Provide a better error for similar methods with Dict

[timholy]

Fixes the following error:

julia> d = Dict(:a=>"a")
Dict{Symbol,String} with 1 entry:
  :a => "a"

julia> map(p->p.second, d) == ["a"]
ERROR: MethodError: no method matching similar(::Dict{Symbol,String}, ::Type{String})
Closest candidates are:
  similar(::SubArray{T,N,P,I,L}, ::Type{T}, ::Tuple{Vararg{Int64,N}}) at subarray.jl:52
  similar{T}(::Array{T,1}, ::Type{T}) at array.jl:122
  similar{T}(::Array{T,2}, ::Type{T}) at array.jl:123
  ...
 in _collect(::Dict{Symbol,String}, ::Base.Generator{Dict{Symbol,String},##1#2}, ::Base.EltypeUnknown, ::Base.HasLength) at ./array.jl:318
 in map(::Function, ::Dict{Symbol,String}) at ./abstractarray.jl:1673

Noticed from the PlotlyJS tests.

[tkelman]

isn't this counter to fc3cbc9 / the closing of #17714 ?

[JeffBezanson]

Yes. I would be against sometimes mapping a Dict to a Dict, and sometimes to an array. Also I believe the collect machinery assumes the container type doesn't depend on the element type --- as it runs, it might pick a new element type but keeps the container type.

[timholy]

Sounds good. Should this be a deprecation or an outright error? In other words, would you be OK with moving the array-creating version to deprecated.jl?

[JeffBezanson]

In theory it could be a deprecation, but the warning should not depend on
type inference, which might be more trouble than it's worth.
On Aug 11, 2016 2:28 PM, "Tim Holy" notifications@github.com wrote:

In test/dict.jl
#17968 (comment):

@@ -388,6 +388,11 @@ d = Dict('a'=>1, 'b'=>1, 'c'=> 3)
@test [d[k] for k in keys(d)] == [d[k] for k in eachindex(d)] ==
[v for (k, v) in d] == [d[x[1]] for (i, x) in enumerate(d)]

+# generators, similar
+d = Dict(:a=>"a")
+@test @inferred(map(identity, d)) == d
+@test @inferred(map(p->p.first=>p.second[1], d)) == Dict(:a=>'a')
+@test @inferred(map(p->p.second, d)) == Any["a"]

Sounds good. Should this be a deprecation or an outright error? In other
words, would you be OK with moving the array-creating version to
deprecated.jl?

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
https://github.com/JuliaLang/julia/pull/17968/files/3cfc74d47ddf7ffc9e9db32e3057ce648438fe1c#r74476669,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAtcbAYxqgPJaOFl02SP9tqPgBnZe1_Eks5qe2nPgaJpZM4JiQcr
.

[timholy]

Hmm, this is amusing:

julia> d = Dict(:a=>"abc", :b=>"def")
Dict{Symbol,String} with 2 entries:
  :a => "abc"
  :b => "def"

julia> ret = [p for p in d]
2-element Array{Pair{Symbol,String},1}:
 :a=>"abc"
 :b=>"def"

julia> ret = [p->p[2] for p in d]
2-element Array{##4#6,1}:
 #4
 #4

julia> ret = [v for (k,v) in d]
2-element Array{String,1}:
 "abc"
 "def"

julia> ret = [v->v[1] for (k,v) in d]
2-element Array{##10#12,1}:
 #10
 #10

EDIT: this is on master, not this PR.

[timholy]

Better?

[JeffBezanson]

Yes, this is better. Whether you get the error depends on inference though, so it might be a bit jarring.

[timholy]

Isn't that true of the current MethodError too?

[tkelman]

What's the status here, is this good to merge?

[JeffBezanson]

Yes, we might as well do this.