You might have read my blog post analyzing the social weather of rOpenSci onboarding, based on a text analysis of GitHub issues. I extracted text out of Markdown-formatted threads with regular expressions. I basically hammered away at the issues using tools I was familiar with until it worked! Now I know there’s a much better and cleaner way, that I’ll present in this note. Read on if you want to extract insights about text, code, links, etc. from R Markdown reports, Hugo website sources, GitHub issues… without writing messy and smelly code!

Introduction to Markdown rendering and parsing

This note will appear to you, dear reader, as an html page, either here on ropensci.org or on R-Bloggers, but I’m writing it as an R Markdown document, using Markdown syntax. I’ll knit it to Markdown and then Hugo’s Markdown processor, Blackfriday, will transform it to html. Elements such as # blabla thus get transformed to <h1 id="blabla">blabla</h1>. Awesome!

The rendering of Markdown to html or XML can also be used as a way to parse it, which is what the spelling package does in order to identify text segments of R Markdown files, before spell checking them only, not code. I had an aha moment when seeing this spelling strategy: why did I ever use regex to parse Markdown for text analysis?! Transforming it to XML first, and then using XPath, would be much cleaner!

As a side-note, realizing how to simplify my old code made me think of Jenny Bryan’s inspiring useR! keynote talk about code smells. I asked her whether code full of regular expressions instead of dedicated parsing tools was a code smell, sadly it doesn’t have a specific name, but she confirmed my feeling that not using dedicated purpose-built tools might mean you’ll end up “re-inventing all of that logic yourself, in hacky way.”. If you have code falling under the definition below, maybe try to re-factor and if needed get help.

It’s that feeling when you want to do something that sounds simple but instead your code is like 10 stack overflow snippets slapped together that you could never explain to another human what they do ???? pic.twitter.com/IF53AX6QvC

— Dr. Alison Hill (@apreshill) 31 d’agost de 2018

From Markdown to XML

In this note I’ll use my local fork of rOpenSci’s website source, and use all the Markdown sources of blog posts as example data. The chunk below is therefore not portable, sorry about that.

roblog <- "C:\\Users\\Maelle\\Documents\\ropensci\\roweb2\\content\\blog"

all_posts <- fs::dir_ls(roblog, regexp = "*.md")
all_posts <- all_posts[all_posts != "_index.md"]

My fork master branch isn’t entirely synced. It has 202 posts.

The code below uses the commonmark package to render Markdown to XML. Commonmark is a standardized specification for Markdown syntax by John McFarlane. The commonmark R package by Jeroen Ooms wraps the official cmark library and is used by e.g. GitHub to render issues and readmes. Note that my function still has a hacky element, it uses a blogdown unexported function to strip the YAML header of posts! If you know a better way feel free to answer my question over at RStudio community discussion forum.

library("magrittr")
get_one_xml <- function(md){
  md %>%
    readLines(encoding = "UTF-8") %>%
    blogdown:::split_yaml_body() %>%
    .$body %>%
    commonmark::markdown_xml(extensions = TRUE) %>%
    xml2::read_xml()
}

See what it gives me for one post.

get_one_xml(all_posts[42])

## {xml_document}
## <document xmlns="http://commonmark.org/xml/1.0">
##  [1] <paragraph>\n  <text>We just released a new version of </text>\n  < ...
##  [2] <heading level="2">\n  <text>First, install and load taxize</text>\ ...
##  [3] <code_block info="r">install.packages("rgbif")\n</code_block>
##  [4] <code_block info="r">library(taxize)\n</code_block>
##  [5] <heading level="2">\n  <text>New things</text>\n</heading>
##  [6] <heading level="3">\n  <text>New functions: class2tree</text>\n</he ...
##  [7] <paragraph>\n  <text>Sometimes you just want to have a visual of th ...
##  [8] <paragraph>\n  <text>Define a species list</text>\n</paragraph>
##  [9] <code_block info="r">spnames <- c("Latania lontaroides", "Randia ...
## [10] <paragraph>\n  <text>Then collect taxonomic hierarchies for each ta ...
## [11] <code_block info="r">out <- classification(spnames, db = "ncbi", ...
## [12] <paragraph>\n  <text>Use </text>\n  <code>class2tree</code>\n  <tex ...
## [13] <code_block info="r">tr <- class2tree(out)\nplot(tr, no.margin = ...
## [14] <paragraph>\n  <image destination="/assets/blog-images/2014-02-19-t ...
## [15] <heading level="3">\n  <text>New functions: get_gbfid</text>\n</hea ...
## [16] <paragraph>\n  <text>The Global Biodiversity Information Facility ( ...
## [17] <paragraph>\n  <text>We added a similar function to our </text>\n   ...
## [18] <code_block info="r">get_gbifid(sciname = "Poa annua", verbose = FA ...
## [19] <code_block>##         1\n## "2704179"\n## attr(,"class")\n## [1] " ...
## [20] <code_block info="r">get_gbifid(sciname = "Pinus contorta", verbose ...
## ...

