A CSV parsing tool written in C#, primarily for my own edification but also with some utility. I intend to continue learning about .NET metaprogramming and improve this code, in looks, features, and performance.

Unlike many CSV parsers that read in a List of List, or a DataTable object, this one uses reflection to produce a List of row-model objects. It also implements a weak form of duck-typing, since as it fills a row-model it looks for properties of a certain name and expects their type to have a Parse method -- but neither makes nor needs any other assumptions on their type.

A "row model" is simply a class with publicly-settable attributes that match the names of the column headers in the input CSV file. These should be typed usefully, so don't make them all strings. However, each of the types must posess a public static Parse(string) method. Examples of types that natively include such a method are double, int, and DateTime.

Oh, you already have a data model to represent your rows? Super. Now a line like

var inputs = ReadStream<TestClass>(inputStream, true);

will read from inputStream, making inputs a List<TestClass>. The secon parameter (true) intsructs the parser to trim the excess whitespace around each comma-separated entity. (You don't want to try false; it's pretty broken.)

Your input file must have column names on the first row. They need to match -- exactly -- the names of the properties in your row-model class.