My take on https://github.com/kanaka/mal

Variations (hopefully improvements!) on the Mal guide's version and the C# reference implementation.

• Reader uses a queue of tokens that can be reloaded - allows multi-line forms
• READ and EVAL broken out to handle multi-form lines
• Numeric ops can have variable length arg lists (e.g. (+ 1 2 3)) - ref just has dyadic numerics
• Tokeniser can handle negative numbers, decimal-only numbers (e.g. .2) and detect bad numbers (e.g. 1.2.3). To achieve this I replaced the regex-based tokeniser in the C#-Mal ref with a hand-rolled lexer.
• Lots more error checking
• Function bodies can have multiple exprs in their body
• Distinction between structured sequences (i.e. hashmaps) and unstructured sequences (Lists and Vectors). This intentionally precludes some list operations from being applied to hashmaps, primarily to maintain key/ value pairs.