Method ambiguity error

[marius311]

The following on 0.5 and on master gives a method ambiguity error:

module test

abstract foo{T<:Real}
immutable bar{T} <: foo{T} end

f(a::foo, b::foo) = 1
f{T<:bar}(a::T, b::T) = 2

f(bar{Int}(),bar{Int}()) # ambiguous

end

I do realize f{T<:bar}(a::T, b::T) = 2 instead of just f(a::bar,b::bar) is weird since bar is concrete, but strictly speaking its not wrong, hence I believe this is a bug.

A weird workaround is to put instead,

immutable bar{T<:Real} <: foo{T} end

which is redundant since subtyping foo{T} should guarantee T<:Real anyway, but nevertheless this seems to fix the problem.

Tested on:

julia> versioninfo()
Julia Version 0.5.0
Commit 3c9d753 (2016-09-19 18:14 UTC)
Platform Info:
  System: Linux (x86_64-pc-linux-gnu)
  CPU: Intel(R) Core(TM) i7-6700HQ CPU @ 2.60GHz
  WORD_SIZE: 64
  BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Haswell)
  LAPACK: libopenblas64_
  LIBM: libopenlibm
  LLVM: libLLVM-3.7.1 (ORCJIT, broadwell)

and

julia> versioninfo()
Julia Version 0.6.0-dev.2856
Commit 34b39f8* (2017-02-19 00:02 UTC)
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: Intel(R) Core(TM) i7-6700HQ CPU @ 2.60GHz
  WORD_SIZE: 64
  BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Haswell)
  LAPACK: libopenblas64_
  LIBM: libopenlibm
  LLVM: libLLVM-3.9.1 (ORCJIT, skylake)

in which case some of the syntax in my example is changed to reflect 0.6.

[yuyichao]

This is not a bug. T<:bar has to be concrete type so the second one cannot match f(::bar{Int}, ::bar{Int32}).

[yuyichao]

Note that this is actually the first example given in the paramatric method section in the doc http://docs.julialang.org/en/latest/manual/methods.html#Parametric-Methods-1 . Not sure if we have a coherent definition and implementation of the dispatch yet but with jb/subtypes we are getting much closer on that.

[marius311]

Sorry, I don't understand the explanation why this isn't a bug. What do you mean "T<:bar has to be concrete type" ?

[yuyichao]

And the fact that the "workaround" works is another bug #20056 (comment).

[yuyichao]

What do you mean "T<:bar has to be concrete type" ?

I mean the T in the method with T<:bar has to be a concrete type or (isleaftype). See the document I linked.

[yuyichao]

Sorry, actually I read the upper bound wrong, in that case it's a dup of #11597 and #20570