Improve matcher isInstanceOf now that Dart has first class types.

[DartBot]

Originally opened as dart-lang/sdk#15381

This issue was originally filed by greg...@gmail.com

---

Current: new isInstanceOf<T>(T t)

New: isInstanceOf2(Type t). // But with a better name.

Consider adding:
  throwsType(Type t)

Also worth updating these docs:

"As types are not first class objects in Dart we can only approximate this test by using a generic wrapper class."

"Note that this does not currently work in dart2js; it will match any type, and isNot(new isInstanceof<T>()) will always fail. This is because dart2js currently ignores template type parameters."

https://api.dartlang.org/docs/channels/stable/latest/matcher/isInstanceOf.html

https://www.dartlang.org/articles/writing-unit-tests-for-pub-packages/

[DartBot]

<img src="https://avatars.githubusercontent.com/u/5479?v=3" align="left" width="48" height="48"hspace="10"> Comment by sethladd

---

Removed Type-Defect label.
Added Type-Enhancement, Area-UnitTest, Triaged labels.

[DartBot]

<img src="https://avatars.githubusercontent.com/u/17034?v=3" align="left" width="48" height="48"hspace="10"> Comment by kevmoo

---

Removed Area-UnitTest label.
Added Area-Pkg, Pkg-Unittest labels.

[DartBot]

<img src="https://avatars.githubusercontent.com/u/17034?v=3" align="left" width="48" height="48"hspace="10"> Comment by kevmoo

---

Removed Pkg-Unittest label.
Added Pkg-matcher label.

[DartBot]

<img src="https://avatars.githubusercontent.com/u/248818?v=3" align="left" width="48" height="48"hspace="10"> Comment by vicb

---

This is something I have implemented in Guinness as:

    bool matches(obj, Map matchState) =>
        mirrors.reflect(obj).type.isSubtypeOf(mirrors.reflectClass(_type));

see https://github.com/vsavkin/guinness/blob/5f9e291f9c0937a19ab278b0afe7c377bfae25fe/lib/src/common/unittest_backend.dart#L191 for the full Matcher

We have started to discuss this with Kevin: https://plus.google.com/u/0/+VictorBerchet/posts/bdDgAfUB9bW?cfem=1

I'm not clear on how this would affect the generated code size:

• we don't need more info as what the current impl (isInstanceOf<T>(T t)) uses, would there be a way to prevent the JS size to sky rocket ?
• even if there is no way to limit the JS size, would that be a pb - considering we are running tests ?

(I've not really considered JS & its code size for my use case. The reasoning being that if the tests pass with Dart, they should pass after dart2js, if not then this is a dart2js and not your own code issue. The tests are supposed to test your own code, not dart2js - still I'm all for testing with dart2js to help stabilize the tools as long as there is no constraint on my development flow).

[DartBot]

This comment was originally written by greg...@gmail.com

---

I was under the impression that this was already possible:

  bool isInstance(obj, Type type) => obj is type;
  
But it looks like this is possible - yet?

"malformed type: line 5 pos 18: using 't' in this context is invalid"

I wonder if there is a bug for implementing this - or if it is not possible. In the VM it must just be an int check. Not sure about dart2js.

[DartBot]

This comment was originally written by greg.lo...@gmail.com

---

Erm - "But it looks like this is NOT possible - yet?"

[DartBot]

<img src="https://avatars.githubusercontent.com/u/444270?v=3" align="left" width="48" height="48"hspace="10"> Comment by seaneagan

---

In issue dart-lang/sdk#10406 gbracha indicates that this will only be possible via mirrors. Matchers that use mirrors are currently put here:

http://www.dartdocs.org/documentation/matcher/0.11.1/index.html#matcher/matcher-mirror_matchers

Also this issue is a duplicate of issue #214 which was wrongly merged with issue dart-lang/sdk#8138.

[DartBot]

<img src="https://avatars.githubusercontent.com/u/444270?v=3" align="left" width="48" height="48"hspace="10"> Comment by seaneagan

---

Once this is fixed, all the custom throwsArgumentError etc. matchers could easily be removed / cleaned up, since you could just do throwsA(ArgumentError).

[nex3]

I believe there are no plans to allow is to work with first-class types, so this isn't feasible.