# SVmatrix package
# Version 0.1.3
# About matrix http://svmatrix.online/
# Contact package author z@svmatrix.online


# Neccessary packages

# Installing (1 time)
install.packages("data.table")
install.packages("ggplot2")
install.packages("dplyr")
install.packages("circlize")
install.packages("RColorBrewer")

# Loading packages  (each time with R launch)
library(data.table)
library(ggplot2)
library(dplyr)
library(circlize)
library(readxl)


library("RColorBrewer") # for nicer colors, not neccessary


dataset <- read_excel("data_template.xlsx") # template file from the website

# Translate the dataset to fit the package
colnames(dataset)[1] <- "company" 
long <- melt(setDT(dataset), id.vars = c("company"), variable.name = "Year") # transform from wide to long

# Getting data to create matrix

svmatrix_prep <- function(data, # long dataset
                          years, # vector of years studied
                          share_column, # number of column with market share
                          year_column, # number of column with year
                          N # number of companies
                          ) {
  matrix_data <- data.frame(matrix(NA, 10, 5))
  colnames(matrix_data) <- c("Lind", "CRSV", "HTSV", "Quadrant", "Year")
  count = 1
  for (i in years) {
    colnames(data)[year_column] <- "Year"
    sv_data <- data%>%
      filter(as.numeric(as.character(Year)) == as.numeric(i))
    colnames(sv_data)[share_column] <- "Value"
    sv_vec <- c(t(sv_data$Value))
    sv_vector <- sort(sv_vec, decreasing = TRUE)
    Qsum <- NA
    CR_i <- cumsum(sv_vector)
    Qmat <- matrix(NA, nrow = N, ncol = N-1)
    for (m in c(1:N-1)) {
      for (l in c((m+1):(N))) {
        Qmat[l, m] <- (CR_i[m]/m)/((CR_i[l]-CR_i[m])/(l-m))
      }
      Qsum <- c(Qsum, sum(Qmat[m+1,], na.rm = TRUE))
    }
    Qsum <- Qsum[2:N]
    L <- c(rep(NA, N))
    for (m in c(2:N)) {
      L[m] <- Qsum[m]*(1/(m*(m-1)))
    }
    Lind <- NA
    
    for (m in c(2:N)) {
      if (L[m+1]>L[m]) {
        Lind = m
        break
      }
    }
    if (is.na(Lind) == TRUE) {
      break
    }
    CRSV <- sum(sv_vector[1:Lind])
    CR_norm <- sv_vector/sum(sv_vector[1:Lind])*c(1:N)
    HTn <- 1/(2*cumsum(CR_norm)-1)
    HTSVn <- (HTn-1/c(1:N))/(1-1/c(1:N))
    HTSV <- HTSVn[Lind]
    if (CRSV > 0.65 & HTSV > 0.1) {
      Quad = "G"
    }else{
      if (CRSV < 0.65 & HTSV > 0.1) {
        Quad = "I"
      }else{
        if (CRSV > 0.65 & HTSV < 0.1) {
          Quad = "B4"
        }else{
          Quad = "RO"
        }
      }
    }
    matrix_data[count,] <- as.numeric(c(as.character(Lind), as.character(CRSV), as.character(HTSV), NA, i))
    matrix_data[count, 4] <- Quad
    count = count + 1
  }
  count = 1
  matrix_data <- na.omit(matrix_data)
  return(matrix_data)
} # function to calculate data to build matrix
SV_data <- svmatrix_prep(long, years = c(2007:2021), share_column = 3, year_column =  2, 15)

