Channel 9 Forums - Coffeehouse - Unconventional use of indexers

Coffeehouse - Unconventional use of indexers

dcuccia — Thu, 11 Sep 2008 06:53:49 GMT

I'm playing around with trying to make array math in C# as easy and concise as in Matlab, and (besides array operators) the biggest language "mismatch" is in indexers, ranges, etc. In Matlab, I can easily get a subarray of A, such as A[0:5:255]

As a user of a fluent interface DSL in C#, would you ever put up with unconventional use of indexers? Like:

A[Get.Range[0,5,255]]

...where Range is actually a static readonly field holding an instance of class Range

    public class Get
    {
        public static readonly Range Range = new Range();
        public static readonly Size Size = new Size();
    }

    public class Range
    {
        public Range this[int from, int to] { get { return new Range(from, 1, to); } }
        public Range this[int from, int every, int to] { get { return new Range(from, every, to); } }

        private Range(int start, int every, int stop) { Start = start; Every = every;  Stop = stop; }
        private Range(int start, int stop) : this(start, 1, stop) { }
        internal Range() { } // for indexer use

        public int Start { get; private set; }
        public int Stop { get; private set; }
        public int Every { get; private set; }
        public int Count() { return (Stop - Start) / Every + 1; }
    }

    public class Size
    {
        public Size this[int d0] { get { return new Size(new int[] { d0 }); } }
        public Size this[int d0, int d1] { get { return new Size(new int[] { d0, d1 }); } }
        public Size this[int d0, int d1, int d2] { get { return new Size(new int[] { d0, d1, d2 }); } }
        public Size this[int d0, int d1, int d2, int d3] { get { return new Size(new int[] { d0, d1, d2, d3 }); } }

        private Size(int[] dimensions) { Dimensions = dimensions.ToArray(); }
        internal Size() { } // for indexer use

        public int[] Dimensions { get; private set; }
    }

Obviously the following would be more C-sharpy:

A[new Range(0,5,255)]

But, isn't it OK to stray a bit as long as there's a consistent convention in the DSL?

Coffeehouse - Unconventional use of indexers

Ion Todirel — Thu, 11 Sep 2008 09:05:11 GMT

you may also want to make the constructor of the Range class internal, to force the user to use it via Get class (if you use first approach), I would suggest to use the second one, but the first one looks cooler...

Coffeehouse - Unconventional use of indexers

stevo_ — Thu, 11 Sep 2008 10:36:09 GMT

Ion Todirel said:
you may also want to make the constructor of the Range class internal, to force the user to use it via Get class (if you use first approach), I would suggest to use the second one, but the first one looks cooler...

I'd be honest and say I don't think it adds anything or makes it really different, to me it just serves as a barrier for people to realize whats actually going on here..

Coffeehouse - Unconventional use of indexers

dcuccia — Fri, 12 Sep 2008 20:13:26 GMT

stevo_ said:

Ion Todirel said:

*snip*

I'd be honest and say I don't think it adds anything or makes it really different, to me it just serves as a barrier for people to realize whats actually going on here..

Thanks for the thoughts Ion and stevo_.

Is the "Get" class a barrier too? If you had factory methods, which would you prefer (assuming "Get" is used for many purposes in the DSL)?

A[Get.Range(0,5,255)]

or a classic

 A[Range.GetRange(0,5,255)]

Coffeehouse - Unconventional use of indexers

JChung2006 — Fri, 12 Sep 2008 21:07:58 GMT

I would prefer A.Range(from, to) and A.Range(from, to, every) rather than the indexer, but that's just my opinion.