## Setup R error handling to go to stderr
options( show.error.messages=F,
         error = function () { cat( geterrmessage(), file=stderr() ); q( "no", 1, F ) } )
options(warn = -1)
library(RColorBrewer)
library(lattice)
library(latticeExtra)
library(grid)
library(gridExtra)
library(optparse)


option_list <- list(
  make_option(c("-i", "--ymin"), type="double", help="set min ylimit. e.g. '-100.0'"),
  make_option(c("-a", "--ymax"), type="double", help="set max ylimit. e.g. '100.0'"),
  make_option(c("-f", "--first_dataframe"), type="character", help="path to first dataframe"),
  make_option(c("-e", "--extra_dataframe"), type="character", help="path to additional dataframe"),
  make_option(c("-n", "--normalization"), type="character", help="space-separated normalization/size factors"),
  make_option("--first_plot_method", type = "character", help="How additional data should be plotted"),
  make_option("--extra_plot_method", type = "character", help="How additional data should be plotted"),
  make_option("--global", type = "character", help="data should be plotted as global size distribution"),
  make_option("--output_pdf", type = "character", help="path to the pdf file with plots")
)

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

# data frames implementation
## first table
Table = read.delim(args$first_dataframe, header=T, row.names=NULL)
if (args$first_plot_method == "Counts" | args$first_plot_method == "Size") {
  Table <- within(Table, Counts[Polarity=="R"] <- (Counts[Polarity=="R"]*-1))
}
n_samples=length(unique(Table$Dataset))
samples = unique(Table$Dataset)
if (args$normalization != "") {
  norm_factors = as.numeric(unlist(strsplit(args$normalization, " ")))
} else {
  norm_factors = rep(1, n_samples)
}
if (args$first_plot_method == "Counts" | args$first_plot_method == "Size" | args$first_plot_method == "Coverage") {
  i = 1
  for (sample in samples) {
    Table[, length(Table)][Table$Dataset==sample] <- Table[, length(Table)][Table$Dataset==sample]*norm_factors[i]
    i = i + 1
  }
}
genes=unique(Table$Chromosome)
per_gene_readmap=lapply(genes, function(x) subset(Table, Chromosome==x))
per_gene_limit=lapply(genes, function(x) c(1, unique(subset(Table, Chromosome==x)$Chrom_length)) )
n_genes=length(per_gene_readmap)
# second table
if (args$extra_plot_method != '') {
  ExtraTable=read.delim(args$extra_dataframe, header=T, row.names=NULL)
  if (args$extra_plot_method == "Counts" | args$extra_plot_method=='Size') {
    ExtraTable <- within(ExtraTable, Counts[Polarity=="R"] <- (Counts[Polarity=="R"]*-1))
  }
  if (args$extra_plot_method == "Counts" | args$extra_plot_method == "Size" | args$extra_plot_method == "Coverage") {
    i = 1
    for (sample in samples) {
      ExtraTable[, length(ExtraTable)][ExtraTable$Dataset==sample] <- ExtraTable[, length(ExtraTable)][ExtraTable$Dataset==sample]*norm_factors[i]
      i = i + 1
    }
  }
  per_gene_size=lapply(genes, function(x) subset(ExtraTable, Chromosome==x))
}

