Truth 0.33

Changes to existing Subjects and Core classes

• [API] Introduce ExpectFailure, a common mechanism for capturing failures in Truth, allowing tests to verify Truth failures without catching AssertionError. Intended primarily for testing Subject implementations.
• [API] Deprecate PrimitiveDoublesArraySubject.hasvaluesWithin() and
  PrimitiveFloatArraySubject.hasvaluesWithin().
• [SPI] Introduce AbstractFailureStrategy; users creating their own custom FailureStrategy implementations are encouraged to migrate to this class if possible.
• [SPI] Add Subject.ignoreCheck() which is similar to Subject.check() but passes a no-op failure strategy, to
  allow chaining to continue when non-short-circuiting failure strategies are used (such as Expect).
• Restrict visibility of the Platform class, which was never intended to be part of the public API. Most usages can simply be removed / inlined, usages that require GWT compatibility should open feature requests.

Extensions

• Protocol Buffers

  • Permit Iterables for field numbers, descriptors, and handles with the protocol buffers FieldScope APIs
  • Fix comparison rules for map fields.

Fixes

• Typo in LiteProtoTruth docs
• assertThat(Double.NaN).isNaN() in gwt/j2cl now working properly.
• Variety of improvements to error messages in MultimapSubject and MapSubject
• ThrowableSubject and Java8 Optional*Subject tests now use ExpectFailure which exposed some bad behavior, now fixed.
• IterableSubject adds type information to errors when string representations are the same, to clarify some
  ambiguous cases

Misc

• Internal clean-up of FailureStrategy usage to prepare for FailureStrategy becoming fully abstract
• Source now formatted via google-java-format
• For users of ErrorProne, add @SuppressWarnings("TruthSelfEquals") since Truth tests sometimes need to
  test equality against the same reference to prove behavior. Generally ErrorProne will flag that as an error,
  since it can never fail (and is therefore usually a badly written test)
• Some docs on OptionalSubject updated relating to complex assertions.
• Fix some sample code containing a bug.

For details, see the complete list of included commits.

Truth 0.32

New Subjects

• OptionalIntSubject (in java8 extension)
• StreamSubject (in java8 extension)

Changes to existing Subjects and Core classes

• add .hasValueThat() methods to primitive optionals which chain to the appropriate subject
• ThrowableSubject.hasMessageThat() which chains to StringSubject, in place of hasMessage(String)
• ThrowableSubject.hasCauseThat() which chains to ThrowableSubject around the cause
• make isInstanceof() and related methods abort hard, under -XdisableClassMetadata.
• Introduce DelegatedVerbFactory as a more powerful (but looser) alternative to SubjectFactory
• make MultimapSubject use usingCorrespondence() to allow custom "Fuzzy" equality for contains ops.
• Allow usingExactEquality on floating point subjects take Number subtypes, not just floating point types.
• Allow one-arg isEqualTo() to be used for primitive arrays of floating point types, implementing an agreed
  upon algorithm
• Add containsCell(Cell) and doesNotContainCell(Cell) to TableSubject.
• Include more readable hexadecimals when isEqualTo fails on a byte[].

New Extensions

• Protocol Buffers
  • Support for lite and heavy (with proto-reflection) protocol buffers
  • IterableProtoSubject

Fixes

• Make asList() chaining methods obey named()

Misc

• Code all formated with google-java-format
• some error-prone notified improvemetns

For details, see the complete list of included commits.

Truth 0.31

New Subjects

• SortedMapSubject and SortedSetSubject

Changes to existing Subjects and Core classes

• Add GWT support to Java8 subjects
• Lock down some of the Subject infrastructure, making several methods final
• Add @SafeVarargs to some methods where appropriate
• Depend on Guava 20.
• More use of usingTolerance() for relevant Subjects
• Various failure message improvements

New Extensions

• Protocol Buffers
  • Support for lite and heavy (with proto-reflection) protocol buffers
  • IterableProtoSubject

For details, see the complete list of included commits.

Truth 0.30

Pushing towards a 1.0 release

New Subjects

• OptionalSubject (for java.util.Optional, accessible via Truth8.assertThat(Optional<T>))
  (see details under "new extensions" below)

Changes to existing Subjects and Core classes

• Preliminary versions of "Fuzzy Equality" for doubles/floats. See DoublesSubject for
  new APIs, and we'll be updating the docs on fuzzy equality soon.

