autotest

This package is edit from the testthat package. The main goal is makeing the error message human readable.

The usage of this packages is also the same with the testthat package.

• Document: autotest-docs.readthedocs.io

• Slides: Autotest with R

Design Logical

Actually there is no new design at all, just edit.

Error Message Hooks

Design

The autotest:::ErrorHandlerClass is a class for register error messages. It offers 6 APIs.

Four for error messages: registerPreMsg, registerPostMsg, addPreMsg, addPostMsg;

And two for turning on/off detecting variables/expressions: turnOnTraceback, turnOffTraceback.

The hidden variable autotest:::ErrorHandler is a instance of ErrorHandlerClass.

How it works?

Call Flow:

1. test_that -> test_code -> eval(...) -> as.expectation.error.

This occurs when running code trigger meet some errors. For exmaple, calling expect_equal(f(1), 2) without defining the function f.

The format.expectation_error functoin calls autotest:::ErrorHandler$msg, which deal with the error message hooks.

2. test_that -> test_code -> expect_XXX -> expect.

This occurs when meet failure expectattions. For exmaple, expect_equal(1, 2).

  if (expectation_broken(exp)) {
      exp$message <- ErrorHandler$msg(exp$message)  # add hooked error message
      stop(exp)
  }

BTW, Originally in the testthat package, handle_expectation does not stop when it meet expect error. Here are the code in function handle_expectation to stop it:

    if (expectation_broken(e)){
      return()
    }

Customized Error Message

This should be defined in each expect_XXX function.

Let us take the expect_equal function as an example. Here is the call flow: expect_equal -> compare.

The compare function is generic function, which contains compare.integer, compare.data.frame, compare.matrix and other methods. That is how expect_equal supports multiple data types and show different kinds of error messages.

If you want to support a new object, let is say, a ggplot object. Implement a compare.ggplot method, set up the test logical and customize error messages. For example, check the compare.data.frame method you will see that it tests the class, dimension and columns step by step.