In openpharma/GithubMetrics: Quickly Get Key Metrics on Github Repositories

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%",
  knitr.kable.NA = ''
)

GithubMetrics

The aim of this package is to provide a wrapper on gh to quickly get you key Github repo information you need.The code here is used within Roche to quickly let me pull answer simple questions like:

How many studies have more than 1 data scientist (and roughly what's the commit split)
What are the common languages being used (proxied through file type distribution within repos)
Pull commit metadata to enrich other study info held in other systems

Installation

You can install the released version of GithubMetrics from CRAN with:

install.packages("GithubMetrics")

Setup

library(GithubMetrics)
library(tidyverse)
library(glue)

organisation <- "openpharma"

Repos in an org

Pull all the repos present within an org (that I can see).

repos_raw <- gh_repos_get(
  org = organisation
  )

repos_clean <- gh_repos_clean(repos_raw)

glimpse(repos_clean)

Realistically, research code is likely to be on Github Enterprise, so the .api_url and .token parameters can be passed through to gh(). Commented code below shows how you can use an on-premise Github server.

# repos_raw <- gh_repos_get(
#   org = organisation,
#   .api_url = "https://github.roche.com/api/v3",
#   .token = Sys.getenv("GITHUB_PAT_ROCHE")
#   )

Commits

Get every commit for all the repos in this organisation.

repo_all_commits <- gh_commits_get(
  repos_clean %>% filter(size > 0) %>% pull(full_name), 
  days_back = 365*10
)

glimpse(repo_all_commits)

People

Pull all the people that have committed in r.

contributors <- repo_all_commits %>%
  group_by(author) %>%
  summarise(
    commits = n()
  ) %>%
  filter(!author %in% c(".gitconfig missing email","actions-user"))

contributors <- contributors %>%
  left_join(
    gh_user_get(contributors$author),
    by = c("author"="username")
  )

contributors %>%
  arrange(-commits) %>%
  mutate(
    last_active = Sys.Date() - last_active,
    contributor = glue('<img src="{avatar}" alt="" height="30"> {author}'),
    blog = case_when(
      blog == "" ~ "",
      TRUE ~ as.character(glue('<a href="{blog}">link</a>'))
      )
    ) %>%
  select(contributor,commits,name,last_active,company,location,blog) %>%
  knitr::kable(

  )

Files

Pull a specific file using gh_file_get().

desc_formatted <- gh_file_get(
  repo = "GithubMetrics",
  org = "OpenPharma",
  file = "DESCRIPTION"
) %>%
  # format the description
  desc::desc(text = .)

# Print it
desc_formatted$get(c("Package","Title","Version")) %>%
  tibble::enframe() %>%
  knitr::kable()

Get all of the files present in the last commit of all the repos using gh_repo_files_get().

repo_files <- gh_repo_files_get(
  repo_commits = repo_all_commits,
  only_last_commit = TRUE
)

glimpse(repo_files)

repo_files %>%
  group_by(repo) %>%
  summarise(
    Files = n(),
    `R files` = sum(lang %in% "R"),
    `Python files` = sum(lang %in% c("Python","Jupyter Notebook"))
  ) %>% knitr::kable(
    caption = "Types of files in the organisation"
  )

results <- gh_repo_search(
  code = "tidyverse",
  organisation = organisation
)

glimpse(results)

helper_gh_repo_search <- function(x, org = "openpharma"){

  ## Slow it down! as search has 30 calls a minute rate limit.
  ## If you prem the search rate limit is higher, so usually not needed
  if(interactive()){message("Wait 5 seconds")}
  Sys.sleep(5)
  ## End slow down


   results <- gh_repo_search(
      code = x,
      organisation = org
    ) 

  if(is.na(results)) {
    results <- return()
  }
  results %>% 
    mutate(Package = x, Organisation = org) %>%
    group_by(Organisation,Package) %>%
    summarise(
      Repos = n_distinct(full_name), .groups = "drop"
    )
}

packages <- c(
  "tidyverse","pkgdown","dplyr","data.table"
  )

package_use <- bind_rows(
  packages %>%
    map_df(
      helper_gh_repo_search, org = "PHCAnalytics"
    ),
  packages %>%
    map_df(
      helper_gh_repo_search, org = "openpharma"
    ),
  packages %>%
    map_df(
      helper_gh_repo_search, org = "AstraZeneca"
    ),
  packages %>%
    map_df(
      helper_gh_repo_search, org = "Roche"
    ),
  packages %>%
    map_df(
      helper_gh_repo_search, org = "Genentech"
    ),
  packages %>%
    map_df(
      helper_gh_repo_search, org = "Novartis"
    )
)


package_use %>%
  pivot_wider(names_from = "Package", values_from = "Repos") %>%
  mutate(Total = rowSums(.[,-1], na.rm = TRUE)) %>%
  arrange(-Total) %>%
  knitr::kable(
    caption = "Package use detected within repositaries in Pharma orgs"
  )