GitHub - armcn/quickcheck: Property Based Testing in R (original) (raw)

quickcheck

CRAN status R-CMD-check Codecov test coverage metacran downloads

Overview

Property based testing in R, inspired byQuickCheck. This package builds on the property based testing framework provided byhedgehog and is designed to seamlessly integrate with testthat.

Installation

You can install the released version of quickcheck fromCRAN with:

install.packages("quickcheck")

And the development version from GitHub with:

install.packages("remotes")

remotes::install_github("armcn/quickcheck")

Usage

The following example uses quickcheck to test the properties of the base R + function.Hereis an introduction to the concept of property based testing, and an explanation of the mathematical properties of addition can be foundhere.

library(testthat) library(quickcheck)

test_that("0 is the additive identity of +", { for_all( a = numeric_(len = 1), property = function(a) expect_equal(a, a + 0) ) }) #> Test passed πŸŽ‰

test_that("+ is commutative", { for_all( a = numeric_(len = 1), b = numeric_(len = 1), property = function(a, b) expect_equal(a + b, b + a) ) }) #> Test passed 😸

test_that("+ is associative", { for_all( a = numeric_(len = 1), b = numeric_(len = 1), c = numeric_(len = 1), property = function(a, b, c) expect_equal(a + (b + c), (a + b) + c) ) }) #> Test passed πŸ˜€

Here we test the properties of thedistinctfunction from the dplyrpackage.

library(dplyr, warn.conflicts = FALSE)

test_that("distinct does nothing with a single row", { for_all( a = any_tibble(rows = 1L), property = function(a) { distinct(a) %>% expect_equal(a) } ) }) #> Test passed 🎊

test_that("distinct returns single row if rows are repeated", { for_all( a = any_tibble(rows = 1L), property = function(a) { bind_rows(a, a) %>% distinct() %>% expect_equal(a) } ) }) #> Test passed 🎊

test_that("distinct does nothing if rows are unique", { for_all( a = tibble_of(integer_positive(), rows = 1L, cols = 1L), b = tibble_of(integer_negative(), rows = 1L, cols = 1L), property = function(a, b) { unique_rows <- bind_rows(a, b) distinct(unique_rows) %>% expect_equal(unique_rows) } ) }) #> Test passed πŸ˜€

Quickcheck generators

Many generators are provided with quickcheck. Here are a few examples.

Atomic vectors

integer_(len = 10) %>% show_example() #> [1] -833 5111 -8831 -3495 -1899 1051 9964 2473 9557 -2465 character_alphanumeric(len = 10) %>% show_example() #> [1] "y5Ph" "8" "B8" "3vOcYf" "qr" "o"
#> [7] "5rW2nHdrA" "88" "umU" "vJpqr" posixct_(len = 10, any_na = TRUE) %>% show_example() #> [1] "1652-02-25 11:34:40 LMT" "1683-08-15 05:26:47 LMT" #> [3] "2339-08-19 19:19:07 PDT" "0244-05-09 12:26:30 LMT" #> [5] "0756-11-24 03:23:10 LMT" "0660-04-16 21:21:08 LMT" #> [7] "2993-05-14 04:45:47 PDT" NA
#> [9] "1301-04-09 00:40:00 LMT" NA

Lists

list_(a = constant(NULL), b = any_undefined()) %>% show_example() #> $a #> NULL #> #> $b #> [1] -Inf flat_list_of(logical_(), len = 3) %>% show_example() #> [[1]] #> [1] TRUE #> #> [[2]] #> [1] TRUE #> #> [[3]] #> [1] TRUE

Tibbles

tibble_(a = date_(), b = hms_(), rows = 5) %>% show_example() #> # A tibble: 5 x 2 #> a b
#>

Hedgehog generators

quickcheck is meant to work with hedgehog, not replace it.hedgehog generators can be used by wrapping them in from_hedgehog.

library(hedgehog)

is_even <- function(a) a %% 2 == 0

gen_powers_of_two <- gen.element(1:10) %>% gen.with(function(a) 2^a)

test_that("is_even returns TRUE for powers of two", { for_all( a = from_hedgehog(gen_powers_of_two), property = function(a) is_even(a) %>% expect_true() ) }) #> Test passed πŸ˜€

Any hedgehog generator can be used with quickcheck but they can’t be composed together to build another generator. For example this will work:

test_that("powers of two and integers are both numeric values", { for_all( a = from_hedgehog(gen_powers_of_two), b = integer_(), property = function(a, b) { c(a, b) %>% is.numeric() %>% expect_true() } ) }) #> Test passed πŸŽ‰

But this will cause an error:

test_that("composing hedgehog with quickcheck generators fails", { tibble_of(from_hedgehog(gen_powers_of_two)) %>% expect_error() }) #> Test passed πŸ₯‡

A quickcheck generator can also be converted to a hedgehog generator which can then be used with other hedgehog functions.

gen_powers_of_two <- integer_bounded(1L, 10L, len = 1L) %>% as_hedgehog() %>% gen.with(function(a) 2^a)

test_that("is_even returns TRUE for powers of two", { for_all( a = from_hedgehog(gen_powers_of_two), property = function(a) is_even(a) %>% expect_true() ) }) #> Test passed πŸ˜€

Fuzz tests

Fuzz testing is a special case of property based testing in which the only property being tested is that the code doesn’t fail with a range of inputs. Here is an example of how to do fuzz testing with quickcheck. Let’s say we want to test that the purrr::map function won’t fail with any vector as input.

test_that("map won't fail with any vector as input", { for_all( a = any_vector(), property = function(a) purrr::map(a, identity) %>% expect_silent() ) }) #> Test passed πŸŽ‰

Repeat tests

Repeat tests can be used to repeatedly test that a property holds true for many calls of a function. These are different from regular property based tests because they don’t require generators. The functionrepeat_test will call a function many times to ensure the expectation passes in all cases. This kind of test can be useful for testing functions with randomness.

test_that("runif generates random numbers between a min and max value", { repeat_test( property = function() { random_number <- runif(1, min = 0, max = 10) expect_true(random_number >= 0 && random_number <= 10) } ) }) #> Test passed πŸŽ‰