var and method reference - unnecessary cast warning (missing?)

[carstenartur]

eclipse.buildId=4.33.0.20240711-1200 java.version=22.0.1 java.vendor=Eclipse Adoptium BootLoader constants: OS=macosx, ARCH=x86_64, WS=cocoa, NL=de_DE

Working on a bug in a jdt.ui cleanup to make use of "var" I experienced some strange behavior. See https://github.com/eclipse-jdt/eclipse.jdt.ui/issues/1524

    public interface MyIntf  {
        void bla(Object... o);
    }
    /**
     * @param o
     */
    private  void bla(Object... o) {

    }

    private void blubb() {
        new ArrayList<Object>().forEach((o) -> bla(o));
        var myin=(MyIntf)this::bla;
    }

When I mark the right side of the line with "var" and extract as lokal variable I get this:

        MyIntf myIntf= (MyIntf)this::bla;
        var myin=myIntf;

In the resulting line the cast is not marked as unneeded. I can remove it without getting an error:

        MyIntf myIntf= this::bla;
        var myin=myIntf;

The question is: is this to be expected?

[stephan-herrmann]

[...]

      MyIntf myIntf= (MyIntf)this::bla;
      var myin=myIntf;

In the resulting line the cast is not marked as unneeded. I can remove it without getting an error:

      MyIntf myIntf= this::bla;
      var myin=myIntf;

The question is: is this to be expected?

I'm afraid, yes. This situation is not covered by normal detection of unnecessary casts, because:

• a reference expression like this::bla does not have a type on its own, so we cannot compare provided type and cast type. The expression can only be inferred to conform to a given target type.
• in the version with cast, the target type is provided by the cast, and this makes inference succeed. Ergo the cast was used to find the inference solution.

If deemed relevant, someone would have to implement an additional analysis, that detects when cast type and the expected type at its location are the same type.

[carstenartur]

Don't know if such additional analysis is necessary. As the cleanup that exchanges explicit type by var causes a "Method reference needs an explicit target-type" error we exclude var from methods references now in the cleanup. That is the right thing to do, isn't it?

[carstenartur]

Please advise, should I close this issue?

[stephan-herrmann]

As the cleanup that exchanges explicit type by var causes a "Method reference needs an explicit target-type" error we exclude var from methods references now in the cleanup.

"we" == the cleanup implementation in jdt.ui?

That is the right thing to do, isn't it?

Sure, anything that removes the target type of a method reference or lambda breaks compilation.

As you're not interested in additional cast-analysis, I'm closing this issue.

[carstenartur]

As the cleanup that exchanges explicit type by var causes a "Method reference needs an explicit target-type" error we exclude var from methods references now in the cleanup.

"we" == the cleanup implementation in jdt.ui?

Yes, the proposed fix from Jeff

https://github.com/eclipse-jdt/eclipse.jdt.ui/pull/1541

That is the right thing to do, isn't it?

Sure, anything that removes the target type of a method reference or lambda breaks compilation.

Ok

As you're not interested in additional cast-analysis, I'm closing this issue.

Hope you do not get it wrong. I had a concrete issue in the cleanup that I would like to see fixed and doubt that I can do the job to implement the additional analysis.

Thanks for sharing your thoughts. That is really helpful.