#!/usr/bin/env Rscript

suppressPackageStartupMessages(library("optparse"))

option_list <- list(
        make_option(c("-i", "--input_dir"), action="store", dest="input_dir", help="IDEAS para files directory"),
        make_option(c("-o", "--output_dir"), action="store", dest="output_dir", help="PDF output directory")
)

parser <- OptionParser(usage="%prog [options] file", option_list=option_list)
args <- parse_args(parser, positional_arguments=TRUE)
opt <- args$options

create_heatmap<-function(data_frame, output_file_name) {
    # Plot a heatmap for a .para / .state combination
    # based on the received data_frame which was created
    # by reading the .para file.
    num_columns = dim(data_frame)[2];
    cat("num_columns: ", num_columns, "\n");
    num_rows = dim(data_frame)[1];
    cat("num_rows: ", num_rows, "\n");
    p = (sqrt(9 + 8 * (num_columns-1)) - 3) / 2;
    cat("9 + 8 * (num_columns-1): ", 9 + 8 * (num_columns-1), "\n");
    cat("sqrt(9 + 8 * (num_columns-1)): ", sqrt(9 + 8 * (num_columns-1)), "\n");
    cat("p: ", p, "\n");
    data_matrix = as.matrix(data_frame[,1+1:p] / data_frame[,1]);
    cat("dim(data_matrix)[1]: ", dim(data_matrix)[1], "\n");
    cat("dim(data_matrix)[2]: ", dim(data_matrix)[2], "\n");
    colnames(data_matrix) = colnames(data_frame)[1+1:p];
    cat("colnames(data_matrix): ", colnames(data_matrix), "\n");
    histone_marks = colnames(data_matrix);
    cat("histone_marks: ", histone_marks, "\n");
    rownames(data_matrix) = paste(1:num_rows-1, " (", round(data_frame[,1]/sum(data_frame[,1])*10000)/100, "%)", sep="");
    cat("rownames(data_matrix): ", rownames(data_matrix), "\n");
    # Open the output PDF file.
    pdf(file=output_file_name);
    # Set graphical parameters.
    par(mar=c(6, 1, 1, 6));
    # Create a vector containing the minimum and maximum values in data_matrix.
    min_max_vector = range(data_matrix);
    cat("min_max_vector: ", min_max_vector, "\n");
    # Create a color palette.
    my_palette = colorRampPalette(c("white", "dark blue"))(n=100);
    defpalette = palette(my_palette);
    # Plot the heatmap for the current .para / .state combination.
    plot(NA, NA, xlim=c(0, p+0.7), ylim=c(0, num_rows), xaxt="n", yaxt="n", xlab=NA, ylab=NA, frame.plot=F);
    axis(1, at=1:p-0.5, labels=colnames(data_matrix), las=2);
    axis(4, at=1:num_rows-0.5, labels=rownames(data_matrix), las=2);
    color = round((t(data_matrix) - min_max_vector[1]) / (min_max_vector[2] - min_max_vector[1]) * 100);
    cat("color: ", color, "\n");
    rect(rep(1:p-1, num_rows), rep(1:num_rows-1, each=p), rep(1:p, num_rows), rep(1:num_rows, each=p), col=color);
    histone_mark_color = t(col2rgb(terrain.colors(ceiling(p))[1:p]));
    cat("histone_mark_color: ", histone_mark_color, "\n");

    # Specify a color for common feature names like "h3k4me3".
    # These are histone marks frequently used to identify
    # promoter activities in a cell, and are often displayed
    # in shades of red.
    for(i in 1:length(histone_marks)) {
        cat("histone_marks[i]: ", histone_marks[i], "\n");
        if(regexpr("h3k4me3", tolower(histone_marks[i])) > 0) {
            histone_mark_color[i,] = c(255, 0, 0);
            cat("histone_mark_color[i]: ", histone_mark_color[i], "\n");
        }
        if(regexpr("h3k4me2", tolower(histone_marks[i])) > 0) {
            histone_mark_color[i,] = c(250, 100, 0);
        }
        if(regexpr("h3k4me1", tolower(histone_marks[i])) > 0) {
            histone_mark_color[i,] = c(250, 250, 0);
        }
        if(regexpr("h3k36me3", tolower(histone_marks[i]))>0) {
            histone_mark_color[i,] = c(0, 150, 0);
        }
        if(regexpr("h2a", tolower(histone_marks[i])) > 0) {
            histone_mark_color[i,] = c(0, 150, 150);
        }
        if(regexpr("dnase", tolower(histone_marks[i])) > 0) {
            histone_mark_color[i,] = c(0, 200, 200);
        }
        if(regexpr("h3k9ac", tolower(histone_marks[i])) > 0) {
            histone_mark_color[i,] = c(250, 0, 200);
        }
        if(regexpr("h3k9me3", tolower(histone_marks[i])) > 0) {
            histone_mark_color[i,] = c(100, 100, 100);
        }
        if(regexpr("h3k27ac", tolower(histone_marks[i])) > 0) {
            histone_mark_color[i,] = c(250, 150, 0);
        }
        if(regexpr("h3k27me3", tolower(histone_marks[i])) > 0) {
            histone_mark_color[i,] = c(0, 0, 200);
        }
        if(regexpr("h3k79me2", tolower(histone_marks[i])) > 0) {
            histone_mark_color[i,] = c(200, 0, 200);
        }
        if(regexpr("h4k20me1", tolower(histone_marks[i])) > 0) {
            histone_mark_color[i,] = c(50, 200, 50);
        }
        if(regexpr("ctcf", tolower(histone_marks[i])) > 0) {
            histone_mark_color[i,] = c(200, 0, 250);
        }
        state_color = get_state_color(data_matrix, histone_mark_color)[,2];
        cat("state_color: ", state_color, "\n");
    }
    rect(rep(p+0.2, num_rows), 1:num_rows-0.8, rep(p+0.8, num_rows), 1:num_rows-0.2, col=state_color);
    palette(defpalette);
    dev.off();
}

