Repository navigation
v0.8.0
Normalization v0.8.0
Revamped the interface
The old Normalization interface had poor type stability and was not very consistent across different normalizations and modifiers.
The new interface is more transparent and type stable, with all 'normalization method' properties now defined as type parameters and all 'normalization instance' parameters defined as fields of concrete AbstractNormalization subtypes.
The modifiers, like Robust and nansafe, now consistently subtype the new AbstractModifier type, and consistent methods are defined for extracting normalization parameters, argument types, and other traits.
These changes also come with various improvements to speed and flexibility, especially for the Robust and NaNSafe modifiers.
The underlying normalization methods are now defined as basic functions of the normalization parameters and scalar values, affording compatibility with InverseFunctions.inverse for automatically calculating the inverse transformation (closes #12).
The DimensionalData, DataFrames, and Unitful extensions have all been improved---particularly UnitfulExt, which can now handle unitful denormalization.
The new interface comes with breaking changes to all syntaxes, so please read the README carefully before updating.