comprehension is indeed fine (and fast). should remember to read/search the
docs and issues more (as usual).
fill(function,dims) is also good.
wondering, if it might be productive to make fill warn or even err when
there is ambiguity and make versions:
`shallowfill`,`copyfill`,`deepcopyfill`. all this is unnecessary, but might
help for readability purposes.
On Sunday, November 29, 2015 at 3:49:19 PM UTC+2, Milan Bouchet-Valat wrote:
>
> Le dimanche 29 novembre 2015 à 04:21 -0800, Dan a écrit :
> > Since this situation comes up once in a while, it might be nice to
> > have a copyfill which does a copy on each cell. In code:
> >
> > function copyfill!{T}(a::Array{T}, x)
> > xT = convert(T, x)
> > for i in eachindex(a)
> > @inbounds a[i] = copy(xT)
> > end
> > return a
> > end
> >
> > copyfill(v, dims::Dims) = copyfill!(Array(typeof(v), dims), v)
> > copyfill(v, dims::Integer...) = copyfill!(Array(typeof(v), dims...),
> > v)
> >
> > And then we have:
> >
> > julia> m = copyfill(Int[1],3,3)
> > 3x3 Array{Array{Int64,1},2}:
> > [1] [1] [1]
> > [1] [1] [1]
> > [1] [1] [1]
> >
> > julia> push!(m[1,1],2)
> > 2-element Array{Int64,1}:
> > 1
> > 2
> >
> > julia> m
> > 3x3 Array{Array{Int64,1},2}:
> > [1,2] [1] [1]
> > [1] [1] [1]
> > [1] [1] [1]
> >
> > As some would expect. Is there a +1 on this?
> This has been discussed in depth here:
> https://github.com/JuliaLang/julia/pull/8759
>
> One issue is that sometimes copy() is enough, but sometimes you want
> deepcopy() instead (when the object itself contains references). Thus,
> a more general solution would be to take as first argument a function
> returning the value:
> https://groups.google.com/d/msg/julia-users/L5fXkHPduBo/UNoq6NTQOnEJ
>
> But as Stefan noted the comprehension form is quite good too. Maybe the
> docs should point to it after the warning about shared references.
>
>
> Regards
>
> > On Sunday, November 29, 2015 at 1:36:03 PM UTC+2, Kristoffer Carlsson
> > wrote:
> > > I guess the simplest would be:
> > >
> > > [Int[] for i = 1:3, j=1:3, k=1:3]
> > >
> > >
> > > And to repeat what Milan already said, you don't want fill! because
> > > then all your arrays point to the same memory location.
> > >
> > > On Sunday, November 29, 2015 at 11:51:28 AM UTC+1, Aleksandr
> > > Mikheev wrote:
> > > > Hi all. Once again I have some questions about Julia.
> > > >
> > > > I know that there is a possibility to create a list of arrays.
> > > > For exmple:
> > > >
> > > > s = fill(Array(Int64,1),4)
> > > >
> > > > And then I can do something like this:
> > > >
> > > > s[1] = [1; 2]
> > > > s[1] = [s[1]; 5]
> > > >
> > > > By parity of reasoning I did this:
> > > >
> > > > s = fill(Array(Int64,1),4,4,4)
> > > >
> > > > And it worked fine. But in both cases I had initial elements in s
> > > > (like when I construct arrays with Array{Int64}(m,n)):
> > > >
> > > > julia> s = fill(Array(Int64,1),3,3,3)
> > > > 3x3x3 Array{Array{Int64,1},3}:
> > > > [:, :, 1] =
> > > > [2221816832] [2221816832] [2221816832]
> > > > [2221816832] [2221816832] [2221816832]
> > > > [2221816832] [2221816832] [2221816832]
> > > >
> > > >
> > > > [:, :, 2] =
> > > > [2221816832] [2221816832] [2221816832]
> > > > [2221816832] [2221816832] [2221816832]
> > > > [2221816832] [2221816832] [2221816832]
> > > >
> > > >
> > > > [:, :, 3] =
> > > > [2221816832] [2221816832] [2221816832]
> > > > [2221816832] [2221816832] [2221816832]
> > > > [2221816832] [2221816832] [2221816832]
> > > >
> > > >
> > > > Is there something I could do to prevent this? I know I could
> > > > easily fix it by:
> > > >
> > > >
> > > > for i = 1:3
> > > > for j = 1:3
> > > > for k = 1:3
> > > > s[i,j,k] = []
> > > > end
> > > > end
> > > > end
> > > >
> > > > But I guess this is a weird solution.
> > > >
> > > > Thank you in advance!
> > > >
> > > >
>
