Inconsistent missing property handling with java objects

[Attila Szegedi]

The intended behavior is to treat Java objects as if they were JS objects that had Object.seal() invoked on them (object is not extensible, its existing properties are non-configurable). Therefore, attempted assignment to a non-existent property in strict mode should fail with a TypeError, and in non-strict mode should silently do nothing. I consider this to be a bug in our implementation (that we aren't throwing TypeError in strict mode when attempting assignment to a non-existent property). However, getting the value of a non-existent property would still return undefined, as per <http://es5.github.io/#x8.12.3>.

Attila.

On Sep 3, 2014, at 5:55 PM, Mike Rettig <mike.rettig at gmail.com> wrote:

> How about this example which demonstrates the inconsistent behavior of js
> objects versus java objects?
> 
>        ScriptEngineManager engineManager =
>                new ScriptEngineManager();
>        final ScriptEngine nashorn =
> engineManager.getEngineByName("nashorn");
>        nashorn.eval("'use sctrict'");
>        nashorn.eval("var jsD={}");
>        nashorn.eval("jsD.badProp=1233");
>        nashorn.eval("print(jsD.badProp)");
> 
>        nashorn.eval("var d= new java.util.Date()");
>        nashorn.eval("d.badProp=4567");
>        nashorn.eval("print(d.badProp)");
> 
> 
> This prints:
> 
> 1233
> undefined
> 
> For the java object, the assignment is allowed to the invalid field, but
> retrieving that value returns 'undefined'.  This is inconsistent with
> javascript objects AND java objects.
> 
> Is the intended behavior for nashorn to treat java objects like js objects
> or like POJO's?
> 
> Mike
> 
> 
> 
> On Wed, Sep 3, 2014 at 10:32 AM, Bullotta, Rick <rick.bullotta at thingworx.com
>> wrote:
> 
>> One could make the case that the Rhino behavior was actually the incorrect
>> behavior.  A missing property in JavaScript is, by definition, undefined...
>> 
>> -----Original Message-----
>> From: nashorn-dev [mailto:nashorn-dev-bounces at openjdk.java.net] On Behalf
>> Of Mike Rettig
>> Sent: Wednesday, September 03, 2014 11:30 AM
>> To: Hannes Wallnoefer
>> Cc: nashorn-dev at openjdk.java.net
>> Subject: Re: Inconsistent missing property handling with java objects
>> 
>> There are a couple reasons why it is problematic for me. The java objects
>> are not dynamic objects. The interface is static. Allowing silent failures
>> leads to all sorts of time consuming and frustrating bugs. Nashorn doesn't
>> wrap the java objects, so it should treat them like java objects.
>> Referencing an unknown property on a java object should fail (as it did in
>> Rhino). The current behavior is the worst of both worlds. The java object
>> is not wrapped so there are no extension points (e.g. __noSuchProperty___),
>> but in the case of a missing property it behaves like a js object by
>> silently failing.
>> 
>> Anyone know the reasoning behind this design decision?
>> 
>> Thanks,
>> 
>> Mike
>> 
>> 
>> 
>> 
>> On Wed, Sep 3, 2014 at 10:00 AM, Hannes Wallnoefer <
>> hannes.wallnoefer at oracle.com> wrote:
>> 
>>> Hi Mike,
>>> 
>>> It was just a design decision to handle missing Java properties this
>>> way in Nashorn.
>>> 
>>> The behaviour is implemented in NashornBottomLinker.linkBean method.
>>> Unfortunately there's currently no way to configure Nashorn to behave
>>> the way you want it to.
>>> 
>>> Regards,
>>> Hannes
>>> 
>>> Am 2014-09-03 um 16:13 schrieb Mike Rettig:
>>> 
>>> With nashorn, it seems that the assignment and accessing of missing
>>>> properties on java objects is handled as 'undefined'.
>>>> 
>>>> Example:
>>>> 
>>>>         String engine = "nashorn";
>>>>         ScriptEngineManager engineManager =
>>>>                 new ScriptEngineManager();
>>>>         final ScriptEngine nashorn = engineManager.getEngineByName(
>>>> engine);
>>>>         nashorn.eval("var d=new java.util.Date();");
>>>>         nashorn.eval("d.badProp=1233");
>>>>         nashorn.eval("print(d.badProp)");
>>>> 
>>>> 
>>>> This will print 'undefined'. It doesn't fail on the 'badProp' on the
>>>> Date object. With Rhino, this would fail on the assignment of the bad
>>>> property with the following error:
>>>> 
>>>> "Exception in thread "main" javax.script.ScriptException:
>>>> sun.org.mozilla.javascript.internal.EvaluatorException: Java class
>>>> "java.util.Date" has no public instance field or method named "badProp".
>>>> (<Unknown source>#1) in <Unknown source> at line number 1"
>>>> 
>>>> I want the script to fail when the bad property is referenced.  I've
>>>> tried 'use strict', the rhino compatibility script, and attempting to
>>>> define __noSuchProperty__ on the java object. I can create a wrapper
>>>> object that simply delegates to the java object and throws if a
>>>> property is undefined, but that seems excessive.
>>>> 
>>>> Is this the correct behavior? Is there a way to work around this
>>>> behavior in nashorn?
>>>> 
>>>> Thanks,
>>>> 
>>>> Mike
>>>> 
>>> 
>>> 
>>