ValueShapes provides Julia types to describe the shape of values, like scalars, arrays and structures.
Shapes provide a generic way to construct uninitialized values (e.g. multidimensional arrays) without using templates.
Shapes also act as a bridge between structured and flat data representations:
Mathematical and statistical algorithms (e.g. optimizers, fitters, solvers,
etc.) often represent variables/parameters as flat vectors of nameless real
values. But user code will usually be more concise and readable if
variables/parameters can have names (e.g. via
NamedTuples) and non-scalar
shapes. ValueShapes provides a duality of view between the two different data
See the documentation for details:
ValueShapes is designed to compose well with ElasticArrays, ArraysOfArrays and TypedTables (and similar table packages). ValueShapes package has some overlap in functionality with TransformVariables, but provides a duality of view instead of transformations (and therefore uses data views instead of data copies, where possible).