# Usage:
# ~$ Rscript runtime_test.R

textplot <- function(...) {
    text <- unlist(list(...))
    if (length(text) > 20) {
        text <- c(text[1:17],
                  '   ...... (skipped, text too long) ......',
                  text[c(-1, 0) + length(text)])
    }

    plot(NA, xlim = c(0, 1), ylim = c(0, 1),
         bty = 'n', xaxt = 'n', yaxt = 'n', xlab = '', ylab = '')
        
    for (i in seq_along(text)) {
        text(0, 1 - (i / 20),
             text[[i]], pos = 4)
    }
}

# library(CVEpureR) # load CVE's pure R implementation
library(CVE) # load CVE

#' Writes log information to console. (to not get bored^^)
tell.user <- function(name, start, i, length) {
    cat("\rRunning Test (", name, "):",
        i, "/", length,
        " - elapsed:", format(Sys.time() - start), "\033[K")
}
#' Computes "distance" of spanned subspaces.
#' @param B1 Semi-orthonormal basis matrix
#' @param B2 Semi-orthonormal basis matrix
#' @return Frobenius norm of subspace projection matrix diff.
subspace.dist <- function(B1, B2){
    P1 <- tcrossprod(B1, B1)
    P2 <- tcrossprod(B2, B2)
    return(norm(P1 - P2, type = 'F'))
}

# Set random seed
set.seed(437)

# Number of simulations
SIM.NR <- 50L
# maximal number of iterations in curvilinear search algorithm
MAXIT <- 50L
# number of arbitrary starting values for curvilinear optimization
ATTEMPTS <- 10L
# set names of datasets
ds.names <- paste0("M", seq(7))
# Set used CVE method
methods <- c("simple", "weighted") # c("legacy", "simple", "linesearch", "sgd")

# Setup error and time tracking variables
error <- matrix(NA, SIM.NR, length(methods) * length(ds.names))
time <- matrix(NA, SIM.NR, ncol(error))
colnames(error) <- kronecker(paste0(ds.names, '-'), methods, paste0)
colnames(time) <- colnames(error)

# Create new log file and write CSV (actualy TSV) header.
# (do not overwrite existing logs)
log.nr <- length(list.files('tmp/', pattern = '.*\\.log'))
file <- file.path('tmp', paste0('test', log.nr, '.log'))
cat('dataset\tmethod\terror\ttime\n', sep = '', file = file)
# Open a new pdf device for plotting into (do not overwrite existing ones)
path <- paste0('test', log.nr, '.pdf')
pdf(file.path('tmp', path))
cat('Plotting to file:', path, '\n')

# only for telling user (to stdout)
count <- 0
start <- Sys.time()
# Start simulation loop.
for (sim in 1:SIM.NR) {
    # Repeat for each dataset.
    for (name in ds.names) {
        tell.user(name, start, (count <- count + 1), SIM.NR * length(ds.names))

        # Create a new dataset
        ds <- dataset(name)
        # Prepare X, Y and combine to data matrix
        Y <- ds$Y
        X <- ds$X
        data <- cbind(Y, X)
        # get dimensions
        k <- ncol(ds$B)

        for (method in methods) {
            dr.time <- system.time(
                dr <- cve.call(X, Y,
                    method = method,
                    k = k,
                    attempts = ATTEMPTS
                )
            )
            dr$B <- coef(dr, k)

            key <- paste0(name, '-', method)
            error[sim, key] <- subspace.dist(dr$B, ds$B) / sqrt(2 * k)
            time[sim, key] <- dr.time["elapsed"]

            # Log results to file (mostly for long running simulations)
            cat(paste0(
                    c(name, method, error[sim, key], time[sim, key]),
                    collapse = '\t'
                ), '\n',
                sep = '', file = file, append = TRUE
            )
        }
    }
}

cat("\n\n## Time [sec] Summary:\n")
print(summary(time))
cat("\n## Error Summary:\n")
print(summary(error))

boxplot(error,
    main = paste0("Error (Nr of simulations ", SIM.NR, ")"),
    ylab = "Error",
    las = 2
)
boxplot(time,
    main = paste0("Time (Nr of simulations ", SIM.NR, ")"),
    ylab = "Time [sec]",
    las = 2
)