require "rinruby"
require 'statsample'
#R.eval(".libPaths(\"C:/Program Files/R/R-3.5.0/library\")")
#R.eval(".libPaths(\"\\\\home.gu.gu.se/home-XB$/xbalek/Documents/R/win-library/3.5\")")
#R.eval(".libPaths(\"U:\\Documents\\R\\win-library\\3.5\")")
#R.eval(".libPaths(\"U:/Documents/R/win-library/4.0\")")

#R.eval(".libPaths(\"C:/Users/xbalek/Work Folders/Documents/R/win-library/3.3\")")
R.eval("library(\"irr\")")
R.eval("library(\"effectsize\")")

o = File.open("alpha_rho_analysis.tsv","w:utf-8")
o.puts "ncoders\tntexts\tmean_alpha_max\tmean_alpha_min\talpha_diff\talpha_wilcox_p\talpha_r_effect_size\tmean_rho_max\tmean_rho_min\trho_diff\trho_wilcox_p\trho_r_effect_size\tn_max\tn_min\tcoders\talphas_max\talphas_min\trhos_max\trhos_min"
def stats(input, type)
    if type == "hash"
        sent_array = input.values
    elsif type == "array"
        sent_array = input
    end
    sent_sum = 0.0
    sent_array.each do |sent|
        sent_sum += sent
    end
    mean = sent_sum/sent_array.length
    
    sumsq = 0.0
    sent_array.each do |sent|
        sumsq += (mean - sent)*(mean - sent)
    end
    sd = Math.sqrt(sumsq/sent_array.length)
    return mean, sd
end

def construct_combinations_length(source,length)
    #combs = Hash.new{|hash, key| hash[key] = Array.new}
    cands = source.permutation(length).to_a
    cands2 = []
    cands.each do |cand| 
        if !cands2.include?(cand.sort)
            cands2 << cand
        end 
    end
    #combs = cands2
    
    return cands2
end


sources = [[1,6,7,8,9,10,11],[6,7,8,9,10,11],[6,7,8,9,10],[6,7,8,11],[7,8,11]]
#sources = [[3,6,7,8,9,10,11]]
#sources = [[7,8,11]]
userids = []
STDOUT.puts "baseuser\tanotheruser\talpha\trho\tcomb"
sources.each do |source|
    ncoders = source.length
    
    f = File.open("sd_source#{source.join(";")}.tsv","r:utf-8")
    
    users = Hash.new{|hash, key| hash[key] = Array.new}
    
    f.each_line.with_index do |line, index|
        if index > 0
            line1 = line.strip.split
            
            for j in 5..5+source.length-1 
                #if source[0] == 1 and j == 5
                #    k = 1
                #else
                #    k = j
                #end
                k = j - 5
                users[userids[k].to_i] << line1[j].to_f
            end        
        else
            userids = line.strip.split("\t")[5..5+source.length-1]
            #STDERR.puts userids.join(";")
        end
    end
    #STDERR.puts users
    nvalues = users.values[0].length
    #STDERR.puts "#{nvalues} judgments in total"
    
    alphas_max = []
    rhos_max = []
    combs_max = construct_combinations_length(source,ncoders-1)
    combs_max.each do |comb|
        baseuser = ""
        source.each do |user|
            if !comb.include?(user)
                baseuser = user
                break
            end
        end
        