# Building matrix (the current version throws an error, but the matrix is built)
svmatrix <- function(svmatrix_prep_dataset, # Dataset made by svmatrix_prep function, must be inputted
                     label_x_offset = -0.005,  # Year label offset on x axis, single number or vector with length of number of years
                     label_y_offset = 0.004, # Year label offset on y axis, single number or vector with length of number of years
                     legend_position = "topleft", # Legend position
                     legend_inset = c(-0.2, 0), # Legend offset (x, y)
                     labels.off = FALSE, # Turn off year labels, TRUE|FALSE
                     Lind.off = FALSE,  # Turn of Lind coefficients on points, TRUE|FALSE
                     legend.off = FALSE,  # Turn off legend, TRUE|FALSE
                     point_size = svmatrix_prep_dataset$Lind, # Size of points
                     B4_label = c(0.825, 0.03), # Position of B4 label on matrix
                     G_label = c(0.825, 0.3), # Position of G label on matrix
                     RO_label = c(0.475, 0.03), # Position of RO label on matrix
                     I_label = c(0.475, 0.3), # Position of I label on matrix
                     xlim =  c(0.3, 1),  # x axis limits
                     ylim = c(0.01, 1), # y axis limits
                     cex = 0.8, # Legend size
                     text.cex = 0.7, # Year labels size
                     bg_col = "white", # Background color
                     plot_col = "white", # Plot window color
                     quad_col = "grey", # Quadrant labels color
                     quad_size = 4, # Quadrant labels size
                     pch = 16, # Point type
                     point_col = 2:nrow(svmatrix_prep_dataset), # Point colors, vector (recommended to use pre-made palettes)
                     axis_font = "Arial", # Axes labels font (from Windows standart fonts)
                     labels_font = "Arial", # Year labels font (from Windows standart fonts)
                     axis_font_type = 2, # Axes labels font type
                     labels_font_type = 1, # Year labels font type
                     add_line = FALSE, # Add line successively connecting points by year, TRUE/FALSE 
                     line_col = "red", # Line color
                     line_size = 1,  # Line width
                     line_type = "dashed" # Line type
                     ){
  windowsFonts(A = windowsFont(axis_font))
  windowsFonts(B = windowsFont(labels_font))
  par(mar = c(5, 5, 4, 8), bg = bg_col)
  plot.new()
  rect(par("usr")[1], par("usr")[3],
       par("usr")[2], par("usr")[4],
       col = plot_col)
  par(new = TRUE)
  plot(x = 1, y = 1, log = "y", xlim = xlim, ylim =  ylim, col = plot_col, pch = 19,
       xlab="CRSV - доля отраслевого рынка, контролируемая группой доминирующих альфа-компаний",
       ylab="HTSV - уровень дифференициации между доминирующими в подотраслях альфами", family = "A", font = axis_font_type) +
    text(B4_label[1], B4_label[2], expression("B4"), cex = quad_size, col = quad_col) +
    text(G_label[1], G_label[2], expression("G"), cex = quad_size, col = quad_col) +
    text(RO_label[1], RO_label[2], expression("RO"), cex = quad_size, col = quad_col) +
    text(I_label[1], I_label[2], expression("I"), cex = quad_size, col = quad_col) + abline(v = 0.65) + abline(h = 0.1) +
    lines(svmatrix_prep_dataset$CRSV*(100*as.integer(add_line == FALSE)+1), svmatrix_prep_dataset$HTSV, col = line_col, lwd = line_size, lty = line_type)  +
    points(svmatrix_prep_dataset$CRSV, svmatrix_prep_dataset$HTSV, cex = point_size, pch = pch,
           col = point_col) +
    text(svmatrix_prep_dataset$CRSV+label_x_offset*point_size/2*(100*as.integer(labels.off == TRUE)+1), svmatrix_prep_dataset$HTSV+label_y_offset*point_size/2, 
         labels = svmatrix_prep_dataset$Year, cex = text.cex, family = "B", font = labels_font_type) +
    text(svmatrix_prep_dataset$CRSV*(100*as.integer(Lind.off == TRUE)+1), svmatrix_prep_dataset$HTSV, labels = svmatrix_prep_dataset$Lind, cex = max(0.25, point_size/8), col = "grey26") +
    legend(legend_position, inset = legend_inset+c(100*as.integer(legend.off == TRUE), 0), legend = c(svmatrix_prep_dataset$Year), cex = cex,
           pch = 19, title = "Year", col = 2:nrow(svmatrix_prep_dataset), xpd = TRUE)
} # Function to build SV matrix

svmatrix(SV_data, label_x_offset = c(rep(0.009, 2), rep(-0.009, 13)), label_y_offset = c(rep(-0.001, 4), rep(0.001, 11)), ylim = c(0.002, 0.1), xlim = c(0.4, 1), text.cex = 1, point_size = SV_data$Lind,
         legend_position = "topleft", legend_inset = c(0,100), cex = 0.75, point_col = brewer.pal(n = 15, name = "Dark2"),
         bg_col = "lightblue2", plot_col = "white", quad_col = "darkgrey", quad_size = 8, axis_font = "Arial", add_line = TRUE, line_col = "red", line_size = 2, pch = 19)