In numbats/cranscrub: Tools for CRAN Data

library(tidyverse)
tocache <- TRUE
knitr::opts_chunk$set(echo = FALSE, 
                      cache = TRUE,
                      cache.path = "cache/",
                      fig.align = 'center', 
                      fig.pos = 'htbp', 
                      fig.width = 6,
                      message = FALSE,
                      warning = FALSE)

theme_set(
  theme(panel.background = element_rect(fill = NA),
        panel.grid = element_line(color = "lightgray"),
        axis.text = element_text(color = "black"),
        axis.line = element_line(color = "black", size = 0.7),
        axis.ticks.length = unit(1.4, "mm"),
        axis.ticks = element_line(color = "black", size = 0.7),
        axis.title = element_text(color = "black", face = "bold"),
        strip.background = element_rect(color = "black",
                                        fill = "black"),
        strip.text = element_text(color = "white"),
        plot.title.position = "plot",
        plot.title = element_text(color = "black", hjust = 0)))

library(tidyverse)
library(lubridate)
library(rvest)
library(glue)
library(dplyr)
library(purrr)
library(ggplot2)
library(feasts)
library(cranlogs)
library(stringr)
library(bookdown)
library(installr)
library(data.table)
library(scales)
library(cranlogs)

Introduction

With the growth of R community, many R packages have been developed as research products. @hornik2012did said that "R packages are the result of scholarly activity and as such constitute scholarly resources which must be clearly identifiable for the respective scientific communities". Majority of the R packages are developed and owned by individual author who has been contributing and sharing their knowledge with the public. It is important to recognise the contribution that these R package developers made to the scientific and academic communities. One of the most important metric for scholarly and scientific research publications is download statistics [@GreeneJosephW2016Wrdi]. Similar to publications, the download statistics is an important part of the metric. @rhub suggests that download counts are a popular way that indicates a package's importance and quality.

To use the R package download statistics as a metric, it must be accurate to be useful and reliable for any purposes such as grant application. As the number of R package downloads is calculated according to the CRANlog entries, the challenge here is to identify whether it is an actual user behind each entry.

dd_start <- "2012-10-01"
dd_end <- Sys.Date() - 1

is_weekend <- function(date) {
  weekdays(date) %in% c("Saturday", "Sunday")
}

total_downloads <- cran_downloads(from = dd_start, to = dd_end) %>% 
  mutate(weekend = is_weekend(date)) %>% 
  filter(row_number() <= n()-1)

The daily total number of R pakcages downloads from October 2012 to July 2021. It is clear that R packages has become popular with the number of R packages downloaded everyday increasing rapidly. There are two unusual number of R package download spikes happened in 2014 and 2018.

total_downloads %>%
  ggplot()  + 
  geom_line(aes(date, count/1000))+
  geom_smooth(aes(date, count/1000),stat = "smooth") +
  ggtitle("Daily number of R pakcages downloads") +
  scale_x_continuous(name ="Year", breaks=as.Date(c("2012-01-01","2013-01-01","2014-01-01","2015-01-01","2016-01-01","2017-01-01","2018-01-01","2019-01-01","2020-01-01","2021-01-01")),labels = c("2012","2013","2014","2015","2016","2017","2018","2019","2020","2021")) +
  scale_y_continuous("Number of R package Downloads", breaks = c(0,2500,5000,7500,10000,15000,20000), labels = c("0","2.5m","5m","7.5m","10m","15m","20m"))+ 
  theme(plot.title = element_text(hjust = 0.5))

load(here::here("paper/Data/spilk_2014.RData"))


ID_2014 <- spilk_2014 %>%  
  group_by(country,ip_id) %>% 
  count()

pkg_ID_2014 <- spilk_2014 %>%  
  group_by(country,ip_id,package) %>% 
  count()


ido <- max(ID_2014$n)/sum(ID_2014$n)

ID_2014_country <- spilk_2014 %>% 
  group_by(country) %>% 
  count() 

ID_2014_country <- ID_2014_country%>% 
  mutate(pre = (n/sum(ID_2014_country$n))*100)

library(rworldmap)

mapDevice('x11')
#join to a coarse resolution map
spdf <- joinCountryData2Map(ID_2014_country, joinCode="ISO2", nameJoinColumn="country")

mapCountryData(spdf, nameColumnToPlot="pre", catMethod="fixedWidth")

Figure shows the daily total number of R packages downloads from October 2012 to November 2021. From the plot, it suggests that there has been an enormous growth of R package downloads overtime. Among the growth of the number of downloads, there are two spikes observed. Zooming in for a closer look at the two spikes, the first one happened on 17th of November 2014. On the 17th of November 2014, r format(round(ido*100, 2), nsmall = 2)\% of the R packages downloads are done by the IP address from Indonesia.