Plotting download counts of various R versions from R Studio CRAN mirror

rstudio-download-logs.R

library(tidyverse)
library(lubridate)
library(scales)

## generate names for log files
start <- as.Date('2013-01-01')
today <- as.Date('2022-04-10')
all_days <- seq(start, today, by = 'day')
year <- as.POSIXlt(all_days)$year + 1900

## place to store the log files
log_dir <- file.path(tempdir(), "cran_logs")
dir.create(log_dir)

## sometimes I get a failures in the download step, this allows us 
## to only re-try the missing log files
missing_days <- setdiff(as.character(all_days), strtrim(tools::file_path_sans_ext(list.files(log_dir), TRUE), 10))
if(length(missing_days)) {
    urls <- paste0('http://cran-logs.rstudio.com/', as.POSIXlt(missing_days)$year + 1900, '/', missing_days, '-r.csv.gz')
} else {
    urls <- paste0('http://cran-logs.rstudio.com/', year, '/', all_days, '-r.csv.gz')
}

## download the files - sometimes I get a failure, so repeat steps if needed.
download.file(urls, destfile = file.path(log_dir, basename(urls)), quiet = TRUE)

## read all files
logs <- readr::read_csv(file = list.files(log_dir, pattern = ".csv.gz$", full.names = TRUE), 
                        col_types = "Dtdccci",
                        progress = FALSE)
    
downloads <- logs %>%
    ## trucate version to whatever comes before the first period '.'
    mutate(version = gsub(version, pattern = "^([[:alnum:]]*).*", replacement = "\\1")) %>%
    group_by(month = floor_date(date, unit = "month"), version) %>% 
    summarise(downloads = n())
    
## calculate a 14 day rolling mean for proportion of unique IP addresses 
ips <- group_by(logs, date) %>% 
    summarise(prop_uniq = length(unique(ip_id)) / n()) %>%
    mutate(prop_uniq_s = RcppRoll::roll_meanr(prop_uniq, 14))

## create plot
ggplot() + 
    geom_bar(data = downloads, 
             aes(x = month, y = downloads, fill = version), 
             stat = "identity") +
    geom_line(data = ips,
              aes(y = prop_uniq_s * 1e6, x = date), 
              alpha = 0.5, lwd = 1.5) +
    scale_x_date(name = "Date") +
    scale_y_continuous(name = "Number of downloads", 
                       labels = comma,
                       sec.axis = sec_axis( trans=~./1e6, name="Proportion of unique IPs")) + 
    scale_fill_brewer(palette = "Set1") +
    theme_minimal()