Tracking issue for modernization efforts

[regexident]

Hi Matt,

I'd like to contribute a series of improvements to Surge.

Surge is already pretty nice for implementing performant number crunching and machine learning algorithms, but I find myself writing my own Vector<Scalar> types and adding extensions for common Matrix<Scalar> operations over and over again (especially the scalar variants), so I figured I might as well upstream them:

• Upgrade to recommended project settings. (https://github.com/mattt/Surge/pull/109)
• Add overload of XCTAssertEqual(_:_:accuracy:) for collections of FloatingPoint. (https://github.com/mattt/Surge/pull/104)
• Add basic suite of unit tests for arithmetic functions. (https://github.com/mattt/Surge/pull/105)
• Add basic suite of unit tests for matrix arithmetic functions. (https://github.com/mattt/Surge/pull/108)
• Add missing named function equivalents of .-operators (e.g. addInPlace(_:_:) for .+=(_:_:)), transforming operators into strictly forwarding calls of explicitly named equivalents. (https://github.com/mattt/Surge/pull/107)
• Add Matrix initializers for common diagonal/eye/identity patterns. (https://github.com/mattt/Surge/pull/111)
• Optimize scalar-variants of arithmetic functions (e.g. func mul(x: [Float], y: Float), etc.) where applicable.
• Add missing scalar-variants of matrix functions (e.g. func mul(x: Matrix<Float>, y: Float), etc.). (https://github.com/mattt/Surge/pull/110)
• Add Vector<Scalar> type and unit tests for performant, yet expressive vector arithmetic. (https://github.com/mattt/Surge/pull/110)
• Add matrix-vector multiplication and other arithmetic functions (using Matrix<Scalar> & Vector<Scalar>). (https://github.com/mattt/Surge/pull/110)

---

Why Vector<Scalar> if we already have [Scalar]?

Simply using [Float] requires one to use .+ & Co, due to otherwise colliding with array concatenation, etc. Having an explicit Vector<Scalar> type also allows for better auto-completion by keeping all the non-vector-space-related Array/Collection gunk out of it. Or and it also doesn't spill every operator/function into the global namespace of Collection.