In gaborcsardi/pressr: Fake Web Apps for HTTP Testing

#| include: false
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

What is webfakes?

Webfakes is an R package that can spin up web servers on your machine to facilitate testing R code. R code that needs an HTTP connection is not trivial to test:

• Connectivity problems might prevent the tests from accessing the web server.
• The web server might need authentication, and it is not easy to convey login information to the test suite in a secure way.
• The web server might have rate limits, i.e, it limits the number of queries per hour or day, causing some spurious test failures.
• You might want to test in non-normal conditions, e.g. with low bandwidth, or when the client is rate limited. These conditions don't normally happen on the web server and they are hard to trigger.

With webfakes you can easily start a custom web app, that is running on the local machine.

• Webfakes does not need a network connection.
• Webfakes does not need authentication. Well, unless you want it to.
• Webfakes does not have rate limits.
• Webfakes can simulate low bandwidth, or a broken connection.

Webfakes vs mocking

Mocking is a general technique to mimic the behavior of a function or object that is needed in test case. In the case of HTTP requests, this typically means that both the request and its response are recorded when the tests run the first time, and saved to disk. Subsequent test runs intercept the HTTP requests, match them against the recorded requests and then replay the corresponding recorded response. See for example the vcr and httptest R packages.

The advantages of using your own webfakes server, over mocking:

• Simpler infrastructure. No separate recording and replaying phases, no recorded files. No request matching.
• You can use any web client you want. E.g. curl and base R's HTTP functions do not explicitly support mocking currently.
• No need to worry about sensitive information in recorded requests or responses.
• Often easier to use when testing non-normal conditions, e.g. errors that are hard to trigger, low bandwidth, or rate limits.
• Works if you stream data from a HTTP connection, instead of reading the whole response at once.
• You can reuse the same app for multiple tests, in multiple packages.
• Easier to use for tests that require multiple rounds of requests.
• Comes with a built-in https://httpbin.org compatible app, chances are, you don't even need to write your testing app, just start writing tests right away.
• Better test writing experience. This is subjective, and your mileage may vary.

Webfakes vs the real API

• No network needed. No more skip_if_offline().
• Much faster.
• No rate limits. But you can simulate one if you want to.
• You can write your custom app.
• Simulate low bandwidth or a broken connection.

Webfakes vs httpbin.org

• No network needed. No more skip_if_offline().
• Much faster.
• You can use the built-in webfakes::httpbin_app() app, so it is easy to switch from httpbin.org.
• You can write your custom app, httpbin.org might not have what you need.

Using webfakes::httpbin_app() with testthat

You can also create a web server for a test file, or even for a single test case. See vignette("how-to") for details how.

Writing apps

If the builtin httpbin_app() is not appropriate for your tests, you can write your own app. You can also extend the httpbin_app() app, if you don't want to start from scratch.

gaborcsardi/pressr documentation built on June 2, 2025, 2:10 a.m.