  • Subject implementers should note changes to the Service Provider Interface (SPI) for writing
    custom Subject subclasses:
    • getSubject() -> actual()
    • getDisplaySubject() -> actualAsString()
    • If you want to provide a subject-specific string representation of the type being tested, a new
      method actualCustomStringRepresentation() can be overridden. The deprecated
      getDisplaySubject() and getSubject() methods should not be overridden (and will be
      deleted in Truth 1.0. The renamed methods are final.
      actualAsString() has logic that not only consumes the string representation (default or custom)
      but also honors .named() and other contextual formatting.

• More evolution of "Fuzzy" truth - near-value approximation for certain types rather than strict
  equality is now supported by way of a "correspondence" mechanism, and the capabilities are
  being extended to many existing subjects.

  • To use:

        assertThat(actualIterable)
            .comparingElementsUsing(correspondence)
            .containsExactlyElementsIn(expectedIterable);

    Currently supports containsExactlyElementsIn(Iterable<E>), doesNotContain(E), and
    contains(E) methods. Subjects for Map, Iterable, and the Multimap types all support
    correspondence. For more detail, see the javadoc for IterableSubject.UsingCorrespondence and
    Correspondence

• .named() now supports a varargs/format construction, such as:

        assertThat(someBooleanFunction())
            .named("processed %s(%s)", foo, bar)
            .isTrue();

  This results in a more suitable error message: Not true that processed foo(bar) <false> is true as
  compared with the default message, and requires less string concatenation where that may prove
  awkward

New Extensions

• Add an extension for Java8 types, initially containing a subject forjava.util.Optional

  • Include "com.google.truth.extensions:truth-java8-extension:0.30" in your build dependencies

  • To use:

        import static com.google.common.truth.Truth8.assertThat;
        import java.util.Optional;
        public class MyTest {
          @Test public void testOptional() {
            Optional<String> o = Optional.of("Foo");
            assertThat(o).isPresent(); //succeeds
            assertThat(o).hasValue("Foo"); //succeeds
            assertThat(o).isEmpty(); // fails
          }
        }

Fixes

• Fix a missing @Nullable in AbstractVerb.that(T)
• Various cleanups of code, docs, readme, javadocs, and contribution

Truth 0.29

Pushing towards a 1.0 release

New Subjects

• AtomicLongMapSubject

Changes to existing Subjects and Core classes

• Preliminary versions of "Fuzzy Equality" for doubles/floats. See DoublesSubject for
  new APIs, and we'll be updating the docs on fuzzy equality soon.
• Prefer isEqualTo instead of equals since that is a method with a fairly precise
  meaning in Java, and we were hijacking it. .equals() shouldn't be called on Subject
  implementations.
• Multidimensional Array support in ObjectArraySubject
• Lots of renames and deprecations.
• SPI/API cleanup
  • marking a lot of subjects or their methods final, or noting where we can't
  • start making parts of the infrastructure more in line with the Open/Closed Principle
    (final methods, etc.)
  • stop storing "failure message" in the TestVerb, part 1

New Extensions

• An extension/contrib submodule, for things we want to ship with Truth,
  but which may need to be separate artifacts, mostly due to dependency
  issues or general bloat. Including our first entry: RE2J

  Support for MessageLite protocol buffers is also in extensions, but not released
  in 0.29.

Fixes

• Some cleanups of generics
• Message improvements:
  • Trim some stack trace of the obvious truth frames, to make it clearer where the
    locus of error is. (i.e. who needs to see Subject.failComparing() in the stack trace?)
  • ThrowableSubject uses string comparision where reasonable (to take advantage of
    JUnit's ComparisonFailure)
  • fix BooleanSubject's handling of null in failure messaging
  • other message improvements

Miscellaneous

• Apply @CheckReturnValue across a wide range of methods (on by default).
  • Users are _strongly_ advised to use error-prone in their builds, to ensure
    that this is checked, and the compiler errors out when people fail to call the
    following chained methods.
• build and continuous test system fixes
• more recent upstream dependencies
• full formating of the codebase using google-java-format
• various improvements from error-prone
• improvements to the examples
• more flesh out some holes in the tests

---

Note: various methods are being deprecated in preparation for 1.0. A penultimate
pre-1.0 release will be cut with those deprecated methods intact. Then a 1.0 will
be released with those methods removed. This should give people a chance to
get the 1.0 features and benefits, but have a nice transition phase.