cran-code.Rmd

---
title: "Untitled"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

```{r}
library(tidyverse)
theme_set(theme_light())

tuesdata <- tidytuesdayR::tt_load("2019-11-12")

cran_code <- tuesdata$loc_cran_packages
```

```{r}
View(cran_code)
```

What the most common programming languages in CRAN packages?

```{r}
cran_code %>%
  count(language, sort = TRUE)
```

```{r}
by_language <- cran_code %>%
  group_by(language) %>%
  summarize(packages = n(),
            code = sum(code),
            comments = sum(comment),
            files = sum(file),
            lines_per_package = code / packages,
            files_per_package = files / packages,
            comment_code_ratio = comments / code) %>%
  arrange(desc(packages))
```

```{r}
by_language %>%
  head(20) %>%
  mutate(language = fct_reorder(language, packages)) %>%
  ggplot(aes(language, packages)) +
  geom_col() +
  coord_flip() +
  labs(x = "",
       y = "# of packages that have code from this language")

library(tidytext)

by_language %>%
  gather(metric, value, packages, code, files) %>%
  group_by(metric) %>%
  top_n(8, value) %>%
  ungroup() %>%
  mutate(language = reorder_within(language, value, metric),
         metric = str_to_title(metric)) %>%
  ggplot(aes(language, value)) +
  geom_col() +
  coord_flip() +
  scale_x_reordered() +
  scale_y_continuous(labels = scales::comma) +
  facet_wrap(~ metric, scales = "free", ncol = 1) +
  labs(x = "Value (# of lines of code, files, or packages)")
```

How much are languages commented?

```{r}
by_language %>%
  filter(packages >= 20) %>%
  ggplot(aes(packages, comment_code_ratio)) +
  geom_point() +
  geom_text(aes(label = language), check_overlap = TRUE, vjust = 1, hjust = 1) +
  scale_x_log10() +
  expand_limits(x = 10) +
  labs(x = "# of packages language is used in",
       y = "Comment/Code ratio")
```

```{r}
by_language %>%
  filter(packages >= 20) %>%
  ggplot(aes(packages, lines_per_package)) +
  geom_point() +
  geom_text(aes(label = language), check_overlap = TRUE, vjust = 1, hjust = 1) +
  scale_x_log10() +
  expand_limits(x = 10) +
  labs(x = "# of packages language is used in",
       y = "Lines per package")
```

How much R code is there in each package?

```{r}
cran_code %>%
  filter(language == "R") %>%
  ggplot(aes(code)) +
  geom_histogram() +
  scale_x_log10(labels = scales::comma)
```

```{r}
cran_code %>%
  filter(language == "R") %>%
  arrange(desc(code))
```

Let's look just at the tidyverse packages

```{r}
packages <- tidyverse_packages() %>%
  str_extract("[a-z\\d]+")

cran_code %>%
  filter(pkg_name %in% packages) %>%
  mutate(pkg_name = fct_reorder(pkg_name, code, sum),
         language = fct_lump(language, 6),
         language = fct_reorder(language, code, sum)) %>%
  ggplot(aes(pkg_name, code, fill = language)) +
  geom_col() +
  guides(fill = guide_legend(reverse = TRUE)) +
  coord_flip() +
  labs(title = "How much code does each tidyverse package have?",
       x = "",
       y = "# of lines of code",
       fill = "Language")
```

```{r}
cran_code %>%
  filter(pkg_name %in% packages) %>%
  filter(language == "R") %>%
  mutate(comment_code_ratio = comment / code) %>%
  arrange(desc(comment_code_ratio)) %>%
  View()
```

```{r}
cran_code %>%
  filter(language == "R", code >= 100) %>%
  mutate(tidyverse = ifelse(pkg_name %in% packages, "Tidyverse", "Other")) %>%
  ggplot(aes(code / comment)) +
  geom_histogram() +
  scale_x_log10(labels = scales::number_format(accuracy = .1)) +
  facet_wrap(~ tidyverse, ncol = 1, scales = "free_y") +
  labs(x = "Code to comment ratio")
```

```{r}
cran_code %>%
  filter(code >= 100, language == "R", comment > 0) %>%
  mutate(code_comment_ratio = code / comment) %>%
  arrange(desc(code_comment_ratio))
```

```{r}
pkgs <- available.packages()

head(pkgs)

# Take package downloads from yesterday
package_downloads <- read_csv("http://cran-logs.rstudio.com/2019/2019-12-19.csv.gz")

downloads_by_package <- package_downloads %>%
  distinct(package, ip_id) %>%
  count(pkg_name = package, sort = TRUE, name = "downloads")
```

```{r}
cran_code %>%
  filter(language == "R") %>%
  inner_join(downloads_by_package, by = "pkg_name") %>%
  arrange(desc(downloads)) %>%
  filter(downloads >= 10) %>%
  ggplot(aes(downloads, code)) +
  geom_point() +
  geom_smooth(method = "lm") +
  scale_x_log10() +
  scale_y_log10()

cran_code %>%
  filter(language == "R") %>%
  inner_join(downloads_by_package, by = "pkg_name") %>%
  arrange(desc(downloads)) %>%
  filter(downloads >= 10, code >= 10) %>%
  ggplot(aes(downloads, code / comment)) +
  geom_point() +
  geom_smooth(method = "lm") +
  scale_x_log10() +
  scale_y_log10()
```

```{r}
package_metadata <- available.packages() %>%
  as_tibble() %>%
  janitor::clean_names() %>%
  select(-version, -file)
  
cran_code %>%
  filter(language == "R") %>%
  inner_join(package_metadata, by = c(pkg_name = "package")) %>%
  mutate(license = fct_lump(license, 10),
         license = fct_reorder(license, code)) %>%
  ggplot(aes(license, code)) +
  geom_boxplot() +
  coord_flip() +
  scale_y_log10()

n_import <- package_metadata %>%
  select(package, imports) %>%
  separate_rows(imports, sep = ",") %>%
  extract(imports, "import", "([A-Za-z\\d\\.]+)") %>%
  filter(!is.na(import)) %>%
  count(import, sort = TRUE, name = "n_reverse_import")
```

```{r}
cran_code %>%
  filter(language == "R") %>%
  left_join(n_import, by = c(pkg_name = "import")) %>%
  replace_na(list(n_reverse_import = 0)) %>%
  filter(n_reverse_import >= 10) %>%
  ggplot(aes(n_reverse_import, code)) +
  geom_point() +
  geom_text(aes(label = pkg_name), vjust = 1, hjust = 1, check_overlap = TRUE) +
  scale_x_log10() +
  scale_y_log10() +
  labs(x = "Number of packages that IMPORT this",
       y = "Lines of R code in package")
```