Authenticated API Testing Using Travis CI

August 6, 2015

(This article was first published on » R, and kindly contributed to R-bloggers)

As I’ve become more serious about contributing in the open-source community, having quality tests for my packages has been something I’ve spent much more time on than when I was just writing quick-and-dirty code for my own purposes. My most used open-sourced package is RSiteCatalyst, which accesses the Adobe Analytics (authenticated) API, which poses a problem: how do you maintain a project on GitHub with a full test suite, while at the same time not hard-coding your credentials in plain sight for everyone to see?

The answer ends up being using encrypted environment variables within Travis CI.


In terms of a testing framework, Hadley Wickham provides a great testing framework in testthat; while I wouldn’t go as far as he does to say that the package makes testing fun, it certainly makes testing easy. Let’s take a look at some of the tests in RSiteCatalyst from the QueueOvertime function:From the gist above, you can see the tests are fairly simplistic; for a given number of permutations of arguments of the function, I test to see if a data frame was returned. This is because, for the most part, RSiteCatalyst is just a means of generating JSON calls, submitting them to the Adobe Analytics API, then parsing the results into an R data frame. Since there is very little additional logic in the package, I don’t spend a bunch of time testing what data is actually returned (i.e. what is returned depends on the Adobe Analytics API, not R).

What is interesting is line 6 of the gist; I reference Sys.getenv() twice in order to pass in my username and key for the Adobe Analytics API, which feels very “interactive R”, but the goal is automated testing. Filling in those two environment variables is where Travis CI comes in.

Travis CI Configuration

In order to have any automation using Travis CI, you need to create a .travis.yml configuration file. While you can read the Travis docs to create the .travis.yml file for R, you’re probably better off just using the use_travis function from devtools (also from Hadley, little surprise!) to create the file for you. In terms of creating encrypted keys to use with Travis, you’ll need to use the Travis CLI tool, which is distributed as a Ruby gem (i.e. package).  If you view the RSiteCatalyst .travis.yml file, you can see that I define two global “secure” variables, the value of which are the output from running a command similar to the following in the Travis CLI tool:
Note that if this seems insecure, every time you run the encrypt command with the same arguments, you get a different value; Travis CI is creating new public and private RSA keys each time.

Setting Up Authenticated Testing Locally

If you get as far as setting up encrypted Travis CI keys and tests using testthat, the final step is really for convenience. With the .travis.yml file, Travis CI sets the R environment variables on THEIR system; on your local machine, the environment variables aren’t set. Even if the environment variables were set, they would be set to the Travis CI hashed values, which is not what I want to pass to my authentication function in my R package.

To set the authentication variables locally, so that each time you hit ‘check’ to build and check against CRAN errors, you just need to modify the .Renviron file for R:
With that minor change, in addition to the .travis.yml file, you’ll have a seamless environment for developing and testing R packages.

Testing Is Like Flossing…

As easy as the testthat and devtools packages make testing, and as inexpensively as Travis CI is as a service (free for open source projects!), there’s really no excuse to provide packaged-up code and not include tests. Hopefully this blog post has demonstrated that it’s possible to include tests even when authentication is necessary without compromising your credentials.

So let’s all be sure to include tests, not just pay lip service to the idea that testing is useful. Code testing only works if you actually do it :)

To leave a comment for the author, please follow the link and comment on their blog: » R. offers daily e-mail updates about R news and tutorials on topics such as: Data science, Big Data, R jobs, visualization (ggplot2, Boxplots, maps, animation), programming (RStudio, Sweave, LaTeX, SQL, Eclipse, git, hadoop, Web Scraping) statistics (regression, PCA, time series, trading) and more...

If you got this far, why not subscribe for updates from the site? Choose your flavor: e-mail, twitter, RSS, or facebook...

Comments are closed.


Never miss an update!
Subscribe to R-bloggers to receive
e-mails with the latest R posts.
(You will not see this message again.)

Click here to close (This popup will not appear again)