extracting(Function...) propagates Object instead of the common base type of all extractors

[xenoterracide]

Feature summary

So, I'm noticing that my attempts to validate that this toString() and getVersion are returning the same thing is kind of painful.

Example

// using semver4j, note this won't pass because my code subclasses and stuff even then, but compiling


      var v010 = Semver.coerce("v0.1.0");

     // asserthat(v010).matchesToString(...) nope doesn't exist

      assertThat(v010)
        .extracting(Semver::getVersion, Semver::toString)
        .allSatisfy(o -> {
          assertThat(o).isInstanceOf(String.class); // I guess chaining off this assertion won't convince allowance of matches
          if (o instanceof String s) { // but couldn't o already have determined we're a string?
            assertThat(s).matches("^0\\.1\\.0-SNAPSHOT-1-g\\p{XDigit}{7}$");
          }
        });

[joel-costigliola]

Have you given a try to extracting(String propertyOrField, InstanceOfAssertFactory) or asInstanceOf ?

Should look like

// ... import static InstanceOfAssertFactories.STRING
assertThat(v010)
        .extracting(Semver::getVersion, Semver::toString)
        .allSatisfy(o -> assertThat(o).asInstanceOf(STRING).matches("^0\\.1\\.0-SNAPSHOT-1-g\\p{XDigit}{7}$"));

[scordio]

What you're looking for is probably asInstanceOf(InstanceOfAssertFactory), as Joel already mentioned.

However, there could be a way to make extracting(Function...) and the following allSatisfy smarter, maybe with some tuning of the type parameters for cases where all the extractors produce a common type which is more specific than Object, with a strategy similar to #1551.

[xenoterracide]

both things sound great, hadn't heard of asInstanceOf before. Even less thinking about it though sounds like a better solution.

[joel-costigliola]

@xenoterracide shall we close this issue or is there something more to discuss?

[scordio]

I would keep it open to see if we can do something cheap like in #1551.

[xenoterracide]

^ same