## functions
globalbc = function(df, global="", ...) {
  if (global == "yes") {
    bc <- barchart(Counts~as.factor(Size)|factor(Dataset, levels=unique(Dataset)),
                   data = df, origin = 0,
                   horizontal=FALSE,
                   col=c("darkblue"),
                   scales=list(y=list(tick.number=4, rot=90, relation="free", cex=0.5, alternating=T), x=list(rot=0, cex=0.6, tck=0.5, alternating=c(3,3))),
                   xlab=list(label=bottom_first_method[[args$first_plot_method]], cex=.85),
                   ylab=list(label=legend_first_method[[args$first_plot_method]], cex=.85),
                   main=title_first_method[[args$first_plot_method]],
                   layout = c(2, 6), newpage=T,
                   as.table=TRUE,
                   aspect=0.5,
                   strip = strip.custom(par.strip.text = list(cex = 1), which.given=1, bg="lightblue"),
                   ...
    )
  } else {
    bc <- barchart(Counts~as.factor(Size)|factor(Dataset, levels=unique(Dataset)),
                   data = df, origin = 0,
                   horizontal=FALSE,
                   group=Polarity,
                   stack=TRUE,
                   col=c('red', 'blue'),
                   scales=list(y=list(tick.number=4, rot=90, relation="free", cex=0.5, alternating=T), x=list(rot=0, cex=0.6, tck=0.5, alternating=c(3,3))),
                   xlab=list(label=bottom_first_method[[args$first_plot_method]], cex=.85),
                   ylab=list(label=legend_first_method[[args$first_plot_method]], cex=.85),
                   main=title_first_method[[args$first_plot_method]],
                   layout = c(2, 6), newpage=T,
                   as.table=TRUE,
                   aspect=0.5,
                   strip = strip.custom(par.strip.text = list(cex = 1), which.given=1, bg="lightblue"),
                   ...
    )
  }
  return(bc)
}
plot_unit = function(df, method=args$first_plot_method, ...) {
  if (exists('ymin', where=args)){
    min=args$ymin
  }else{
    min=''
  }
  if ((exists('ymax', where=args))){
    max=args$ymax
  }else{
    max=''
  }
  ylimits=c(min,max)
  if (method == 'Counts') {
    p = xyplot(Counts~Coordinate|factor(Dataset, levels=unique(Dataset))+factor(Chromosome, levels=unique(Chromosome)),
               data=df,
               type='h',
               lwd=1.5,
               scales= list(relation="free", x=list(rot=0, cex=0.7, axs="i", tck=0.5), y=list(tick.number=4, rot=90, cex=0.7)),
               xlab=NULL, main=NULL, ylab=NULL, ylim=ylimits,
               as.table=T,
               origin = 0,
               horizontal=FALSE,
               group=Polarity,
               col=c("red","blue"),
               par.strip.text = list(cex=0.7),
               ...)
    p=combineLimits(p)
  } else if (method != "Size") {
    p = xyplot(eval(as.name(method))~Coordinate|factor(Dataset, levels=unique(Dataset))+factor(Chromosome, levels=unique(Chromosome)),
               data=df,
               type='p',
               pch=19,
               cex=0.35,
               scales= list(relation="free", x=list(rot=0, cex=0.7, tck=0.5), y=list(tick.number=4, rot=90, cex=0.7)),
               xlab=NULL, main=NULL, ylab=NULL, ylim=ylimits,
               as.table=T,
               origin = 0,
               horizontal=FALSE,
               group=Polarity,
               col=c("red","blue"),
               par.strip.text = list(cex=0.7),
               ...)
  } else {
    p = barchart(Counts~as.factor(Size)|factor(Dataset, levels=unique(Dataset))+Chromosome, data = df, origin = 0,
                 horizontal=FALSE,
                 group=Polarity,
                 stack=TRUE,
                 col=c('red', 'blue'),
                 scales=list(y=list(rot=90, relation="free", cex=0.7), x=list(rot=0, cex=0.7, axs="i", tck=c(1,0))),
                 xlab = NULL,
                 ylab = NULL,
                 main = NULL,
                 as.table=TRUE,
                 par.strip.text = list(cex=0.6),
                 ...)
    p=combineLimits(p)
  }
  return(p)
}


## function parameters

#par.settings.firstplot = list(layout.heights=list(top.padding=11, bottom.padding = -14))
#par.settings.secondplot=list(layout.heights=list(top.padding=11, bottom.padding = -15), strip.background=list(col=c("lavender","deepskyblue")))
par.settings.firstplot = list(layout.heights=list(top.padding=-2, bottom.padding=-2),strip.background=list(col=c("lightblue","lightgreen")))
par.settings.secondplot=list(layout.heights=list(top.padding=-1, bottom.padding=-1),strip.background=list(col=c("lightblue","lightgreen")))
title_first_method = list(Counts="Read Counts", Coverage="Coverage depths", Median="Median sizes", Mean="Mean sizes", Size="Size Distributions")
title_extra_method = list(Counts="Read Counts", Coverage="Coverage depths", Median="Median sizes", Mean="Mean sizes", Size="Size Distributions")
legend_first_method =list(Counts="Read count", Coverage="Coverage depth", Median="Median size", Mean="Mean size", Size="Read count")
legend_extra_method =list(Counts="Read count", Coverage="Coverage depth", Median="Median size", Mean="Mean size", Size="Read count")
bottom_first_method =list(Counts="Coordinates (nbre of bases)",Coverage="Coordinates (nbre of bases)", Median="Coordinates (nbre of bases)", Mean="Coordinates (nbre of bases)", Size="Sizes of reads")
bottom_extra_method =list(Counts="Coordinates (nbre of bases)",Coverage="Coordinates (nbre of bases)", Median="Coordinates (nbre of bases)", Mean="Coordinates (nbre of bases)", Size="Sizes of reads")