        basevalues = users[baseuser]
        sum_values = []
        ave_values = []
        comb.each do |user|
            users[user].each.with_index do |value, index|
                sum_values[index] = sum_values[index].to_f + value
            end
        end
        sum_values.each do |value|
            ave_values << value/(ncoders-1).to_f
        end
        #STDERR.puts "max"
        #STDERR.puts "base: #{basevalues.join(";")}"
        #STDERR.puts "ave: #{ave_values.join(";")}"
        R.assign "base", basevalues
        R.assign "ave", ave_values
        R.eval "m1 <- matrix(c(base,ave),nrow=2,byrow=TRUE)"
        alpha = R.pull "kripp.alpha(m1,\"interval\")$value"
        alphas_max << alpha
        rho = Statsample::Bivariate.spearman(Daru::Vector.new(basevalues), Daru::Vector.new(ave_values))
        rhos_max << rho
        pearson = Statsample::Bivariate::Pearson.new(Daru::Vector.new(basevalues), Daru::Vector.new(ave_values))
        pearson_r = pearson.r
        if source == [6,7,8,9,10,11]
            #STDOUT.puts "baseuser\talpha\trho\tr\tcomb"
            #STDOUT.puts "#{baseuser}\t#{alpha}\t#{rho}\t#{pearson_r}\t#{comb}"
        end
    end
    
    alphas_min = []
    rhos_min = []
    combs_min = construct_combinations_length(source,2)
    combs_min.each do |comb|
        baseuser = comb[0]
        anotheruser = comb[1]
        #STDERR.puts "min"
        #STDERR.puts "base: #{users[baseuser].join(";")}"
        #STDERR.puts "ave: #{users[anotheruser].join(";")}"
        
        R.assign "base", users[baseuser]
        R.assign "ave", users[anotheruser]
        R.eval "m1 <- matrix(c(base,ave),nrow=2,byrow=TRUE)"
        alpha = R.pull "kripp.alpha(m1,\"interval\")$value"
        alphas_min << alpha
        rho = Statsample::Bivariate.spearman(Daru::Vector.new(users[baseuser]), Daru::Vector.new(users[anotheruser]))
        rhos_min << rho
        #pearson = Statsample::Bivariate::Pearson.new(Daru::Vector.new(baseuser), Daru::Vector.new(anotheruser))
        #pearson_r = pearson.r
        if source == [6,7,8,9,10,11]
            STDOUT.puts "#{baseuser}\t#{anotheruser}\t#{alpha}\t#{rho}\t#{comb}"
        end
    end
    #STDERR.puts alphas_max.join(",")
    #STDERR.puts stats(alphas_max,"array")[0]
    #STDERR.puts alphas_min.join(",")
    #STDERR.puts stats(alphas_min,"array")[0]
    R.assign "alphas_max", alphas_max
    R.assign "alphas_min", alphas_min
    p = R.pull "wilcox.test(alphas_max,alphas_min)$p.value"
    r = R.pull "rank_biserial(alphas_max,alphas_min)$r"
    R.assign "rhos_max", rhos_max
    R.assign "rhos_min", rhos_min
    p_rho = R.pull "wilcox.test(rhos_max,rhos_min)$p.value"
    r_rho = R.pull "rank_biserial(rhos_max,rhos_min)$r"
    #STDERR.puts p
    #STDERR.puts r
    #o.puts "ncoders\tntexts\tmean_alpha_max\tmean_alpha_min\talpha_diff\talpha_wilcox_p\talpha_r_effect_size\tmean_rho_max\tmean_rho_min\trho_diff\trho_wilcox_p\trho_r_effect_size\tn_max\tn_min\tcoders\talphas_max\talphas_min\trhos_max\trhos_min"

    o.puts "#{ncoders}\t#{nvalues}\t#{stats(alphas_max,"array")[0]}\t#{stats(alphas_min,"array")[0]}\t#{stats(alphas_max,"array")[0]-stats(alphas_min,"array")[0]}\t#{p}\t#{r}\t#{stats(rhos_max,"array")[0]}\t#{stats(rhos_min,"array")[0]}\t#{stats(rhos_max,"array")[0]-stats(rhos_min,"array")[0]}\t#{p_rho}\t#{r_rho}\t#{alphas_max.length}\t#{alphas_min.length}\t#{source.join(",")}\t#{alphas_max.join(",")}\t#{alphas_min.join(",")}\t#{rhos_max.join(",")}\t#{rhos_min.join(",")}"
    #o.puts alphas_max.join(",")
    #o.puts 

    f.close 
    
    
    
end