Handling a multi-dimensional data structure in .Net 3.5 and above

Question

I want to build a 2 dimensional (non ragged at this point) object array.

I can easily build a 2 dimensional Array[,], and will do so if it is the best option available, but have tended to avoid arrays in favour of the advanced functionality of .NET's List and Dictionary structures.

I could also use a List<List<T>> to store a 2 dimensional array, but was wondering if there was any best-practice or implemented data structures in .NET 3.5 or above to handle Typed 2 - n dimensional structures with more flexible / comprehensive functionality than an array?

I am not interested in SSAS/OLAP style answers.

Answer 1

I'd look at Tuple implementations for .NET 3.5. Tuples will be native in .NET 4.0, so this will be forward compatible:

• http://stackoverflow.com/questions/152019/tuples-in-c
• Using anonymous types to simulate a tuple (used quite a bit by the MVC devs)

Answer 2

Arrays are one of the CLI fundamental data types. However, the "SzArray" (single dimensional, 0-based indexing) is the fastest of the arrays. It's the one you see declared int[]. Either way, if you have a fixed-size set of data, then an array (single- or multi-dimensional) will provide the best performance.

I wouldn't be at all surprised if internal optimizations made the two dimensional array faster than manually using row-major indexing for a single dimensional array. I also wouldn't be suprised if it was the other way around. Make sure to use a quality profiler if this matters heavily (either way should be fast for practical uses, I'm talking matrix math here).

Answer 3

• A two dimensional array would work really well, it would consume the least amount of ram, and have the fastest lookup times. It would generally work well with read-only data
• A list of lists, or dictionary of lists, or dictionary of arrays, or dictionary of dictionaries would work well too depending on how you need to access the data etc