Headings, code blocks… all properly delimited and one XPath query away from us! Let me convert all posts before diving into parsing examples.

all_posts %>%
  purrr::map(get_one_xml) -> blog_xml

Parsing the XML

URLs parsing

Let’s say I want to find out which domains are the most often linked from rOpenSci’s blog. No need for any regular expression thanks to commonmark, xml2 and urltools!

get_urls <- function(post_xml){
  post_xml %>%
    xml2::xml_find_all(xpath = './/d1:link', xml2::xml_ns(post_xml)) %>%
    xml2::xml_attr("destination") %>%
    urltools::url_parse()
}

# URLs
blog_xml %>%
  purrr::map_df(get_urls) %>%
  dplyr::count(domain, sort = TRUE) %>%
  head(n = 10) %>%
  knitr::kable()

More Twitter than CRAN! We probably could do with less own-domain use since / would get us here too.

R code parsing

Remember that cool post by Matt Dancho analyzing David Robinson’s code? In theory you could clone any of your favorite blogs (David Robinson’s blog, Julia Silge’s blog, etc.) to analyze them, no need to even webscrape first! Note that you can git clone from R using the git2r package.

get_functions <- function(post_xml){
  post_xml %>%
    # select all code chunks
    xml2::xml_find_all(xpath = './/d1:code_block', xml2::xml_ns(.)) %>%
    # select chunks with language info
    .[xml2::xml_has_attr(., "info")] %>%
    # select R chunks
    .[xml2::xml_attr(., "info") == "r"] %>%
    # get the content of these chunks
    xml2::xml_text() %>%
    glue::glue_collapse(sep = "\n") -> code_text
  
  # Base R code parsing tools
  parsed_code <- try(parse(text = code_text,
        keep.source = TRUE) %>%
    utils::getParseData(),
    silent = TRUE)
  
  if(is(parsed_code, "try-error")){
    # this happens because of output sometimes
    # stored in R chunks when not using R Markdown
    return(NULL)
  }
  
  if(is.null(parsed_code)){
    return(NULL)
  }
    
  dplyr::filter(parsed_code,
                grepl("FUNCTION", token))
    
}

blog_xml %>%
  purrr::map_df(get_functions) %>%
  dplyr::count(text, sort = TRUE) %>%
  head(n = 10) %>%
  knitr::kable()

Function definititions (function), basic stuff (c, library) and tidyverse functions seem to be the most popular on the blog!

Text parsing

After complementing our commonmark-xml2 combo with urltools and with R base code parsing facilities… let’s pair it with tidytext! What are the words most commonly use d n rOpenSci’s blog posts?

get_text <- function(post_xml){
  xml2::xml_find_all(post_xml,
                     xpath = './/d1:text', xml2::xml_ns(post_xml)) %>%
    xml2::xml_text(trim = TRUE) %>%
    glue::glue_collapse(sep = " ") %>%
    as.character() -> text
  
  tibble::tibble(text = text)
}

blog_xml %>%
  purrr::map_df(get_text) %>%
  tidytext::unnest_tokens(word, text, token = "words") %>%
  dplyr::filter(!word %in% tidytext::stop_words$word) %>%
  dplyr::count(word, sort = TRUE) %>%
  head(n = 10) %>%
  knitr::kable()

This beats my old code! There’s really something to be said for purpose-built tools.

Conclusion

I hope this note will inspire you to use commonmark and xml2 when analyzing Markdown files. As mentioned earlier, Hugo or Jekyll website sources are Markdown files and GitHub issue threads are too so it should open up quite a lot of data! If you’re new to XPath, I’d recommend reading this introduction. The results of XML-parsing are also better parsed without (your writing) regular expressions: I have shown urltools for URL parsing, that base R has code parsing tools (parse, getParsedData), and I’ve used tidytext.

Note that if you’re into blog analysis, don’t forget you can also get information out of the YAML header using… the yaml package, not regular expressions!

As a bonus, maybe seeing this wrangling inspired you to modify Markdown files programmatically? E.g. what if I wanted to automatically replace all level 1 headers with level 2 headers? We’re working on that, stay tuned and if you want follow this GitHub thread!

Thanks to Jeroen Ooms, Jenny Bryan and Jim Hester for their answering my XML parsing (meta) questions.