[julia-users] Re: Vector to column matrix

[Matt Bauman]

It'd be nice to allow up to one missing dimension in the dimension tuple, 
and calculate it automatically.  Something like this works nicely, where 
the missing dimension is specified by ():

function Base.reshape{N}(a::AbstractArray, dims::NTuple{N,Union(Int,())})
missing_mask = [isa(x,Tuple) for x in dims]
sum(missing_mask) == 1 || throw(DimensionMismatch(new dimensions 
$(dims) may only have up to one omitted dimension))
sz = length(a) / sum(dims[!missing_mask])
sz == int(sz) || throw(DimensionMismatch(array size $(length(a)) must 
be divisible by the product of the new dimensions $(dims[!missing_mask])))
reshape(a,map(x-isa(x,Tuple) ? int(sz) : x, dims))
end

Then: reshape(linspace(0, 10), ((), 1)) would do the trick.


On Tuesday, April 29, 2014 2:07:41 PM UTC-4, Tom Nickson wrote:

 Is there an built-in way to convert a vector to a column matrix without 
 transposing twice?
 I can make a matrix into a vector with vec or squeeze, but I can't see a 
 simple way to convert from a 1d array to a 2d without initialising it, 
 calling reshape and having to measure the size.

 I've put this definition in my .juliarc:

 colmat{T}(x::Array{T, 1}) = reshape(x, (length(x), 1))

 and do colmat(linspace(0, 10)), for example.

 Is there a better way?

[julia-users] Re: Vector to column matrix

[Matt Bauman]

Oops:  prod, not sum:

function Base.reshape{N}(a::AbstractArray, dims::NTuple{N,Union(Int,())})
missing_mask = [isa(x,Tuple) for x in dims]
sum(missing_mask) == 1 || throw(DimensionMismatch(new dimensions 
$(dims) may only have up to one omitted dimension))
sz = length(a) / prod(dims[!missing_mask])
sz == int(sz) || throw(DimensionMismatch(array size $(length(a)) must 
be divisible by the product of the new dimensions $(dims[!missing_mask])))
reshape(a,map(x-isa(x,Tuple) ? int(sz) : x, dims))
end

[julia-users] Re: Vector to column matrix

[Matt Bauman]

Ah, this was tried once before to get included in Base, and it was turned 
down.  Colon makes much more sense here (and works better, too). 
 https://github.com/JuliaLang/julia/pull/4263

One final iteration (I'll see if we can get this version into Base):

function Base.reshape{N}(a::AbstractArray, dims::NTuple{N, Union(Int, 
Colon)})
missing_mask = [isa(x, Colon) for x in dims]
sum(missing_mask) == 1 || throw(DimensionMismatch(new dimensions 
$(dims) may only have up to one omitted dimension))
sz = length(a) / prod(dims[!missing_mask])
sz == int(sz) || throw(DimensionMismatch(array size $(length(a)) must 
be divisible by the product of the new dimensions $(dims[!missing_mask])))
reshape(a, map(x-isa(x, Colon) ? int(sz) : x, dims))
end