Find Upstream Links in Network

network <- kwb.graph::exampleNetwork(n_links = 25L, index = 13L)

# Define a function to plot the network (requires the "igraph" package)
plot_net <- function(network, ..., marked = NULL, seed = 1L) {
  
  if (! is.null(seed)) {
    set.seed(seed)
  }
  
  
  if (is.null(marked)) {
    edge.color <- "gray"
  } else {
    edge.color <- rep("lightgray", nrow(network))
    edge.color[marked] <- "red"
  }
  
  igraph::plot.igraph(
    x = igraph::graph_from_data_frame(network), 
    layout = igraph::layout.fruchterman.reingold, 
    vertex.size = 5, 
    vertex.label.cex = 0.8,
    vertex.label.dist = 1.5,
    edge.arrow.size = 0.5, 
    edge.color = edge.color,
    ...
  )
}

# Plot the two networks
plot_net(network)

# For each link, find all links upstream
us_links <- kwb.graph::getConnectedLinks(network)

# Mark the upstream links for three different start links
link_sets <- us_links[lengths(us_links) > 3L]

plot_net(network, marked = link_sets[[1]])

plot_net(network, marked = link_sets[[3]])

plot_net(network, marked = link_sets[[10]])

# Get the full example network that is stored in the package
network <- kwb.graph::exampleNetwork(n_links = -1L)

# For the comparison of run times, initialise vectors holding run times
runtime.R <- vector()
runtime.C1 <- vector()
runtime.C2 <- vector()

# Function to run code in "exp" and return elapsed time
elapsed <- function(exp) system.time(exp)["elapsed"]

# Shortcut to the main function. Arguments "resultSize" and "queueSize" are
# required by the C-versions of getConnectedLinks(), TODO: find proper values
# within getConnectedLinks()!
run <- function(version) kwb.graph::getConnectedLinks(
  network, 
  version = version, 
  resultSize = 2054851, 
  queueSize = 100*1024
)

# Number of repetitions for run time comparison
n <- 3L

# Compare run times of three different implementations of the "collect links
# upstream" algorithm within getConnectedLinks()
for (i in seq_len(n)) {
  cat("run", i, "/", n, "\n")
  runtime.R[i]  <- elapsed(x1 <- run(version = 1L))
  runtime.C1[i] <- elapsed(x2 <- run(version = 2L))
  runtime.C2[i] <- elapsed(x3 <- run(version = 3L))
}
#> run 1 / 3 
#> run 2 / 3 
#> run 3 / 3

(runtimeData <- data.frame(
  version = 1:3,
  implementation = c("R-functions", "C-functions(1)", "C-functions(2)"),
  mean_runtime = sapply(list(runtime.R, runtime.C1, runtime.C2), mean)
))
#>   version implementation mean_runtime
#> 1       1    R-functions    1.1076667
#> 2       2 C-functions(1)    0.3926667
#> 3       3 C-functions(2)    0.3920000

boxplot(mean_runtime ~ version, data = runtimeData)