Better @assert default failure message, like that given by @test

[oxinabox]

when you fail a @test, you get a great message telling you what the expect value was and what the actual value was.
With @assert, you do not.

For example

julia> using Base.Test

julia> a=(2,4)
(2,4)

julia> @test a[1]==3
Test Failed
  Expression: a[1] == 3
   Evaluated: 2 == 3
ERROR: There was an error during testing
 in record(::Base.Test.FallbackTestSet, ::Base.Test.Fail) at ./test.jl:397
 in do_test(::Base.Test.Returned, ::Expr) at ./test.jl:281

julia> @assert a[1]==3
ERROR: AssertionError: a[1] == 3

I often fine myself writing things like:

julia> @assert(a[1]==3, "a[1]==$(a[1]) != 3")
ERROR: AssertionError: a[1]==2 != 3

but @test shows that this can be done automatically.
I think that makes a better default, than just printing out what the expression was, without showing the values.

[TotalVerb]

Just curious: why not simply use Test.@test?

[oxinabox]

They are not semantically the same operation.
That feels rather dirty, but it would work, yes.

But eventually @asserts will be (to some lesser or greater extent),
be removed when running in some form of optimized mode.
#7732
#10614

Also @test throws the error on the wrong line -- it alway throws an error from ./test.jl:397
when it fails to find a outer testset to eat the Error.

[StefanKarpinski]

I agree with this general direction – if we're going to keep asserts. There's the question of whether we should or not, however, since people tend to abuse asserts for error checking instead of strictly using them for verifying the internal logic of their algorithms when testing.

[samoconnor]

Related #15495

[al6x]

+1 for printing the actual value, currently you see message like

AssertionError: base ≈ 12

and you need to add println(base) and run it again before you can fix the code.

[Wikunia]

Maybe there can be a different macro for this behavior otherwise @StefanKarpinski or should one always do using Test; @test for this usecases?