Generating tests for "all-your-base" exercise

[tommyschaefer]

Hi there!

I've been working on adding a test generator for the all-your-base exercise for exercism/xruby and I've run into a bit of a snag.

Everything works perfectly until I get to the last 13 (if I counted correctly) exercises that are left undefined. As far as I'm able to tell, and I certainly could be missing something, these tests aren't easy (possible?) to generate from individual tracks. Is that correct?

If so, I'm wondering if theres a way that we could add more information to the canonical-data so that different tracks could generate their tests based on the following (from the data documentation):

It's up to each track do decide:

1. What's the canonical representation of zero?
    - []?
    - [0]?

2. What representations of zero are allowed?
     - []?
     - [0]?
     - [0,0]?

3. Are leading zeroes allowed?

4. How should invalid input be handled?

Though, I'm really not to sure how that would work...

Have any other tracks successfully implemented a test generator for all-your-base? I looked through a couple of the other language repositories, but I wasn't able to find anything.

Thank you so much for your time!!

[Insti]

It's useful to link to a file you are discussing: https://github.com/exercism/x-common/blob/master/exercises/all-your-base/canonical-data.json

[Insti]

Aren't the last examples with expected: null error cases?
So you need to return/raise an appropriate error value.

Can you help us by asking a specific question about a specific case?

[dvberkel]

@stevejb71 has created a test generated and converted all-your-base with it for the xocaml track.

He probably faces similar questions, so maybe he can weigh in on the matter?

[tommyschaefer]

@Insti

Aren't the last examples with expected: null error cases?

Not necessarily, for example the leading zeros test:

{
  "description"  : "leading zeros",
  "input_base"   : 7,
  "input_digits" : [0, 6, 0],
  "output_base"  : 10,
  "expected"     : null
}

This is a case where tracks can either implement leading zeros as valid or invalid.

So in the Ruby implementation, we do allow leading zeros:

def test_leading_zeros
    skip
    digits = [0, 6, 0]
    input_base = 7
    output_base = 10
    expected = [4, 2]

    ...

But the Javascript implementation does not:

xit('leading zeros', function () {
  expect(function () {
    converter.convert([0, 6, 0], 7, 10);
  }).toThrow(new Error('Input has wrong format'));
});

@dvberkel Thanks so much! I'll check out that implementation 😄

[stevejb71]

Hi @tommyschaefer

I also allowed leading zeroes in OCaml (it felt more natural, even though it complicates the implementation).

The OCaml test generator works off a template file (one for each exercise), and substitutes in a None (OCaml equivalent of null, to indicate failure) for the null it sees in the json. Since this isn't what I wanted, I just hand-edited.

I see the test generator as a tool to help in translating exercises into OCaml, and to help updating tests when the json changes, but it will still need some human intervention for certain exercises.

[tommyschaefer]

Thank you so much for the help @Insti, @dvberkel, and @stevejb71!

I'm going to go ahead and close this issue since, with help from you all, I was able to get this working 😄

Thanks again everyone!!