RFR [9] 8177738: Runtime.Version must be a value-based class

[Paul Sandoz]

> On 3 Apr 2017, at 16:40, Pavel Rappo <pavel.rappo at oracle.com> wrote:
> 
> 
>> On 3 Apr 2017, at 20:39, Paul Sandoz <paul.sandoz at oracle.com> wrote:
>> 
>> Hi,
>> 
>> 962 * [1-9][0-9]*((\.0)*\.[1-9][0-9]*)*
>> 
>> You removed the initial “^” which is still mentioned in JDK-8148822
>> 
>> < `^[1-9][0-9]*(((\.0)*\.[1-9][0-9]*)*)*$`. The sequence may be of arbitrary
>> ---
>>> `^[1-9][0-9]*((\.0)*\.[1-9][0-9]*)*$`. The sequence may be of arbitrary
>> 
>> JEP 223 is still using the former. Which is correct?
> 
> The patch from the thread [1] I referred to in my previous email removes ^ and $.
> So in this case I simply reused the patch and, once again [2], I do agree this
> is an odd place to use boundary matchers, given $VNUM is only a prefix in $VSTR
> and not the whole regex.
> 
> As for the JEP question, it's more for original designers/authors. If we agree
> on it I will update the JEP.
> 

Ok, no objections, i was just unsure as to which was the proposed authoritative answer. Perhaps for the record you should also update the description of JDK-8148822?


>> Since Version is now value based i think you need to check where it’s identity is used and replace with equals (which means ensuring that equals is efficient and could use identity under the covers). See usages in JarFile.
> 
> Will do, thanks!
> 
>> —
>> 
>> Separately i wish we could avoid using List<Integer> as the internal representation, it’s very inefficient, int[] is much better, the cost is then on the version() method but it’s easy to create an immutable List wrapping the array. (I logged an issue for that, but cannot remember the number.)
> 
> What are you concerns specifically?

It’s just not particularly efficient to box int to Integer into a List on the heap esp for the expected number of version components, and i suspect the version() method probably not something that will often be called, so taking a hit there with a AbstractList wrapper is fine. It’s quite a simple change, esp. with the the foundation you have set up, so seems worthwhile (separate issue, not this one).

Paul.


> 
> The startup time?
> The amount of simultaneously existing Version objects at runtime?
> The use of `Integer` wrapper?
> 
> --------------------------------------------------------------------------------
> [1] http://mail.openjdk.java.net/pipermail/core-libs-dev/2016-October/044277.html
> [2] http://mail.openjdk.java.net/pipermail/core-libs-dev/2017-March/046941.html
> 
>