triangle: actively ignore cases with float input

[marionebl]

No description provided.

[sshine]

I don't think that this is a good idea.

Here we perform structural equality on floats.

I think the exercise is complicated significantly if we are to introduce rounding-error testing.

I would also say that this is a bug in OCaml. See Why can't I compare reals in Standard ML?.

[marionebl]

I don't think that this is a good idea.

Here we perform structural equality on floats.

I think the exercise is complicated significantly if we are to introduce rounding-error testing.

I would also say that this is a bug in OCaml. See Why can't I compare reals in Standard ML?.

The test dataset contains floating point numbers though. Should we filter out those test cases then?

[sshine]

The test dataset contains floating point numbers though. Should we filter out those test cases then?

Yes. triangle's canonical data gives this comment:

      "comments": [
        " Your track may choose to skip this test    ",
        " and deal only with integers if appropriate "
      ],

When exercism/problem-specifications#1518 merges, we can push a change to triangle's canonical tests that changes these comments into an optional: "floating-point" and filter based on this in the test generator.

[marionebl]

Ok, so filtering cases was more involved than I would have guessed. Anyway - adapted the test-generator as requested.

[sshine]

Excellent work.

I went and pushed for the merge of the "optional": "..." feature in canonical-data.json to make this easier, but it seems you found a way with Yojson to filter out the non-integer test cases. Nice job.