.dev/_BENCHMARK_RESHAPE.R

In easystats/datawizard: Easy Data Wrangling and Statistical Transformations

library(tidyr)
library(dplyr)
library(datawizard)

### DATA_TO_LONG ==========================================


# SLOW (5M rows)

wide_data <- data.frame(replicate(5, rnorm(10)))
tmp <- list()
for (i in 1:500000) {
  tmp[[i]] <- wide_data
}

tmp <- data.table::rbindlist(tmp) |>
  as_tibble()

ex1_l <- bench::mark(
  old = old_data_to_long(tmp),
  new = data_to_long(tmp),
  tidyr = pivot_longer(tmp, cols = everything()),
  iterations = 10
)


ex2_l <- bench::mark(
  old = relig_income %>%
    old_data_to_long(-"religion", names_to = "income", values_to = "count"),
  new = relig_income %>%
    data_to_long(-"religion", names_to = "income", values_to = "count"),
  tidyr = relig_income %>%
    pivot_longer(!religion, names_to = "income", values_to = "count"),
  iterations = 100
)

ex3_l <- bench::mark(
  old = billboard %>%
    old_data_to_long(
      cols = starts_with("wk"),
      names_to = "week",
      values_to = "rank"
    ),
  new = billboard %>%
    data_to_long(
      cols = starts_with("wk"),
      names_to = "week",
      values_to = "rank"
    ),
  tidyr = billboard %>%
    pivot_longer(
      cols = starts_with("wk"),
      names_to = "week",
      values_to = "rank"
    ),
  iterations = 50
)

ex4_l <- bench::mark(
  old = who |>
    old_data_to_long(
      cols = 5:60,
      names_to = c("diagnosis", "gender", "age"),
      names_sep = "_",
      values_to = "count"
    ),
  new = who |>
    data_to_long(
      cols = 5:60,
      names_to = c("diagnosis", "gender", "age"),
      names_sep = "_",
      values_to = "count"
    ),
  tidyr = who |>
    pivot_longer(
      cols = 5:60,
      names_to = c("diagnosis", "gender", "age"),
      names_sep = "_",
      values_to = "count"
    ),
  iterations = 10
)

ex5_l <- bench::mark(
  old = who |>
    old_data_to_long(
      cols = 5:60,
      names_to = c("diagnosis", "gender", "age"),
      names_pattern = "new_?(.*)_(.)(.*)",
      values_to = "count"
    ),
  new = who |>
    data_to_long(
      cols = 5:60,
      names_to = c("diagnosis", "gender", "age"),
      names_pattern = "new_?(.*)_(.)(.*)",
      values_to = "count"
    ),
  tidyr = who |>
    pivot_longer(
      cols = 5:60,
      names_to = c("diagnosis", "gender", "age"),
      names_pattern = "new_?(.*)_(.)(.*)",
      values_to = "count"
    ),
  iterations = 10
)



### DATA_TO_WIDE ==========================================

ex1_w <- bench::mark(
  old = fish_encounters %>%
    old_data_to_wide(
      names_from = "station",
      values_from = "seen",
      values_fill = 0
    ),
  new = fish_encounters %>%
     data_to_wide(
      names_from = "station",
      values_from = "seen",
      values_fill = 0
    ),
  tidyr = fish_encounters %>%
    pivot_wider(
      names_from = "station",
      values_from = "seen",
      values_fill = 0
    ),
  iterations = 100
)




production <- expand_grid(
  product = letters,
  country = paste0(letters, "I"),
  year = 2000:2025
) %>%
  mutate(production = rnorm(nrow(.)))


ex2_w <- bench::mark(
  old = production %>%
    old_data_to_wide(
      names_from = c("product", "country"),
      values_from = "production"
    ),
  new = production %>%
     data_to_wide(
      names_from = c("product", "country"),
      values_from = "production"
    ),
  tidyr = production %>%
    pivot_wider(
      names_from = c(product, country),
      values_from = production
    ),
  iterations = 10
)


ex3_w <- bench::mark(
  old = production %>%
    old_data_to_wide(
      names_from = c("product", "country"),
      values_from = "production",
      names_glue = "prod_{product}_{country}"
    ),
  new = production %>%
     data_to_wide(
      names_from = c("product", "country"),
      values_from = "production",
      names_glue = "prod_{product}_{country}"
    ),
  tidyr = production %>%
    pivot_wider(
      names_from = c(product, country),
      values_from = production,
      names_glue = "prod_{product}_{country}"
    ),
  iterations = 10
)


tmp <- list()
for (i in 1:1000) {
  tmp[[i]] <- us_rent_income
}

tmp <- data.table::rbindlist(tmp) |>
  as_tibble()
tmp$GEOID <- rep(1:52000, each = 2)
tmp$NAME <- as.character(rep(1:52000, each = 2))

ex4_w <- bench::mark(
  old = tmp %>%
    old_data_to_wide(
      names_from = "variable",
      values_from = c("estimate", "moe")
    ),
  new = tmp %>%
     data_to_wide(
      names_from = "variable",
      values_from = c("estimate", "moe")
    ),
  tidyr = tmp %>%
    pivot_wider(
      names_from = "variable",
      values_from = c("estimate", "moe")
    ),
  iterations = 10
)

# SLOW (1M rows) ============

set.seed(123)
contacts <- tibble(
  id = rep(1:500000, each = 2),
  field = rep(c("a", "b"), 500000),
  value = sample(letters, 1000000, replace = TRUE)
)

ex5_w <- bench::mark(
  old = contacts %>%
    old_data_to_wide(names_from = "field", values_from = "value"),
  new = contacts %>%
     data_to_wide(names_from = "field", values_from = "value"),
  tidyr = contacts %>%
    tidyr::pivot_wider(names_from = field, values_from = value),
  iterations = 1
)


# SLOWER (10M rows) ============

set.seed(123)
contacts <- tibble(
  id = rep(1:5000000, each = 2),
  field = rep(c("a", "b"), 5000000),
  value = sample(letters, 10000000, replace = TRUE)
)

ex6_w <- bench::mark(
  old = contacts %>%
    old_data_to_wide(names_from = "field", values_from = "value"),
  new = contacts %>%
     data_to_wide(names_from = "field", values_from = "value"),
  tidyr = contacts %>%
    tidyr::pivot_wider(names_from = field, values_from = value),
  iterations = 1
)




reprex:::prex({
  ### DATA_TO_LONG ==========================================

  ex1_l
  ex2_l
  ex3_l
  ex4_l
  ex5_l

  ### DATA_TO_WIDE ==========================================

  ex1_w
  ex2_w
  ex3_w
  ex4_w
  ex5_w
  ex6_w
})