Awesome

NOTE: I am putting my support behind the {crrri} package and you should, too.

decapitated

Headless ‘Chrome’ Orchestration

Description

The ‘Chrome’ browser https://www.google.com/chrome/ has a headless mode which can be instrumented programmatically. Tools are provided to perform headless ‘Chrome’ instrumentation on the command-line, including retrieving the javascript-executed web page, PDF output or screen shot of a URL.

IMPORTANT

You'll need to set an envrionment variable HEADLESS_CHROME to use this package.

If this value is not set, a location heuristic is used on package start which looks for the following depending on the operating system:

Windows(32bit): C:/Program Files/Google/Chrome/Application/chrome.exe
Windows(64bit): C:/Program Files (x86)/Google/Chrome/Application/chrome.exe
macOS: /Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome
Linux: /usr/bin/google-chrome

If a verification test fails, you will be notified.

It is HIGHLY recommended that you use decapitated::download_chromium() to use a standalone version of Chrome with this packge for your platform.

It's best to use ~/.Renviron to store this value.

Working around headless Chrome & OS security restrictions:

Security restrictions on various operating systems and OS configurations can cause headless Chrome execution to fail. As a result, headless Chrome operations should use a special directory for decapitated package operations. You can pass this in as work_dir. If work_dir is NULL a .rdecapdata directory will be created in your home directory and used for the data, crash dumps and utility directories for Chrome operations.

tempdir() does not always meet these requirements (after testing on various macOS 10.13 systems) as Chrome does some interesting attribute setting for some of its file operations.

If you pass in a work_dir, it must be one that does not violate OS security restrictions or headless Chrome will not function.

Helping it “always work”

The three core functions have a prime parameter. In testing (again, especially on macOS), I noticed that the first one or two requests to a URL often resulted in an empty <body> response. I don’t use Chrome as my primary browser anymore so I’m not sure if that has something to do with it, but requests after the first one or two do return content. The prime parameter lets you specify TRUE, FALSE or a numeric value that will issue the URL retrieval multiple times before returning a result (or generating a PDF or PNG). Until there is more granular control over the command-line execution of headless Chrome.

What’s in the tin?

The following functions are implemented:

CLI-based ops

downlaod_chromium: Download a standalone version of Chromium (recommended)
chrome_dump_pdf: "Print" to PDF
chrome_read_html: Read a URL via headless Chrome and return the raw or rendered '<body>' 'innerHTML' DOM elements
chrome_shot: Capture a screenshot
chrome_version: Get Chrome version
get_chrome_env: get an envrionment variable 'HEADLESS_CHROME'
set_chrome_env: set an envrionment variable 'HEADLESS_CHROME'

`gepetto`-based ops

Helpers to get gepetto installed:

install_gepetto: Install gepetto
start_gepetto: Start/stop gepetto
stop_gepetto: Start/stop gepetto

API interface functions:

gepetto: Create a connection to a Gepetto API server
gep_active: Get test whether the gepetto server is active
gep_debug: Get "debug-level" information of a running gepetto server
gep_render_har: Render a page in a javascript context and serialize to HAR
gep_render_html: Render a page in a javascript context and serialize to HTML
gep_render_magick: Render a page in a javascript context and take a screenshot
gep_render_pdf: Render a page in a javascript context and rendero to PDF

More information on gepetto is forthcoming but you can take a sneak peek here.

Installation

devtools::install_github("hrbrmstr/decapitated")

Usage

library(decapitated)

# current verison
packageVersion("decapitated")

## [1] '0.3.0'

chrome_version()

chrome_read_html("http://httpbin.org/")

## {xml_document}
## <html>
## [1] <head>\n<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">\n<meta http-equiv="content-type" valu ...
## [2] <body id="manpage">\n<a href="http://github.com/kennethreitz/httpbin"><img style="position: absolute; top: 0; rig ...

chrome_dump_pdf("http://httpbin.org/")

chrome_shot("http://httpbin.org/")

##   format width height colorspace filesize
## 1    PNG  1600   1200       sRGB   215680