -Chris > > On 01/12/11 03:53 PM, Jing LV wrote: >> Hello everyone, >> >> (As I am new in OpenJDK, please tell me if I post on the wrong >> mailing-list). >> I find an odd behaviour of HashMap. If we put a value which is an >> instance of Proxy, it would return false while checking >> containsValue(proxyinstance). I checked use Proxy instance as a key, it >> works well. >> I suppose this is a bug of HashMap. By reading the code closely, I find >> the problem was that, for Proxy instance, Proxy.equals(Proxy) may return >> false (there are many discussions on this already). >> HashMap.containsKey() will check if there is a key in the map equals >> and/or the same with the given key(so it bypass equals check) with >> getEntry() method but HashMap.containsValue() does not. So a quick fix >> will be: check there is a value in the map that equals and/or the same >> with the given value. >> >> I'd like to report this bug to bugzilla of openjdk, with my testcase and >> patch. Please tell me if any comments. Thanks a lot. >> >> Best wishes! >> Jing LV >> -- - DML
