Amber unit test framework

[lens0021]

This repository contains both tests for rust and tests for amber. This discussion is for the latter.

Motivation 1. As an amber user, I want to test the amber scripts I write.
Motivation 2. As (a kind of) a maintainer of Amber, I want to

• improve the readability of test cases. Currently, the Output comments show the all output on the top altogether, the test result does now tell why it should be, etc.
• use the common way to write unit tests, for example, assert* functions.

And the following is some questions:

• This will be provided as an amber sub-command, or an external library?
  • If it implemented in the amber itself, the assertion functions will be built-ins or stdlibs?
• It will be allowed to test all functions, or only pub functions?
• How to implement code coverage measurement?
• How assertion functions fail. Like rust panic!()? Or setting an environment value?
• How to suggest the naming conversions of the test files and functions.
• Whether a compiler flag similar to #[cfg(test)] will be introduced.
• Whether to introduce the concepts of test suite, tests groups or something.
• How to make a test case skippable conditionally.
• After implementation, How to gradually introduce the framework to our current test codes.

[lens0021]

This is a form of the ideal unit test code what I am thinking. Modified from https://github.com/lens0021/test-chamber/blob/e4b7f7c511a12da5fc795b8f9c08696a60273f17/stdlib-examples/array_contains.ab

import * from "std/array"
import { assert_true, assert_false } from "std/test"

#[test]
pub failable fun test_array_contains() {
    const array = [1, 2, 3]
    assert_true(array_contains(array, 2), "array_contains should return true when given an item")?
    assert_false(array_contains(array, 0), "array_contains should return false when given a non existing item")?
}

assert_* functions fails when the test failed.

[Mte90]

The cool part about the Amber tests that is pure Amber code. I used it to check the bash output quality as we have in this way tons of code.
Also I can run the test as single script to speed up the development, in this way we are complicating this two usage of the tests.