get_state_color <- function(data_matrix, histone_mark_color) {
    cat("XXX histone_mark_color: ", histone_mark_color, "\n");
    #histone_mark_color = rep("", dim(data_matrix)[2]);
    #cat("XXX histone_mark_color: ", histone_mark_color, "\n");
    #cat("XXX order(apply(data_matrix, 2, sd), decreasing=T): ", order(apply(data_matrix, 2, sd), decreasing=T), "\n");
    #cat("XXX dim(data_matrix)[2]-1: ", dim(data_matrix)[2]-1, "\n");
    #cat("XXX histone_mark_color: ", histone_mark_color, "\n");
    #histone_mark_color[order(apply(data_matrix, 2, sd), decreasing=T)] = hsv((1:dim(data_matrix)[2]-1) / dim(data_matrix)[2], 1, 1);
    #cat("XXX histone_mark_color: ", histone_mark_color, "\n");
    #histone_mark_color = t(col2rgb(histone_mark_color));
    #cat("XXX histone_mark_color: ", histone_mark_color, "\n");
    rg = apply(data_matrix, 1, range);
    cat("XXX rg: ", rg, "\n");
    mm = NULL;
    cat("XXX dim(data_matrix): ", dim(data_matrix), "\n");
    cat("XXX dim(data_matrix)[1]: ", dim(data_matrix)[1], "\n");
    for(i in 1:dim(data_matrix)[1]) {
        mm = rbind(mm, (data_matrix[i,] - rg[1, i] + 1e-10) / (rg[2, i] -rg[1, i] + 1e-10));
        cat("XXX mm: ", mm, "\n");
    }
    mm = mm^5;
    cat("XXX dim(mm): ", dim(mm), "\n");
    cat("XXX dim(mm)[2]: ", dim(mm)[2], "\n");
    if(dim(mm)[2] > 1) {
        mm = mm / (apply(mm, 1, sum) + 1e-10);
        cat("XXX mm: ", mm, "\n");
    }
    state_color = mm%*%histone_mark_color;
    cat("XXX state_color: ", state_color, "\n");
    s = apply(data_matrix, 1, max);
    cat("XXX s: ", s, "\n");
    s = (s - min(s)) / (max(s) - min(s) + 1e-10);
    cat("XXX s: ", s, "\n");
    state_color = round(255 - (255 - state_color) * s/0.5);
    cat("XXX state_color: ", state_color, "\n");
    state_color[state_color<0] = 0;
    cat("XXX state_color: ", state_color, "\n");
    rt = paste(state_color[,1], state_color[,2], state_color[,3], sep=",");
    cat("XXX rt: ", rt, "\n");
    h = t(apply(state_color, 1, function(x){rgb2hsv(x[1], x[2], x[3])}));
    cat("XXX h: ", h, "\n");
    h = apply(h, 1, function(x){hsv(x[1], x[2], x[3])});
    cat("XXX h: ", h, "\n");
    rt = cbind(rt, h);
    cat("XXX rt: ", rt, "\n");
    return(rt);
}

# Read the inputs.
para_files <- list.files(path=opt$input_dir, pattern="\\.para$", full.names=TRUE);
for (i in 1:length(para_files)) {
    para_file <- para_files[i];
    para_file_base_name <- strsplit(para_file, split="/")[[1]][2]
    output_file_name <- gsub(".para", ".pdf", para_file_base_name)
    output_file_path <- paste(opt$output_dir, output_file_name, sep="/");
    data_frame <- read.table(para_file, comment="!", header=T);
    create_heatmap(data_frame, output_file_path);
}