gistfile1.txt

#code to the video: https://youtu.be/hs7q64diU_c

library(tidyverse)

# prepare johnson df
data(JohnsonJohnson)

johnson <- as_tibble(matrix(JohnsonJohnson, ncol = 4, nrow= 21, byrow = TRUE)) %>%
  rename(quarter1 = V1,
         quarter2 = V2,
         quarter3 = V3,
         quarter4 = V4) %>%
  mutate(year = seq(1960, 1980))

# the tidy way
# capture the expressions using quote
# for benchmarking
# you can remote the call to quote() to just 
# get the data. but capturing the call allows
# me to benchmark it.
tidy_expr <- quote(johnson %>%
  pivot_longer(cols = starts_with("quarter"), names_to = "quarter", values_to = "values") %>%
  group_by(year) %>%
  summarise(mean(values)))
  

# c_across solution
c_across_expr <- quote(johnson %>%
  rowwise() %>%  
  mutate(year_mean = mean(c_across(starts_with("quarter")))))

# "base" solution

base_expr <- quote(johnson %>%
  mutate(year_mean = rowMeans(across(starts_with("quarter")))))

# Now the benchmark
microbenchmark::microbenchmark(
                  eval(tidy_expr),
                  eval(c_across_expr),
                  eval(base_expr),
                  times = 100
                )

# Another example, with the EUSILC dataset
# here you actually want to work on a per row basis
options(pillar.sigfig = 10)

data("eusilcP", package = "simPop")

silc <- as_tibble(eusilcP)

silc

silc %>%
  mutate(across(starts_with("py"), as.character)) %>%
  head

silc %>%
  rowwise() %>%  
  mutate(eq_sum = sum(c_across(starts_with("eq")))) %>%  
  select(starts_with("eq")) %>%  
  head

silc %>%
  mutate(eq_sum = rowSums(across(starts_with("eq")))) %>%  
  select(starts_with("eq")) %>%  
  head