I believe typed comprehensions may do scalar conversion, but assume right from the start that the output array is of the specified type (so no array conversion).
On Oct 29, 2014, at 11:08 AM, Zenna Tavares <[email protected]> wrote: > Ok, thanks. > > Final quick question, does the syntax John used - A = ASCIIString[fun(i) for > i = 1:3] - have the same convert semantics, or is it different? > > On Wednesday, October 29, 2014 2:02:47 PM UTC-4, Stefan Karpinski wrote: > Assignment expressions always return the right hand side, not the left hand > side, so that chaining assignments don't produce weird results doing > something like > > y = v[i] = x > > With this rule, this is always equivalent to doing y = x; v[i] = x (and the > order doesn't matter). If assignment returned the left hand side, then it > would be equivalent to v[i] = x; y = v[i]. > > So in the first version, you're getting the result of [fun(i) for i = 1:3], > not the value that gets assigned to A. > > On Wed, Oct 29, 2014 at 1:57 PM, Zenna Tavares <[email protected]> wrote: > Ok, but then why does it return the Vector{Any}. The following two functions > give different results > > function makestring(fun) > A::Array{ASCIIString} = [fun(i) for i = 1:3] > end > > > function makestring(fun) > A::Array{ASCIIString} = [fun(i) for i = 1:3] > convert(Array{ASCIIString},A) > end > > I thought maybe it was a weird REPL thing, but it's not. > > On Wednesday, October 29, 2014 1:52:01 PM UTC-4, John Myles White wrote: > I'm pretty sure this sort of thing always works since type declarations on > variables behave like convert calls: > > julia> function foo() > a::Int64 = 0x01 > return a > end > foo (generic function with 1 method) > > julia> foo() > 1 > > -- John > > On Oct 29, 2014, at 10:50 AM, Zenna Tavares <[email protected]> wrote: > > > Also, the following runs but still returns a Vector{Any}. How is this > > possible? > > > > function makestring(fun) > > A::Array{ASCIIString} = [fun(i) for i = 1:3] > > end > > > > > > On Wednesday, October 29, 2014 1:45:51 PM UTC-4, Zenna Tavares wrote: > > As shown in the following example, I am getting differently typed arrays > > depending on where I use the list comprehension. > > > > f(i) = "$i" > > A = [f(i) for i = 1:3] > > function makestring(fun) > > A = [fun(i) for i = 1:3] > > end > > B = makestring(f) > > > > In this example A has type Vector{ASCIIString} while B has type > > Vector{Any}. What gives? And is there a workaround such that we get a more > > specific type? > > > > I understand there are some open issues related to this, but I am not sure > > if this case comes under what's already been discussed. > > > > > >
