Create a Table Row Type and expose as a Vector on In-Memory Table with .rows property

[radeusgd]

Pull Request Description

Implements https://www.pivotaltracker.com/story/show/182307026

Important Notes

Checklist

Please include the following checklist in your PR:

• The documentation has been updated if necessary.
• All code conforms to the
  Scala,
  Java,
  and
  Rust
  style guides.
• All code has been tested:
  • Unit tests have been written where possible.
  • If GUI codebase was changed: Enso GUI was tested when built using BOTH
    ./run ide build and ./run ide watch.

[JaroslavTulach]

The implementation would have to be made ready to run on fast path. Anyway, I am not very excited to see this duck typing idea overall...

[JaroslavTulach]

I believe the ArrayProxy implementation is correct. I suggest to expand the VectorBenchmarks class with ArrayProxy case and compare it compiles well. Use ProxyList benchmark as an inspiration.

[jdunkerley]

libs part looks good - happy for this to go in once @JaroslavTulach and @hubertp happy with the built-in part.

[radeusgd]

I added the benchmarks as requested, for comparison I added the AbstractList benchmark @JaroslavTulach created too.

The results on my machine are following:

Benchmark	Time [μs/op]
averageAbstractList	119.85
averageOverArrayProxy	46.87
averageOverArray	18.24
averageOverPolyglotArray	18.05
averageOverPolyglotVector	16.74
averageOverSlice	17.25

So the proxy is roughly 2.5x slower than direct access (probably due to the additional layer of indirection), but still almost 2.6x faster than the polyglot AbstractList.

[JaroslavTulach]

averageOverArrayProxy 46.87
averageOverArray 18.24
So the proxy is roughly 2.5x slower than direct access

That is very slow. It should be on par with direct access. I'd try the benchmark without the bounds check. E.g. after removing:

    if (index >= length || index < 0) {
      throw InvalidArrayIndexException.create(index);
    }

[hubertp]

nit regarding range check

[hubertp]

I don't think that's correct or even necessary? If it is outside of the range but you tried to access it you should get UnsupportedMessageException? Right @JaroslavTulach ?

[radeusgd]

It is necessary to get the semantics that one would normally expect.

If I remove it, the following test:

        Test.specify "should correctly delegate to the callback" <|
            arr = Array_Proxy.new 3 (ix -> ix + 10)
            arr.length . should_equal 3
            arr.at 0 . should_equal 10
            arr.at 1 . should_equal 11
            arr.at 2 . should_equal 12
            arr.at 3 . should_fail_with Index_Out_Of_Bounds_Error_Data

will fail like so:

Array_Proxy:  [168ms]
    - [FAILED] should correctly delegate to the callback [103ms]
        Reason: Expected an error Constructor<Index_Out_Of_Bounds_Error_Data> but no error occurred, instead got: 13 (at /home/radeusgd/NBO/enso/test/Tests/src/Data/Array_Proxy_Spec.enso:20:13-70).