JDK 9 RFR of JDK-8159330: Improve deprecation text for Class.newInstance

[joe darcy]

Hi Roger,

On 6/13/2016 6:59 AM, Roger Riggs wrote:
> Hi,
>
> That example kind of glosses over the need to handle 3 new exceptions.
> That's been my annoyance with this change.
>
> throws InstantiationException, IllegalAccessException, 
> InvocationTargetException

The patch sent to the list 
(http://mail.openjdk.java.net/pipermail/core-libs-dev/2016-June/041754.html), 
did note addition exception types are thrown by the replacement code. 
The proposed text with Max's improvement is

+     * <p>The call
+     *
+     * <pre>{@code
+     * clazz.newInstance()
+     * }</pre>
+     *
+     * can be replaced by
+     *
+     * <pre>{@code
+     * clazz.getConstructor().newInstance()
+     * }</pre>
+     *
+     * The latter sequence of calls is inferred to be able to throw
+     * the additional exception types {@link
+     * InvocationTargetException} and {@link
+     * NoSuchMethodException}. Both of these exception types are
+     * subclasses of {@link ReflectiveOperationException}.
+     *

Class.newInstance is declared to throw InstantiationException and 
IllegalAccessException. Its implementation internally catches 
NoSuchMethodException and rethrows it as something else.

I considered adding more text to guide developers more explicitly to 
either use multi-catch

     catch(InstantiationException  | IllegalAccessException e) =>
     catch(InstantiationException  | IllegalAccessException | 
InvocationTargetException | NoSuchMethodException e)

or catching ReflectiveOperationException

     catch(InstantiationException  | IllegalAccessException e) =>
     catch(ReflectiveOperationException e) =>

With the more precise rethrow feature added in 7, catching 
ReflectiveOperationException and then rethrowing it doesn't lose any 
type information.

I was hesitant to add such text as to not overwhelm further the 
non-deprecated text of the method. However, if people think such 
guidance is warranted, I'll craft something.

Thanks,

-Joe


>
> Roger
>
>
>
> On 6/12/2016 11:43 PM, joe darcy wrote:
>>
>> On 6/12/2016 7:21 PM, Wang Weijun wrote:
>>> Why not just clazz.getConstructor().newInstance()?
>>
>> That seems to work to and is much shorter; thanks :-)
>>
>> -Joe
>>
>>>
>>>> +     * can be replaced by
>>>> +     *
>>>> +     * <pre>{@code
>>>> +     * clazz.getConstructor(new 
>>>> Class<?>[0]).newInstance((Object[])null);
>>>> +     * }</pre>
>>
>