## ----include = FALSE----------------------------------------------------------
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

## ----setup, message=FALSE-----------------------------------------------------
library(rgph)
library(dplyr)
library(tidyr)
library(ggplot2)

## ----example calculation------------------------------------------------------
ex_file <- system.file("extdata", "running_example.txt", package = "rgph")
ex_reeb <- read_reeb_graph(ex_file)
( ex_multi <- reeb_graph_pairs(ex_reeb, method = "multi_pass") )
( ex_single <- reeb_graph_pairs(ex_reeb, method = "single_pass") )
all.equal(ex_multi, ex_single)

## ----example validation-------------------------------------------------------
all.equal(ex_multi, ex_single, check.attributes = FALSE)

## ----example benchmark--------------------------------------------------------
bench::mark(
  multi = reeb_graph_pairs(ex_reeb, method = "multi_pass"),
  single = reeb_graph_pairs(ex_reeb, method = "single_pass"),
  check = FALSE
)

## ----flower data--------------------------------------------------------------
# print only a few edges for illustration
print(flower, n = 4)

## ----flower isolates----------------------------------------------------------
# print only a few pairs for illustration
print(reeb_graph_pairs(flower, method = "multi_pass"), n = 4)

## ----flower benchmark---------------------------------------------------------
flower <- rgph:::drop_reeb_graph_points(flower)
bench::mark(
  multi = reeb_graph_pairs(flower, method = "multi_pass"),
  single = reeb_graph_pairs(flower, method = "single_pass"),
  check = FALSE
)

## ----random (split) trees benchmark-------------------------------------------
# collect split tree Reeb graphs
tree_files <- vapply(
  c(
    `10` = "10_tree_iterations.txt",
    `100` = "100_tree_iterations.txt",
    `1000` = "1000_tree_iterations.txt"
  ),
  function(f) system.file("extdata", f, package = "rgph"),
  ""
)
tree_reebs <- lapply(tree_files, read_reeb_graph)
tree_reebs <- lapply(tree_reebs, function(rg) { rg$values <- -rg$values; rg })
# aggregate benchmark comparisons
tree_bench <- tibble()
for (i in seq_along(tree_reebs)) {
  bm <- bench::mark(
    multi = reeb_graph_pairs(tree_reebs[[i]], method = "multi_pass"),
    single = reeb_graph_pairs(tree_reebs[[i]], method = "single_pass"),
    check = FALSE
  )
  bm <- transmute(
    bm,
    method = as.character(expression),
    n_itr, time, memory
  )
  bm <- relocate(mutate(bm, size = as.integer(names(tree_files)[[i]])), size)
  tree_bench <- bind_rows(tree_bench, bm)
}

## ----random (split) trees plot, fig.width=8, fig.height=4, out.width="100%"----
# plot runtime results
tree_bench %>%
  select(size, method, time) %>%
  unnest(time) %>%
  ggplot(aes(x = as.factor(size), y = time * 1e3)) +
  geom_boxplot(aes(color = method, shape = method)) +
  scale_y_continuous(
    transform = "log1p",
    labels = scales::label_number(suffix = "ms")
  ) +
  labs(x = "tree size", y = "run time")

## ----random (split) trees medians---------------------------------------------
tree_bench %>%
  transmute(size, method, median = vapply(time, median, 0.)) %>%
  pivot_wider(id_cols = size, names_from = method, values_from = median) %>%
  transmute(size, ratio = single / multi)