---

# Read JSON files and extract the prices
library("rjson")

# Read any files on disk
files <- list.files(path = "tmp", pattern = "*.json", full.names = TRUE)
stopifnot(length(files) > 0)

# Extract the interesting bits
extract <- function(f){
  json <- fromJSON(file = f)
  sapply(json$Data, function(x) { mean(c(x$open, x$high, x$low)) })[seq(1999)]
}

prices <- as.data.frame(sapply(files, extract))

# Add a random series
random_prices <- cumsum(rnorm(nrow(prices)))
random_prices <- random_prices + 8000
prices <- cbind(TURBO = random_prices, prices)

# Buy strategy
buy_now <- function(s) {
  tail(s, 1) < (mean(s) - (2 * sd(s)))
}

# Plot prices and linear regression
linear_regression <- function(name) {

  # Get prices for token
  y <- unlist(prices[name])
  x <- seq(length(y))

  # Create plot titles
  token <- sub("tmp/", "", sub(".json", "", name))
  main_title <- token
  sub_title <- paste("sd", sd(y))

  # Plot with calculated line
  spot <- y[length(y)]
  yield <- 1.02
  projection <- seq(spot, spot * yield, ((spot * yield) - spot) / length(y))
  plot(c(y, projection), type = "l", col = "black", main = main_title, sub = sub_title,
    xlab = "minutes", ylab = "GBP")

  # Add linear regression and grid
  relation <- lm(y ~ x)
  abline(relation, col = "orange")
  # grid(10, 10)

  # Initialise trade report
  trade_window_size <- 80
  r <- rep(NA, each = length(y))

  for (i in seq(length(y) - trade_window_size)) {
    trade_window <- y[i:(i+trade_window_size)]
    if (buy_now(trade_window))
      r[i + trade_window_size] = y[i + trade_window_size]
  }

  points(r, col = "red")
}

plots <- sapply(colnames(prices), linear_regression)