:- use_module(to_be_tested). % define helper predicates here :- use_module(library(tap)). % define test predicates here 'two plus two is four' :- 4 is 2+2. 'zero not equal to one'(fail) :- 0 =:= 1. 6 is 3*2.
Run tests with standard TAP tools like prove:
$ prove -v -e 'swipl -q -t main -s' test/examples.pl TAP version 13 1..4 ok 1 - simplest possible test case ok 2 - simplest failing test case ok 3 - generates a diagnostic message # I'm a diagnostic message ok 4 - long-running test case # time=1.0ms # tests 4 # pass 4
The Test Anything Protocol is a text-based interface between test scripts and a test harness. A wide range of tools exist for running, rendering and analyzing test results. By writing your Prolog tests with TAP, you get access to all this testing infrastructure. For example, interactive HTML output.
TAP tests traditionally reside in a t/ directory in your project's root. Each file beneath t/ encapsulates a collection of tests related to a specific topic. During development, one can run all test files or just an interesting subset. In its most basic form, a test file is a script which generates TAP output. library(tap) helps you write these scripts.
To write a test file with library(tap), load all code that you'll need for testing. Define any helper predicates. Then load library(tap). All predicates defined after loading library(tap) are considered test cases. The predicate's name is the test name. By default, a predicate must succeed without leaving any choicepoints for the test to pass. See Arguments section below to change that behavior.
For small tests (see
6 is 3*2 above), the name can be omitted. The test body is then used as the test name.
library(tap) does not yet support the entire TAP specification and is missing many features found in PlUnit. Both are temporary shortcomings. I expect the library to fill these gaps eventually.
A test predicate can optionally include arguments to change TAP's expectations about the test. Arguments look like this:
'test with arguments'(Arg1, Arg2, ...) :- ...
Acceptable arguments are:
error(E)- same as
throws(E). Supported for symmetry with PlUnit.
fail- test is expected to fail
fixme(Reason)- same as
todo(Reason). Supported for symmetry with PlUnit.
todo(Reason)- test is known to fail but report it in TAP output as "TODO Reason". TAP tools treat these tests differently.
todo- same as
throws(E)- throws exception
It's common for each test case in a test file to follow a similar pattern. For example, we might have tests for the length/2 predicate:
:- use_module(library(tap)). 'length([a,b,c],3)' :- length([a,b,c], N), N = 3. 'length([a,b],2)' :- length([a,b], N), N = 2. ...
Because of all the similarity, that's tedious to write and tedious to read. We can factor out the redundancy by creating a macro:
% ... macro definition goes here ... :- use_module(library(tap)). [a,b,c] -> 3. [a,b] -> 2.
That's much better. A regular term_expansion/2 macro that calls tap:register_test/1 does the job:
term_expansion(List -> Length, (Head :- Test)) :- format(atom(Head), 'length(~w, ~w)', [List, Length]), Test = ( length(List, Len), Len = Length ), tap:register_test(Head).
Without registering, our nicely constructed test case won't run. Macros are especially convenient when testing multiple modes of a single predicate. You can decribe the relationship once and have the macro write a separate test case for each mode.
Using SWI-Prolog 7.1 or later:
Source code available and pull requests accepted at https://github.com/fnogatz/tap
This module uses semantic versioning.
Projects using library(tap)
Here is a non-exhaustive list of projects that use library(tap). Most of them use the macro approach, so they can be a good inspiration on how to define tests in Prolog:
Please open an issue if you want to have your project listed here.