github-commit-pattern.R

library(tidyverse)
library(gh)
library(lubridate)
library(glue)

repo <- tibble(json = unclass(gh("/user/repos", .limit = Inf)))

repo <- repo %>% 
  hoist(json, 
    owner = c("owner", "login"),
    repo = "name",
    private = "private",
    updated = "updated_at",
    size = "size",
    .remove = FALSE
  ) %>% 
  mutate(updated = parse_datetime(updated)) %>% 
  filter(year(updated) >= 2019, size > 0)
repo

commits_json <- pmap(repo, function(owner, repo, ...) {
  gh("/repos/:owner/:repo/commits", 
    owner = owner, 
    repo = repo, 
    author = "hadley",
    since = "2019-01-01",
    until = "2020-01-01",
    .limit = Inf
  )
})
saveRDS(commits_json, "~/Desktop/commits.rds")

commits <- tibble(json = commits_json %>% 
  discard(is.character) %>% 
  map(unclass) %>% 
  unlist(recursive = FALSE))

commits <- commits %>% hoist(json,
  author = c("author", "login"),
  committer = c("committer", "login"),
  date = c("commit", "author", "date"),
  .remove = FALSE
) %>% mutate(
  date = parse_datetime(date),
  day = as.Date(date),
  wday = wday(date, label = TRUE),
  time = update(with_tz(date, "US/Central"), yday = 1, year = 2019)
)

commits %>% 
  count(day) %>% 
  ggplot(aes(day, n)) + 
  geom_point()

commits %>% 
  ggplot(aes(fct_shift(wday, 1), time)) + 
  ggbeeswarm::geom_quasirandom(bandwidth = 0.2) + 
  scale_y_datetime(date_labels = "%H:%M")

commits %>% 
  ggplot(aes(fct_shift(wday, 1), time)) + 
  geom_jitter() + 
  scale_y_datetime(date_labels = "%H:%M")

commits %>% 
  ggplot(aes(time, fct_rev(fct_shift(wday, 1)))) + 
  ggridges::geom_density_ridges2(bandwidth = 30 * 60)