test/runtests.jl

using Artifacts
using DerekoVecs
using DataFrames
using Test


@testset "DerekoVecs.jl" begin

    wpd19 = load(joinpath(artifact"wpd19_10000", "wpd19_10000", "wpd19_10000.vecs"))

    @testset "DerekoVecs.jl: loading" begin
        @test wpd19.m == 200
        @test wpd19.n >= 10000
    end

    @testset "DerekoVecs.jl: similarities" begin
        @test isapprox(DerekoVecs.cos_sim(wpd19, "war", "war"), 1)
        @test cos_sim(wpd19, "wurden", "war") == cos_sim(wpd19, "war", "wurden")
        @test cos_sim(wpd19, "wurde", "wurden") > cos_sim(wpd19, "wurde", "ich")
        @test cos_sim(wpd19, "wurden", "war") == cos_sim(wpd19, wpd19, "war", "wurden")
        @test isapprox(cos_sim(wpd19, wpd19, "war"), 1)
        @test isapprox(cos_sim(wpd19, wpd19, 50), 1)
        @test isapprox(cos_sim(wpd19, 50, 50), 1)
    end

    @testset "DerekoVecs.jl: knn" begin
        @test "dieser" in knn(wpd19, "der", 3)
        @test "wurden" in knn(wpd19, "wurde", 3)
    end

    @testset "DerekoVecs.jl: kld" begin
        mykld = kld(wpd19, wpd19)
        @test mykld.common_type_count == length(wpd19.vocabdict)
        @test isapprox(mykld.common_type_share, 100)
        @test isapprox(mykld.kld, 0)
        @test wpd19.total_tokens == mykld.common_token_count
        @test isapprox(mykld.common_token_share, 100)
    end

    @testset "DerekoVecs.jl: load freq list only" begin
        wpd19_freqlist = load(joinpath(artifact"wpd19_10000", "wpd19_10000", "wpd19_10000.vocab"))
        @test wpd19.total_tokens == wpd19_freqlist.total_tokens
        @test isapprox(kld(wpd19_freqlist, wpd19).kld, 0)
    end

    @testset "DerekoVecs.jl: collocation analysis" begin
        if (!isnothing(wpd19.cdb))
            df = get_collocates(wpd19, "werden")
            @test df.collocate[1] == "kann"
            @test df.ldaf[1] > 10
            @test df.ldaf[1] > df.ldaf[3]
            @test nrow(get_collocates(wpd19, 3, 1)) == 1
            @test nrow(get_collocates(wpd19, 3, 2)) == 2
        end
    end

end