Skip to content

NEWS.md

Latest commit

 

History

History
405 lines (222 loc) · 10.8 KB

NEWS.md

File metadata and controls

405 lines (222 loc) · 10.8 KB

User visible changes in TypeUtils

This page describes the most important changes in TypeUtils. The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

Changed

  • Minimal Julia version is 1.0. This bound is imposed by the InverseFunctions package.

Version 1.14.0 (2026-01-06)

This new minor version essentially provides lazymap (see below).

Added

  • B = lazymap([T::Type,], f, A[, f_inv]) yields an object B that lazily maps function f to array or iterator A. Optional argument T is the element-type of B which is automatically inferred if not specified. If A is an (abstract) array, f_inv may be provided to specify the inverse of f. Related type alias LazyMap{T,N,F,typeof(A)} can be used to specialize on the type B with N = ndims(A) and F = typeof(f).

Change

  • The former function as_eltype(T,A) is now a shortcut for lazymap(T,identity,A) unless T == eltype(A) holds in which case A is returned.

  • Private type TypeUtils.AsEltype (formerly used by as_eltytpe) no longer exists.

Version 1.13.0 (2025-12-19)

Added

  • units_of(x) and units_of(typeof(x)) yield the units of the number x. The units are a trait which only depends on the type of the argument. This function is a replacement for Unitful.unit when this package is not yet loaded.

Version 1.12.0 (2025-10-21)

Fixed

  • get_precision on n-tuples yields a correct result and is inferable.

Added

  • get_precision() yields AbtractFloat instead of throwing a MethodError.

Version 1.11.0 (2025-10-16)

Fixed

  • Fix tests for Julia 1.12 and above.

  • Syntax @public a, b, c is allowed.

Version 1.10.0 (2025-07-25)

Added

  • With multiple arguments, get_precision(x, y...) yields the best precision of all the given arguments.

  • nearest(T, x) can deal with dimensionful numbers and types.

Version 1.9.0 (2025-06-20)

This new minor version adds methods to deal with the numerical precision of numbers and objects.

Added

  • get_precision(x) yields the numerical precision of x.

  • adapt_precision(T, x) yields a version of x with numerical precision T.

  • destructure(Vector, x) and destructure(Vector{T}, x) yield a vector while destructure(Tuple, x) yields a tuple as destructure(x).

Version 1.8.0 (2025-06-19)

Added

  • convert_eltype(T,A) for A an instance of LinearAgebra.Factorization.

  • Tests with Aqua.jl.

Fixed

  • A few minor things detected by Aqua.jl have been fixed.

  • Better rule for element type conversion of AbstractUnitRange instances.

  • Fix specialization of Base.return_types which shall return a vector of types.

Version 1.7.1 (2025-05-19)

Fixed

  • is_signed(x) is false if x is a complex number with unsigned rational parts.

Version 1.7.0 (2025-05-13)

Added

  • New is_signed method to infer whether a value of a given numeric type can be negated while retaining the same type.

  • Extend similar for the type of arrays returned by as_eltype.

Version 1.6.0 (2025-04-15)

Added

  • Macro @assert_floating_point to assert whether arguments or variables of a function do use floating-point for storing their value(s).

  • New function assert_floating_point to assert whether an object does use floating-point for storing its value(s). This method is a trait: its result only depend on the type of its argument and this type may be directly specified.

Fixed

  • @public can take macro names (prefixed by @) as argument.

Version 1.5.1 (2025-04-12)

Changed

  • Relax type restriction on T in convert_bare_type(T, x), convert_real_type(T, x), and convert_floating_point_type(T, x).

Fixed

  • Fix new_array(T) and new_array(T,()) to yield a 0-dimensional array.

Version 1.5.0

Added

  • Non-exported public constructor c = TypeUtils.Converter(f,T::Type) such that c is callable and c(x) yields f(T,x) for any x. This replaces and generalizes private type As.

  • Methods as(T), nearest(T), convert_bare_type(T), convert_real_type(T), convert_floating_point_type(T), and convert_eltype(T) yield converters to type T.

Fixed

  • Inference works for more than 3 arguments for bare_type(args...), real_type(args...), and floating_point_type(args...).

Version 1.4.2

Added

  • New nearest(T,x) method to return the value of type T that is the nearest to x. For T integer and x real, it can be seen as rounding with clamping.

Changed

  • Non-exported TypeUtils.@public macro is now public.

Version 1.4.1

Added

  • New macro TypeUtils.@public used to declare non-exported public symbols. Does nothing for Julia versions older than 1.11.

  • For other packages, it may be sufficient to extend convert_eltype(T, X) to a given type X to have convert_eltype(T, x::X) work for instances x of that type.