## Plotting Functions

double_plot <- function(...) {
  page_height = 15
  rows_per_page = 10
  graph_heights=c(40,30,40,30,40,30,40,30,40,30,10)
  page_width=8.2677 * n_samples / 2
  pdf(file=args$output_pdf, paper="special", height=page_height, width=page_width)
  for (i in seq(1,n_genes,rows_per_page/2)) {
    start=i
    end=i+rows_per_page/2-1
    if (end>n_genes) {end=n_genes}
    if (end-start+1 < 5) {graph_heights=c(rep(c(40,30),end-start+1),10,rep(c(40,30),5-(end-start+1)))}
    first_plot.list = lapply(per_gene_readmap[start:end], function(x) update(useOuterStrips(plot_unit(x, par.settings=par.settings.secondplot), strip.left=strip.custom(par.strip.text = list(cex=0.5)))))
    second_plot.list = lapply(per_gene_size[start:end], function(x) update(useOuterStrips(plot_unit(x, method=args$extra_plot_method, par.settings=par.settings.firstplot), strip.left=strip.custom(par.strip.text = list(cex=0.5)), strip=FALSE)))
    plot.list=rbind(first_plot.list, second_plot.list)
    args_list=c(plot.list, list( nrow=rows_per_page+1, ncol=1, heights=unit(graph_heights, rep("mm", 11)),
                                 top=textGrob(paste(title_first_method[[args$first_plot_method]], "and", title_extra_method[[args$extra_plot_method]]), gp=gpar(cex=1), vjust=0, just="top"),
                                 left=textGrob(paste(legend_first_method[[args$first_plot_method]], "/", legend_extra_method[[args$extra_plot_method]]), gp=gpar(cex=1), vjust=0, hjust=0, x=1, y=(-0.38/4)*(end-start-(3.28/0.38)), rot=90),
                                 sub=textGrob(paste(bottom_first_method[[args$first_plot_method]], "/", bottom_extra_method[[args$extra_plot_method]]), gp=gpar(cex=1), just="bottom", vjust=2)
    )
    )
    do.call(grid.arrange, args_list)
  }
  devname=dev.off()
}


single_plot <- function(...) {
  width = 8.2677 * n_samples / 2
  rows_per_page=8
  graph_heights=c(rep(40,8),10)
  pdf(file=args$output_pdf, paper="special", height=15, width=width)
  for (i in seq(1,n_genes,rows_per_page)) {
    start=i
    end=i+rows_per_page-1
    if (end>n_genes) {end=n_genes}
    if (end-start+1 < 8) {graph_heights=c(rep(c(40),end-start+1),10,rep(c(40),8-(end-start+1)))}
    first_plot.list = lapply(per_gene_readmap[start:end], function(x) update(useOuterStrips(plot_unit(x, par.settings=par.settings.firstplot),strip.left=strip.custom(par.strip.text = list(cex=0.5)))))
    plot.list=rbind(first_plot.list)
    args_list=c(plot.list, list( nrow=rows_per_page+1, ncol=1, heights=unit(graph_heights, rep("mm", 9)),
                                 top=textGrob(title_first_method[[args$first_plot_method]], gp=gpar(cex=1), vjust=0, just="top"),
                                 left=textGrob(legend_first_method[[args$first_plot_method]], gp=gpar(cex=1), vjust=0, hjust=0, x=1, y=(-0.41/7)*(end-start-(6.23/0.41)), rot=90),
                                 sub=textGrob(bottom_first_method[[args$first_plot_method]], gp=gpar(cex=1), just="bottom", vjust=2)
    )
    )
    do.call(grid.arrange, args_list)
  }
  devname=dev.off()
}

# main

if (args$extra_plot_method != '') { double_plot() }
if (args$extra_plot_method == '' & !exists('global', where=args)) {
  single_plot()
}
if (exists('global', where=args)) {
  pdf(file=args$output, paper="special", height=11.69)
  Table <- within(Table, Counts[Polarity=="R"] <- abs(Counts[Polarity=="R"])) # retropedalage
  library(reshape2)
  ml = melt(Table, id.vars = c("Dataset", "Chromosome", "Polarity", "Size"))
  if (args$global == "nomerge") {
    castml = dcast(ml, Dataset+Polarity+Size ~ variable, function(x) sum(x))
    castml <- within(castml, Counts[Polarity=="R"] <- (Counts[Polarity=="R"]*-1))
    bc = globalbc(castml, global="no")
  } else {
    castml = dcast(ml, Dataset+Size ~ variable, function(x) sum(x))
    bc = globalbc(castml, global="yes")
  }
  plot(bc)
  devname=dev.off()
}