updates for 0.7

src/AverageShiftedHistograms.jl

@@ -1,6 +1,7 @@
 module AverageShiftedHistograms
 
-import UnicodePlots, Distributions, RecipesBase
+import UnicodePlots, RecipesBase
+using LinearAlgebra, Statistics
 import StatsBase: nobs, fweights
 export ash, ash!, extendrange, xy, xyz, nout, nobs, Kernels
 
@@ -15,7 +16,7 @@ Create a `LinSpace` of length `n` starting at `s` standard deviations below
 """
 function extendrange(y::AbstractVector, s = 0.5, n = 500)
     σ = std(y)
-    linspace(minimum(y) - s * σ, maximum(y) + s * σ, n)
+    range(minimum(y) - s * σ, stop = maximum(y) + s * σ, length = n)
 end
 
 include("kernels.jl")

src/bivariate.jl

@@ -1,4 +1,4 @@
-mutable struct Ash2{R1 <: Range, R2 <: Range, F1 <: Function, F2 <: Function}
+mutable struct Ash2{R1 <: AbstractRange, R2 <: AbstractRange, F1 <: Function, F2 <: Function}
     kernelx::F1
     kernely::F2
     rngx::R1               # rng for x
@@ -10,7 +10,7 @@ mutable struct Ash2{R1 <: Range, R2 <: Range, F1 <: Function, F2 <: Function}
     nobs::Int64
 
     function Ash2(kx::F1, ky::F2, rngx::R1, rngy::R2, mx::Int, my::Int) where
-            {F1 <: Function, F2 <: Function, R1 <: Range, R2 <: Range}
+            {F1 <: Function, F2 <: Function, R1 <: AbstractRange, R2 <: AbstractRange}
 
         v = zeros(Int, length(rngy), length(rngx))
         z = zeros(Float64, length(rngy), length(rngx))
@@ -19,12 +19,12 @@ mutable struct Ash2{R1 <: Range, R2 <: Range, F1 <: Function, F2 <: Function}
 end
 function Base.show(io::IO, o::Ash2)
     println(io, "Ash2")
-    f, l, s = round.((first(o.rngx), last(o.rngx), step(o.rngx)), 4)
+    f, l, s = round.((first(o.rngx), last(o.rngx), step(o.rngx)), digits=4)
     print(io, "X:")
     println(io, "  > edges  | $f : $s : $l")
     println(io, "    > kernel | $(o.kernelx)")
     println(io, "    > m      | $(o.mx)")
-    f, l, s = round.((first(o.rngy), last(o.rngy), step(o.rngy)), 4)
+    f, l, s = round.((first(o.rngy), last(o.rngy), step(o.rngy)), digits=4)
     print(io, "Y:")
     println(io, "  > edges  | $f : $s : $l")
     println(io, "    > kernel | $(o.kernely)")
@@ -33,8 +33,8 @@ function Base.show(io::IO, o::Ash2)
 end
 
 function Base.:(==)(o::Ash2, o2::Ash2) 
-    fns = fieldnames(o)
-    all(getfield.(o, fns) .== getfield.(o2, fns))
+    fns = fieldnames(typeof(o))
+    all(getfield.(Ref(o), fns) .== getfield.(Ref(o2), fns))
 end
 
 function _histogram!(o::Ash2, x::AbstractVector, y::AbstractVector)
@@ -87,7 +87,7 @@ end
 
 
 function ash(x::AbstractVector, y::AbstractVector;
-        rngx::Range = extendrange(x), rngy::Range = extendrange(y),
+        rngx::AbstractRange = extendrange(x), rngy::AbstractRange = extendrange(y),
         mx = 5, my = 5, kernelx = Kernels.biweight, kernely = Kernels.biweight)
     o = Ash2(kernelx, kernely, rngx, rngy, mx, my)
     _histogram!(o, x, y)
@@ -111,16 +111,16 @@ end
 xyz(o::Ash2) = (o.rngx, o.rngy, copy(o.z))
 nobs(o::Ash2) = o.nobs
 nout(o::Ash2) = nobs(o) - sum(o.v)
-function Base.mean(o::Ash2)
-    meanx = mean(o.rngx, fweights(vec(sum(o.v, 1))))
-    meany = mean(o.rngy, fweights(vec(sum(o.v, 2))))
+function Statistics.mean(o::Ash2)
+    meanx = mean(o.rngx, fweights(vec(sum(o.v, dims=1))))
+    meany = mean(o.rngy, fweights(vec(sum(o.v, dims=2))))
     [meanx; meany]
 end
-function Base.var(o::Ash2)
-    varx = var(o.rngx, fweights(vec(sum(o.v, 1))); corrected=true)
-    vary = var(o.rngy, fweights(vec(sum(o.v, 2))); corrected=true)
+function Statistics.var(o::Ash2)
+    varx = var(o.rngx, fweights(vec(sum(o.v, dims=1))); corrected=true)
+    vary = var(o.rngy, fweights(vec(sum(o.v, dims=2))); corrected=true)
     [varx; vary]
 end
-Base.std(o::Ash2) = sqrt.(var(o))
+Statistics.std(o::Ash2) = sqrt.(var(o))
 
