allOf should an option to list only the matcher that failed about javahamcrest

[EarthCitizen]

Can you provide a detailed example of the problem with what you are expecting and what you are getting?

[dcorbin]

Well, maybe. Consider this code sample:
import static org.hamcrest.CoreMatchers.allOf;
import static org.hamcrest.CoreMatchers.startsWith;
import static org.hamcrest.collection.IsIn.isOneOf;
import static org.junit.Assert.assertThat;

import org.hamcrest.Matcher;
import org.junit.Test;

public class MatcherExampleTest {
@SuppressWarnings(“unchecked”)
@test
public void testAll() {
final String value = “Red”;
final Matcher blueOrGreen = isOneOf(“Blue”, “Green”);
assertThat(value, allOf(startsWith(“R”), blueOrGreen));
}

@Test
public void testBad() {
    final String value = “Red”;
    final Matcher<String> blueOrGreen = isOneOf(“Blue”, “Green”);
    assertThat(value, blueOrGreen);
}

}

I would like the failure output for both of these tests to be the same.

The /reason/ is this: We have a large tree of "objects". They're really map-like and list-like. We're almost testing for equality, but there are few fields that are about equality. We've built a framework for assembling a tree of matchers where when the matcher fails, it describes the "tree-path" to that particular "field" that failed. We're not currently using #allOf. The reason (if my memory serves correctly) is that the Expected phrase is so overwhelming noisy to be of no value, but that's where we've been able to put the "tree-path" to identify the point of failure.

Our case might be an extremely non-standard way. Even our solution might be very bad. But when in the example above, it fails for being not being "Blue or Green", I don't care that I expect it start with "R".

(Please forgive that the example in testAll cannot be made to pass).

[valeraz]

In theory, describeMismatch should work for this case, but allOf describes only the first mismatch and returns: https://github.com/hamcrest/JavaHamcrest/blob/master/hamcrest-core/src/main/java/org/hamcrest/core/AllOf.java#L27.

Fore more user-friendly error messages, it would be more useful to aggregate all mismatches.

[tomwhoiscontrary]

See also the different but related #31 "AllOf diagnostics inaccurate".

There are two issues with allOf. Firstly, that the mismatch description is reliably incomprehensible. Secondly, that only the first mismatch is described.

There's some discussion in #31 pointing out that we have to be careful about verbosity, because when you compose verbose matchers, things can get out of hand.

However, if i could find a way to improve allOf without driving up the verbosity (maybe using @npryce's suggestion), would a PR be considered?

I realise that this comment should probably be a separate issue.