examples/sitemap.R

In jeroenooms/curl: A Modern and Flexible Web Client for R

### R sitemap example, Jeroen Ooms, 2016
#
# This code demonstrates the new multi-request features in curl 2.0. It creates
# an index of all files on a web server with a given prefix by recursively following
# hyperlinks that appear in HTML pages.
#
# For each URL, we first perform a HTTP HEAD (via curlopt_nobody) to retrieve
# the content-type header of the URL. If the server returns 'text/html', then we
# perform a subsequent request which downloads the page to look for hyperlinks.
#
# The network is stored in an environment like this: env[url] = (vector of links)
#
# WARNING: Don't target small servers, you might accidentally take them down and
# get banned for DOS. Hits up to 300req/sec on my home wifi.

library(curl)
library(xml2)

stopifnot(packageVersion('xml2') >= 1.0)
stopifnot(packageVersion('curl') >= 2.0)

# Extracts hyperlinks from HTML page
get_links <- function(html, url){
  tryCatch({
    doc <- xml2::read_html(html)
    nodes <- xml2::xml_find_all(doc, "//a[@href]")
    links <- xml2::xml_attr(nodes, "href")
    links <- xml2:::url_absolute(links, url)
    links <- grep("^https?://", links, value = TRUE)
    links <- sub("#.*", "", links)
    links <- sub("/index.html$", "/", links)
    unique(sub("/$", "", links))
  }, error = function(e){character()})
}

crawl <- function(root, timeout = 300){
  total_visited = 0
  pages <- new.env()
  pool <- curl::new_pool(total_con = 50, host_con = 6, multiplex = TRUE)

  crawl_page <- function(url){
    pages[[url]] <- NA
    h <- curl::new_handle(failonerror = TRUE, nobody = TRUE)
    curl_fetch_multi(url, handle = h, pool = pool, done = function(res){
      total_visited <<- total_visited + 1
      if(!identical(url, res$url)){
        pages[[url]] = res$url
        url = res$url
      }
      headers <- curl::parse_headers(res$headers)
      ctype <- headers[grepl("^content-type", headers, ignore.case = TRUE)]
      cat(sprintf("[%d] Found '%s' @ %s\n", total_visited, ctype, url))
      if(isTRUE(grepl("text/html", ctype))){
        handle_setopt(h, nobody = FALSE, maxfilesize = 1e6)
        curl::curl_fetch_multi(url, handle = h, pool = pool, done = function(res){
          total_visited <<- total_visited + 1
          links <- get_links(res$content, res$url)
          cat(sprintf("[%d] Extracted %d hyperlinks from %s\n", total_visited, length(links), url))
          followlinks <- grep(paste0("^", root), links, value = TRUE)
          pages[[url]] <- followlinks
          lapply(followlinks, function(href){
            if(is.null(pages[[href]]))
              crawl_page(href)
          })
        }, fail = function(errmsg){
          cat(sprintf("Fail: %s (%s)\n", url, errmsg))
        })
      }
    }, fail = function(errmsg){
      cat(sprintf("Fail: %s (%s)\n", url, errmsg))
    })
  }
  crawl_page(root)
  curl::multi_run(pool = pool, timeout = timeout)
  as.list(pages)
}

# Create a sitemap
sitemap <- crawl(root = 'https://cloud.r-project.org/web/packages', timeout = Inf)

# Show me all PDF files!
grep("\\.pdf$", names(sitemap), value = TRUE)