This is a package I use to handle numerical-model parameters, thus the name. However, it should be useful otherwise too. It has two main features:
- keyword type constructors with default values for
- unpacking and packing of composite types and dicts (mostly via UnPack.jl).
Checkout my ten minute JuliaCon 2018 talk.
@with_kw which decorates a type definition to
allow default values and a keyword constructor:
julia> using Parameters julia> @with_kw struct A a::Int = 6 b::Float64 = -1.1 c::UInt8 end julia> A(c=4) A a: 6 b: -1.1 c: 4 julia> A() ERROR: Field 'c' has no default, supply it with keyword. julia> A(c=4, a = 2) A a: 2 b: -1.1 c: 4
The macro also supports constructors for named tuples with default values; e.g.
julia> MyNT = @with_kw (x = 1, y = "foo", z = :(bar)) (::#5) (generic function with 2 methods) julia> MyNT() (x = 1, y = "foo", z = :bar) julia> MyNT(x = 2) (x = 2, y = "foo", z = :bar)
Unpacking is done with
@pack! is similar):
struct B a b c end @unpack a, c = B(4,5,6) # is equivalent to BB = B(4,5,6) a = BB.a c = BB.c
Defining several constants
@consts begin a = 1 b = 2.0 c = "a" end
The features are:
- a keyword constructor for the type
- allows setting default values for the fields inside the type definition
- allows assertions on field values inside the type definition
- a constructor which allows creating a type-instance taking its defaults from another type instance
- packing and unpacking macros for the type:
*is the type name.
- generic packing and unpacking macros
@unpack(work with any types, via UnPack.jl)
@constsmacro to defined a bunch of constants
The keyword-constructor and default-values functionality will probably make it into Julia (# 10146, #533 and #6122) although probably not with all the features present in this package. I suspect that this package should stay usable & useful even after this change lands in Julia. Note that keyword functions are currently slow in Julia, so these constructors should not be used in hot inner loops. However, the normal positional constructor is also provided and could be used in performance critical code.
NEWS.md keeps tabs on updates.
- EponymTuples.jl packing/unpacking of named tuples.
- NamedTupleTools.jl has many named-tuple helper functions
- Setfield.jl for setting immutable fields (i.e. similar to the here provided packing).
Implementing similar things:
Base.@kwdefhas functionality similar to
@with_kwbut more limited. However, with Julia v1.1 its capabilities will be much enhanced, see #29316. If that is enough, ditch the Parameters.jl dependency.
- QuickTypes.jl also provides key-word constructors and default values. Provides a more compact format. I'd say QuickTypes.jl is more suited for REPL and other write-only code, whereas Parameters.jl may be more suited for code which is also read.
- SimpleStructs.jl also provides key-word constructors, default values and assertions. But the syntax is less nice than Parameters.jl.
- ExtractMacro.jl also has
- FieldDefaults.jl also has keyword defaults. You can use it as a minimalist replacement for Parameters.jl with the aid of FieldMetadata.jl and Flatten.jl.
- do copy of fields on (re-)construct?
- think about mutables