Changed

  • AbstractUnitRange{<:Integer} has been replaced by AbstractRange{<:Integer} as the eligible type for specifying an array axis. Methods as_array_axis, as_array_axes, and as_array_shape convert index ranges to AbstractUnitRange{Int} throwing an exception if any range does not have unit step. The downside is that this must be explicitly checked for non-AbstractUnitRange{<:Integer} ranges while it can be inferred by type (it is a trait) for AbstractUnitRange{<:Integer} ranges.

  • The exported alias RelaxedArrayShape{N} reflects the above change. Exported alias ArrayShape{N} is restricted to integers and integer-valued unit-ranges which better correspond to Julia's way of representing an N-dimensional array shape.

Fixed

  • TypeStableFunction(f, argtypes...) better tries to infer a suitable concrete type for f with arguments of types argtypes....

Version 1.4.0

Added

  • Non-exported type TypeUtils.Unsupported may be used to provide a fallback implementation of a method for given types of arguments that is only supported when some extension is loaded.

  • new_array(T, inds...) creates an array with elements of type T and shape defined by inds.... The returned array is an OffsetArray{T} if inds... contains any index range other than Base.OneTo and an Array{T} otherwise. In the former case, an exception is thrown if the package OffsetArrays has not been loaded.

Version 1.3.0

Added

Add a few types and methods related to array size and axes:

  • ArrayAxis = AbstractUnitRange{eltype(Dims)} is an alias to the possible canonical types representing a single array index range.

  • ArrayAxes{N} = NTuple{N,ArrayAxis} is an alias to the possible canonical types representing N-dimensional array axes.

  • ArrayShape{N} is an alias to the N-tuple of array dimensions and/or index ranges to which as_array_shape, as_array_size, or as_array_axes are applicable.

  • as_array_shape converts its argument(s) to canonical array axes or to canonical array size.

  • as_array_axes converts its argument(s) to canonical array axes, that is a N-tuple of type ArrayAxes{N}. as_array_axis converts its argument to a single array axis of type ArrayAxis.

  • as_array_size converts its argument(s) to a canonical array size, that is a N-tuple of type Dims{N}. as_array_dim converts its argument to a single array dimension length of type eltype(Dims).

Version 1.2.0

Added

  • to_same_type(x...) is a substitute to promote(x...) that warrants that returned instances have the same type and that calls as(T,x), not convert(T,x), if any conversion is needed.

  • to_same_concrete_type(T...) is a substitute to promote_type(T...) that throws an exception if types T... cannot be promoted to a common concrete type.

Version 1.1.0

Fixed

  • convert_eltype can be applied to a number.

Version 1.0.0

Changed

  • All methods and types formerly provided by Unitless are now provided by TypeUtils which supersedes Unitless.

Version 0.3.8

Changed

  • TwoDimensional is no longer an extension because TwoDimensional version 0.5 directly extends convert as expected by TypeUtils.

Version 0.3.7

Added

  • Method as_return(T, f) builds an instance of TypeStableFunction{T} which is a sub-type of AbstractTypeStableFunction{T}. These two types are both exported.

Version 0.3.6

Changed

  • Methods destructure, destructure!, and restructure are inlined.

Version 0.3.5

Changed

  • Methods destructure, destructure!, restructure, and struct_length can deal with tuples.

Version 0.3.4

Added

  • New methods vals = destructure(obj) or destructure!(vals, obj), and obj = restructure(T, vals) to destructure and object obj as a tuple or vector of its values and, conversely, to rebuild an object of type T from its values.

  • New method struct_length yields the number of values needed to destructure an object.

Version 0.3.3

Fixed

  • Fix convert_eltype(T,A) when A is a range.

Changed

Version 0.3.2

Changed

  • convert_eltype(T,A) yields a tuple if A is a tuple.

Version 0.3.1

Changed

  • convert_eltype(T,A) yields a range if A is a range.

Version 0.3.0

This is the first version as an official Julia package.

Version 0.2.4

Fixed

  • Fix a typo causing as_return to fail when applied to an AsReturn object.

Version 0.2.3

Added

  • New method as_return(T, f) yields a callable object that behaves like f when called as a function except that it lazily converts the value returned by f to type T.

  • New method return_type(f, argtypes...) to infer the type of the result returned by the callable object f when called with arguments of types argtypes....

Version 0.2.2

Added

  • New method promote_eltype(args...) to yield the promoted element type of args....

  • New method convert_eltype(T,A) to convert the element type of A to be T.

  • New method as_eltype(T,A) to lazily convert the element type of A to be T.

Version 0.2.1

Added

  • Method parameterless(T) to get the type T without parameter specifications. For example:

    julia> parameterless(Vector{Float32})
Array

Version 0.2.0

Added

  • Methods as(T,x) to convert x to type T.

  • Call f = as(T) to build a callable object such that f(x) is the same as as(T,x).

  • Extension for TwoDimensonal.

Changed

  • Package was previously named AsType and CastType.