SourceVersion::feature

[Joseph D. Darcy]

PS See https://github.com/openjdk/jdk/pull/5973

-Joe

On 10/14/2021 1:53 PM, Joseph D. Darcy wrote:
> On 10/14/2021 10:23 AM, Michael Bien wrote:
>> is this the right mailing list for javax.lang.model.* discussions?
>
>
> The compiler-dev list would be appropriate as well, but core-libs will 
> work.
>
> First, I understand the desire for a method like this. One of the 
> potential issues is SourceVersion models language versions and while, 
> to date, there has been a simple linear progression, that is not 
> guaranteed to be case arbitrarily far into the future, even though it 
> is the most likely outcome.
>
> Since java.lang.Runtime.Version is in the platform now, I think this 
> request would be better satisfied via a
>
>     public static SourceVersion valueOf(Runtime.Version version)
>
> method on SourceVersion. That way the modeling of SourceVersion can 
> absorb any non-linearity in versioning and presumably provide 
> sufficient information for a  Runtime.Version -> SourceVersion mapping 
> to query.
>
> I've filed the RFE
>
>     JDK-8275308: Add valueOf(Runtime.Version) factory to SourceVersion
>
> Cheers,
>
> -Joe
>
>
>>
>> if yes:
>>
>> instead of adding
>>     /**
>>      * Returns the version as int representing the feature release.
>>      * @see Runtime.Version#feature()
>>      */
>>     public int feature() {
>>         return this.ordinal();
>>     }
>> to SourceVersion.
>>
>> a note could be added to the doc that the ordinal can be used as 
>> feature version. Since this statement would apply to the past too, 
>> the note could be backported to all maintained JDKs. This comes with 
>> the usual downside of not being able to add enums for in-between 
>> versions in future.
>>
>> (doing both would be an option too of course)
>>
>>
>> To not use the ordinal, client code has to go through some enum init 
>> rituals to be able to do version comparisons:
>>
>>    final static SOURCE_VERSION_RELEASE_18;
>>    static {
>> SourceVersion tmp18;
>>         try {
>>             tmp18 = SourceVersion.valueOf("RELEASE_18");
>>         } catch (IllegalArgumentException ex) {
>> tmp18 = null;
>>         }
>>         SOURCE_VERSION_RELEASE_18 = tmp18;
>>         //... more versions
>>    }
>> just to be able to
>>
>>     if (SOURCE_VERSION_RELEASE_18 != null && 
>> version.compareTo(SOURCE_VERSION_RELEASE_18) >= 0) {}
>>
>> or
>>
>>     if 
>> (Integer.parseInt(version.name().substring(version.name().indexOf('_')+1)) 
>> >= 18) {}
>>
>> which is shorter but not a good solution either.
>>
>> what the client code actually wants is:
>>
>>     if (SourceVersion.latest().feature() >= 18) {}
>>
>>
>> it was a wise decision for java.lang.Runtime.Version to not use enums :)
>>
>> best regards,
>> michael
>>
>>
>>
>> On 09.10.21 20:58, Michael Bien wrote:
>>> Hello,
>>>
>>> could javax.lang.model.SourceVersion receive a feature() method 
>>> returning the version as an int, analog to java.lang.Runtime.Version?
>>>
>>> if (SourceVersion.latest().feature() >= 18) {}
>>>
>>> is simpler than comparing enums which may or may not exist dependent 
>>> on the deployed java version and cleaner than ordinal() tricks.
>>>
>>> best regards,
>>>
>>> michael
>>>