ETL: input elements define function parameter count

[bencoman]

This supersedes stale PR #1485. (It was easier to redo off HEAD than resolve conflict there)

Background - In practical terms for test generation, the the number of "input" elements is the number of parameters of the function-under-test. Considering the following as an example which specifies a function with three parameters as input...
all-your-base Canonical-data

   {
      "description": "single bit one to decimal",
      "property": "rebase",
      "input": {
          "inputBase": 2,
          "digits": [1],
          "outputBase": 10
       },

This is reflected by several language implementations:

all-your-base CSharp

Assert.Equal(expected, AllYourBase.Rebase(inputBase, digits, outputBase));    

all-your-base Ruby

converted = BaseConverter.convert(input_base, digits, output_base)

all-your-base ObjectiveC

NSArray<NSNumber *> *result = [ AllYourBase outputDigitsForInputBase:2 inputDigits:@[@1] outputBase:10 ];

---

By that pattern, the following...
etl canonical-data

        {
          "description": "multiple scores with multiple letters",
          "property": "transform",
          "input": {
            "1": ["A", "E"],
            "2": ["D", "G"]
          },

        {
          "description": "multiple scores with differing numbers of letters",
          "property": "transform",
          "input": {
             "1": ["A", "E", "I", "O", "U", "L", "N", "R", "S", "T"],
             "2": ["D", "G"],
             "3": ["B", "C", "M", "P"],
             "4": ["F", "H", "V", "W", "Y"],
             "5": ["K"],
             "8": ["J", "X"],
            "10": ["Q", "Z"]
          },

...would imply two separate functions, one with two parameters and another with seven parameters.
But actually "1","2","3","4","8","10" are part of the data, not part of the function definition.
The example language implementations correctly inferred the intent was to call a single-parameter function...

etl CSharp

    Assert.Equal(expected, Etl.Transform(input));

etl Ruby

    assert_equal expected, ETL.transform(old)

etl ObjectiveC

    NSDictionary<NSString *, NSNumber * > *results = [Etl transform: old];

It is better to be explicit that the transform function takes a single parameter of a legacy data map.
Thus the following canonical-data is proposed...

        {
          "description": "multiple scores with multiple letters",
          "property": "transform",
          "input": {
              "legacy": { 
                  "1": ["A", "E"],
                  "2": ["D", "G"] 
              }
          },

[bencoman]

Hi @ErikSchierboom, I used CSharp as an example which I understand you are involved with. I'm curious if if/how your generator inferred the input data was for a function taking a single dictionary rather than a function taking several parameters.

[ErikSchierboom]

Good question. There is nothing in the canonical data itself that encodes this, which is why there is special handling in this exercise's generator: https://github.com/exercism/csharp/blob/ed87d8c9818ea46a7db7856cc9a294ef6a2a10ae/generators/Exercises/Generators/Etl.cs#L26

Note: there are a few occasions that need special handling, and dictionary-related exercises are one of them.

[ErikSchierboom]

Oh and by the way, I agree that one input parameter should be defined for the combination of data. 👍

[ErikSchierboom]

LGTM! Just one thing: the version number needs to be bumped.

[bencoman]

AH yes, the version CI checks are working well. thx @ErikSchierboom
I guess this is "changing the type of one of the test data keys" ==> bump major.

[ErikSchierboom]

I guess this is "changing the type of one of the test data keys" ==> bump major.

Yep!

[ErikSchierboom]

Awesome work!

[rpottsoh]

LGTM too! Thanks for adding clarity.

[SleeplessByte]

🚀 🚀 🚀