Effectively provide an AbstractArray view of a generator or iterator.
That means it will be possible to pass a generator to methods expecting
arrays. The only exported function is called array and array(g) can be
used like a read-only abstract array.
The restrictions to the generators and iterators are that they need to have
a size.
Usage:
using GeneratorArrays
a = array(x^2 for x in 0:0.01:1)
a[1] == 0
a[51] ≈ 0.25
a[101] == 1
For example you can exploit the binary-search functionality of
searchsorted without evaluating the function at all grid points.
using GeneratorArrays
n = 10^6
a = randn(ComplexF64, n);
p = sortperm(a, by=real);
# positions in a[p] where numbers in corresponding vertical stripes are found
res = Int[]
for h in -3:0.1:3
x = searchsortedfirst(array(real(a[p[i]]) for i in 1:n), h)
push!(res, x)
end
mysin(x) = begin println(x); sin(x); end
# finding inverse value of a monotonic function
searchsortedfirst(array(mysin(x) for x in 0:0.001:3.14), 0.5)
using Primes
# first prime greater than 10^6 ( note: primes(10^6, 10^6+100) is much better)
searchsortedfirst(array(prime(i) for i = 5*10^4:10^5), 10^6)