Site icon R-bloggers

Using {pagedown} in Docker

R | datawookie

[This article was first published on R | datawookie, and kindly contributed to R-bloggers]. (You can report issue about the content on this page here)
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

I’m building an automated reporting system which generates PDF reports. My approach is to use R Markdown to write the report and render to PDF using the excellent {pagedown} package.

Ultimately the system needs to be packaged in Docker and deployed in the cloud.

Setup

To illustrate what I’m doing, we’ll use a simple dummy document, test.Rmd.

---
title: "Test Document"
output: html_document
---

This is a test document.

To convert this into PDF run:

pagedown::chrome_print("test.Rmd")

I got this all running in my local environment quite easily. However, I ran into a snag when trying to package the code with Docker.

The Chrome Problem

I created a Dockerfile based on rocker/r-ver, adding Chrome and {pagedown}, then copying across test.Rmd.

FROM rocker/r-ver:4.1.0

RUN apt-get update -qq && \
    apt-get install -y -qq --no-install-recommends \
        libz-dev \
        libpoppler-cpp-dev \
        pandoc \
        curl

RUN curl -L http://bit.ly/google-chrome-stable -o chrome.deb && \
    apt-get -y install ./chrome.deb && \
    rm chrome.deb

RUN install2.r --error --deps TRUE pagedown

COPY test.Rmd .

Running pagedown::chrome_print() from a container produces an error.

Error in is_remote_protocol_ok(debug_port, verbose = verbose) : 
  Cannot find headless Chrome after 20 attempts

Bummer!

What’s going on here? Chrome is clearly installed, so why is R failing to find it? Well, I think what’s happening is R is actually finding Chrome it but it’s failing to run it.

And the problem appears to relate to Chrome’s sandbox. This is a safety feature built into Chrome. However, in this instance we need to circumvent it to get things working.

Cartoon from the Google Chrome Comic.

Print to PDF using Docker

The {pagedown} documentation suggests two approaches to solving the problem.

Use --no-sandbox Argument

One solution is to send --no-sandbox to Chrome via the extra_args argument.

pagedown::chrome_print("test.Rmd", extra_args = c("--no-sandbox"))

This is perfectly reasonable. And it works! So, from a pragmatic perspective, it’s perfect.

However, I’m going to be making a bunch of calls to pagedown::chrome_print() and, in the interests of simplicity, I’d prefer not to have to provide the extra argument every time.

Specify Security Options

An alternative is to use docker run with --security-opt to specify some custom security options. Again, this works, but it’s just added complexity! Also, I prefer a solution that’s actually baked into the Docker image.

A Chrome Solution

A Chrome Shim

I created a BASH script shim, google-chrome, with the following contents:

#!/bin/bash

/usr/bin/google-chrome --no-sandbox $*

It basically executes Chrome, passing along all command line arguments plus --no-sandbox.

I made the script executable.

chmod u+x google-chrome

An Environment File

I also added the root folder, /, to the PATH environment variable in a file called Renviron.

PATH="/:${PATH}"

Tweaking the Dockerfile

The Dockerfile requires two small tweaks:

  • copy the google-chrome script across to /usr/local/bin/; and
  • copy Renviron as .Renviron.

The revised Dockerfile looks like this:

FROM rocker/r-ver:4.1.0

RUN apt-get update -qq && \
    apt-get install -y -qq --no-install-recommends \
        libz-dev \
        libpoppler-cpp-dev \
        pandoc \
        curl

RUN curl -L http://bit.ly/google-chrome-stable -o chrome.deb && \
    apt-get -y install ./chrome.deb && \
    rm chrome.deb

RUN install2.r --error --deps TRUE pagedown

COPY test.Rmd .

COPY Renviron /.Renviron
COPY google-chrome /usr/local/bin/

The actual Chrome executable is located at /usr/bin/google-chrome. But /usr/local/bin/ comes before /usr/bin/ in PATH, so when R looks for Chrome it finds the shim script first. This in turn adds in the --no-sandbox argument and my PDFs are then happily built by pagedown::chrome_print().

A Chromium Solution

How about using Chromium instead of Chrome? We need to make some changes to the Dockerfile to get Chromium installed.

FROM rocker/r-ver:4.1.0

# Install Chromium with apt not snap!
COPY bionic-updates.list /etc/apt/sources.list.d/
COPY chromium-deb-bionic-updates /etc/apt/preferences.d/

RUN apt-get update -qq && \
    apt-get install -y -qq --no-install-recommends \
        libz-dev \
        libpoppler-cpp-dev \
        pandoc \
        chromium-browser

RUN install2.r --error --deps TRUE pagedown

COPY test.Rmd .

COPY Renviron /.Renviron
COPY chromium-browser /usr/local/bin/

Adding in a shim script which supplies the --no-sandbox option to Chromium and we’re sorted! ?

Admittedly this is a relatively deep rabbit hole for such a simple (and probably inconsequential) issue. But it was fun and instructive.

Resources

If you want to try this out yourself, here are the files you’ll need:

Chrome —

Chromium —

To leave a comment for the author, please follow the link and comment on their blog: R | datawookie.
R-bloggers.com offers daily e-mail updates about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.