 RecipesBase.@recipe f(o::Ash2) = o.rngx, o.rngy, o.z

src/univariate.jl

@@ -1,24 +1,24 @@
-mutable struct Ash{R <: Range, F <: Function}
+mutable struct Ash{R <: AbstractRange, F <: Function}
     density::Vector{Float64}    # ash estimate
     rng::R                      # range of x values
     counts::Vector{Int}         # histogram estimate
     kernel::F                   # kernel function
     m::Int                      # smoothing parameter
     nobs::Int                   # number of observations
-    function Ash(rng::R, kernel::F, m::Int) where {R<:Range, F<:Function}
+    function Ash(rng::R, kernel::F, m::Int) where {R<:AbstractRange, F<:Function}
         m > 0 || throw(ArgumentError("Smoothing parameter must be > 0"))
         new{R, F}(zeros(Float64, length(rng)), rng, zeros(Int, length(rng)), kernel, m, 0)
     end
 end
 function Base.show(io::IO, o::Ash)
     println(io, "Ash")
-    f, l, s = round.((first(o.rng), last(o.rng), step(o.rng)), 4)
+    f, l, s = round.((first(o.rng), last(o.rng), step(o.rng)), digits=4)
     println(io, "  > edges  | $f : $s : $l")
     println(io, "  > kernel | $(o.kernel)")
     println(io, "  > m      | $(o.m)")
     println(io, "  > nobs   | $(o.nobs)")
     x, y = xy(o)
-    inds = find(y)
+    inds = findall(x -> x != 0, y)
     print(io, UnicodePlots.lineplot(x[inds], y[inds]; grid = false))
 end
 
@@ -91,7 +91,7 @@ Ash objectes can be updated with new data, smoothing parameter(s), or kernel(s).
     ash!(obj, newx, newy; kw...)
 
 """
-function ash(x::AbstractArray; rng::Range = extendrange(x), m = ceil(Int, length(rng)/100), kernel = Kernels.biweight)
+function ash(x::AbstractArray; rng::AbstractRange = extendrange(x), m = ceil(Int, length(rng)/100), kernel = Kernels.biweight)
     o = Ash(rng, kernel, m)
     _histogram!(o, x)
     _ash!(o)
@@ -128,8 +128,8 @@ end
 Base.merge(o::Ash, o2::Ash) = merge!(copy(o), o2)
 Base.copy(o::Ash) = deepcopy(o)
 function Base.:(==)(o::Ash, o2::Ash) 
-    fns = fieldnames(o)
-    all(getfield.(o, fns) .== getfield.(o2, fns))
+    fns = fieldnames(typeof(o))
+    all(getfield.(Ref(o), fns) .== getfield.(Ref(o2), fns))
 end
 
 "return the range and density of a univariate ASH"
@@ -150,8 +150,8 @@ Statistics.quantile(o::Ash, p = [0, .25, .5, .75, 1]) = quantile(o.rng, fweights
 Statistics.std(o::Ash) = sqrt(var(o))
 
 function Base.extrema(o::Ash)
-    imin = findfirst(x -> x>0, o.counts)
-    imax = findlast(x -> x>0, o.counts)
+    imin = findnext(x -> x > 0, o.counts, 1)
+    imax = findlast(x -> x > 0, o.counts)
     o.rng[imin], o.rng[imax]
 end
 
@@ -160,7 +160,7 @@ end
 
 Return the estimated density at the point `x`.
 """
-function Distributions.pdf(o::Ash, x::Real)
+function pdf(o::Ash, x::Real)
     rng = o.rng
     y = o.density
     i = searchsortedlast(rng, x)
@@ -176,7 +176,7 @@ end
 
 Return the estimated cumulative density at the point `x`.
 """
-function Distributions.cdf(o::Ash, x::Real)
+function cdf(o::Ash, x::Real)
     cdf = cumsum(o.density) * step(o.rng)
     i = searchsortedlast(o.rng, x)
     if i == 0

test/runtests.jl

@@ -1,11 +1,11 @@
 module Tests
-using AverageShiftedHistograms, StatsBase, Base.Test, Distributions
+using AverageShiftedHistograms, Test, StatsBase, Statistics, LinearAlgebra
 
-info("Messy Output")
+@info("Messy Output")
 show(ash(randn(1000)))
 show(ash(randn(100), randn(100)))
 println("\n\n")
-info("Begin Tests")
+@info("Begin Tests")
 
 
 @testset "Kernels" begin
@@ -33,20 +33,20 @@ end
         @test f(o) ≈ f(x) atol=.1
     end
     AverageShiftedHistograms.histdensity(o)
-    @test quantile(o, 0) ≈ o.rng[findfirst(o.counts)]
+    @test quantile(o, 0) ≈ o.rng[findnext(x -> x != 0, o.counts, 1)]
 
     # check that histogram is correct
-    h = fit(Histogram, x, (-4:.1:4.1)-.05; closed = :left)
+    h = fit(Histogram, x, (-4:.1:4.1) .- .05; closed = :left)
     @test h.weights == o.counts
 
     ash!(o, x)
     @test nobs(o) == 20_000
-    @test pdf(o, -5) == 0
-    @test pdf(o, 5) == 0
-    @test pdf(o, 0) > 0
-    @test cdf(o, -5) == 0
-    @test cdf(o, 0) ≈ .5 atol=.05
-    @test cdf(o, 5) ≈ 1
+    @test AverageShiftedHistograms.pdf(o, -5) == 0
+    @test AverageShiftedHistograms.pdf(o, 5) == 0
+    @test AverageShiftedHistograms.pdf(o, 0) > 0
+    @test AverageShiftedHistograms.cdf(o, -5) == 0
+    @test AverageShiftedHistograms.cdf(o, 0) ≈ .5 atol=.05
+    @test AverageShiftedHistograms.cdf(o, 5) ≈ 1
 
     o = ash([.1, .1]; rng = -1:.1:1)
     @test nout(o) == 0