Favourite Things

R packages & functions that make doing data science a joy based on usage across projects

Each project closes with a table summarising the R tools used. By aggregating the package and function usage across all projects, there’s an opportunity to:

• Spot the use of superseded functions like map_dfr

• Check for usage consistency, e.g. clock::date_today versus Sys.Date

• See where it would be most useful to keep an eye on package version updates

Since starting this in 2017, functions like tidyr’s spread and gather have been superseded by pivot_wider and pivot_longer. Newer packages have emerged like tidyclust, which brings cluster modelling to tidymodels (now used in Finding Happiness in ‘The Smoke’). bslib has brought improvements to the latest shiny app version embedded in Plots Thicken. The paletteer package has put it’s arms around the myriad palette packages out there. And scales’ cut_short_scale assisted with plot labelling.

Most recently, the latest versions of dplyr and purrr have presented a host of enhancements. For example, the .by argument in mutate and friends offers a neat alternative to group_by and ungroup in many situations. Joins have been enhanced (e.g. to support inequality conditions) with the addition of join_by to dplyr. And changes in the map_ family introduce list_rbind and associates with map_dfr, for example, now superseded.

library(conflicted)
library(tidyverse)
conflict_prefer_all("dplyr", quiet = TRUE)
library(tidytext)
library(rvest)
library(paletteer)
library(janitor)
library(glue)
library(ggwordcloud)
library(patchwork)
library(clock)
library(usedthese)

conflict_scout()

theme_set(theme_bw())

n <- 4
palette <- "harrypotter::hufflepuff"

cols <- paletteer_c(palette, n = n)

tibble(x = 1:n, y = 1) |>
  ggplot(aes(x, y, fill = cols)) +
  geom_col(colour = "white") +
  geom_label(aes(label = cols |> str_remove("FF$")), 
             size = 4, vjust = 2, fill = "white") +
  annotate(
    "label",
    x = (n + 1) / 2, y = 0.5,
    label = palette,
    fill = "white",
    alpha = 0.8,
    size = 6
  ) +
  scale_fill_manual(values = as.character(cols)) +
  theme_void() +
  theme(legend.position = "none")

Separation of tidyverse and non-tidyverse packages may be achieved by using the likes of tidyverse_packages which lists all packages in the tidyverse.

• tidyverse
• tidymodels
• tidyverts

tidy <-
  c(
    tidyverse::tidyverse_packages(),
    fpp3::fpp3_packages(),
    tidymodels::tidymodels_packages()
  ) |>
  unique()

used_here() has already been used in Quantum Jitter projects to generate a usage table at the foot of each project with the CSS class usedthese. used_there() may now be used to web-scrape all the tables with this class to aggregate package and function usage data.

base_packages <- c(
  "stats",
  "graphics",
  "grDevices",
  "utils",
  "datasets",
  "methods",
  "base"
)

used_df <-
  used_there("https://www.quantumjitter.com/project/") |>
  mutate(multiverse = case_match(
    Package,
    tidy ~ "tidy",
    base_packages ~ "base",
    .default = "special"
  ))

n_url <- used_df |> summarise(n_distinct(url)) |> pull()

pack_df <- used_df |>
  count(Package, multiverse, wt = n) |>
  mutate(name = "package")

fun_df <- used_df |>
  count(Function, multiverse, wt = n) |>
  mutate(name = "function")

packfun_df <- pack_df |>
  bind_rows(fun_df) |>
  arrange(desc(n)) |>
  mutate(
    packfun = coalesce(Package, Function),
    name = fct_rev(name),
    .by = name
  )

With the latest versions of dplyr and purrr, usage of group_by, ungroup and map_dfr has fallen away whilst usage of list_rbind and join_by has freshly emerged.

p1 <- packfun_df |>
  filter(name == "package") |> 
  ggplot(aes(fct_reorder(packfun, n), n, fill = multiverse)) +
  geom_col(show.legend = FALSE) +
  coord_flip() +
  geom_label(aes(label = n), hjust = "inward", size = 2, fill = "white") +
  scale_fill_manual(values = cols[c(4, 3, 1)]) +
  labs(x = NULL, y = NULL, 
       subtitle = glue("Package Usage"))

min_n <- 4

p2 <- packfun_df |>
  filter(name == "function", n >= min_n) |> 
  ggplot(aes(fct_reorder(packfun, n), n, fill = multiverse)) +
  geom_col() +
  coord_flip() +
  geom_label(aes(label = n), hjust = "inward", size = 2, fill = "white") +
  scale_fill_manual(values = cols[c(4, 3, 1)]) +
  labs(x = NULL, y = NULL, 
       subtitle = glue("Function Usage >= {min_n}"))

p1 + p2 +
  plot_annotation(
    title = glue("Favourite Things"),
    subtitle = glue(
      "Used Across {n_url} Projects as at ",
      "{date_format(date_today('Europe/London'), format = '%B %d, %Y')}"
    )
  )

This last code chunk generates the word cloud for use as the feature image for this project.

set.seed = 123

packfun_df |>
  mutate(angle = 45 * sample(-2:2, n(), 
                             replace = TRUE, 
                             prob = c(1, 1, 4, 1, 1))) |>
  ggplot(aes(
    label = packfun,
    size = n,
    colour = multiverse,
    angle = angle
  )) +
  geom_text_wordcloud(
    eccentricity = 1,
    grid_margin = 0.95,
    seed = 789
  ) +
  scale_size_area(max_size = 20) +
  scale_colour_manual(values = cols[c(4, 2, 1)]) +
  theme_void() +
  theme(plot.background = element_rect(fill = cols[3]))

R Toolbox

This project’s code too should be included in my “favourite things”.

used_here()

Package	Function
base	as.character[1], c[6], library[11], sample[1], unique[1]
clock	date_format[1], date_today[1]
conflicted	conflict_prefer_all[1], conflict_scout[1]
dplyr	arrange[1], bind_rows[1], case_match[1], coalesce[1], count[2], desc[1], filter[2], mutate[5], n[1], n_distinct[1], pull[1], summarise[1]
forcats	fct_reorder[2], fct_rev[1]
ggplot2	aes[7], annotate[1], coord_flip[2], element_rect[1], geom_col[3], geom_label[3], ggplot[4], labs[2], scale_colour_manual[1], scale_fill_manual[3], scale_size_area[1], theme[2], theme_bw[1], theme_set[1], theme_void[2]
ggwordcloud	geom_text_wordcloud[1]
glue	glue[4]
paletteer	paletteer_c[1]
patchwork	plot_annotation[1]
stringr	str_remove[1]
tibble	tibble[1]
tidymodels	tidymodels_packages[1]
tidyverse	tidyverse_packages[1]
usedthese	used_here[1], used_there[1]