The following sections explain a few aspects of idiomatic Julia coding style. None of these rules are absolute; they are only suggestions to help familiarize you with the language and to help you choose among alternative designs.
Use 4 spaces per indentation level.
Writing code as a series of steps at the top level is a quick way to get started solving a problem, but you should try to divide a program into functions as soon as possible. Functions are more reusable and testable, and clarify what steps are being done and what their inputs and outputs are. Furthermore, code inside functions tends to run much faster than top level code, due to how Julia's compiler works.
It is also worth emphasizing that functions should take arguments, instead of operating directly
on global variables (aside from constants like pi).
Code should be as generic as possible. Instead of writing:
Complex{Float64}(x)it's better to use available generic functions:
complex(float(x))The second version will convert x to an appropriate type, instead of always the same type.
This style point is especially relevant to function arguments. For example, don't declare an argument
to be of type Int or Int32 if it really could be any integer, expressed with the abstract
type Integer. In fact, in many cases you can omit the argument type altogether,
unless it is needed to disambiguate from other method definitions, since a
MethodError will be thrown anyway if a type is passed that does not support any
of the requisite operations. (This is known as
duck typing.)
For example, consider the following definitions of a function addone that returns one plus its
argument:
addone(x::Int) = x + 1 # works only for Int
addone(x::Integer) = x + oneunit(x) # any integer type
addone(x::Number) = x + oneunit(x) # any numeric type
addone(x) = x + oneunit(x) # any type supporting + and oneunitThe last definition of addone handles any type supporting oneunit (which returns 1 in
the same type as x, which avoids unwanted type promotion) and the + function with
those arguments. The key thing to realize is that there is no performance penalty to defining
only the general addone(x) = x + oneunit(x), because Julia will automatically compile specialized
versions as needed. For example, the first time you call addone(12), Julia will automatically
compile a specialized addone function for x::Int arguments, with the call to oneunit
replaced by its inlined value 1. Therefore, the first three definitions of addone above are
completely redundant with the fourth definition.
Instead of:
function foo(x, y)
x = Int(x); y = Int(y)
...
end
foo(x, y)use:
function foo(x::Int, y::Int)
...
end
foo(Int(x), Int(y))This is better style because foo does not really accept numbers of all types; it really needs
Int s.
One issue here is that if a function inherently requires integers, it might be better to force the caller to decide how non-integers should be converted (e.g. floor or ceiling). Another issue is that declaring more specific types leaves more "space" for future method definitions.
Instead of:
function double(a::AbstractArray{<:Number})
for i in eachindex(a)
a[i] *= 2
end
return a
enduse:
function double!(a::AbstractArray{<:Number})
for i in eachindex(a)
a[i] *= 2
end
return a
endJulia Base uses this convention throughout and contains examples of functions
with both copying and modifying forms (e.g., sort and sort!), and others
which are just modifying (e.g., push!, pop!, splice!). It
is typical for such functions to also return the modified array for convenience.
Functions related to IO or making use of random number generators (RNG) are notable exceptions:
Since these functions almost invariably must mutate the IO or RNG, functions ending with ! are used to signify a mutation other than mutating the IO or advancing the RNG state.
For example, rand(x) mutates the RNG, whereas rand!(x) mutates both the RNG and x; similarly, read(io) mutates io, whereas read!(io, x) mutates both arguments.
Types such as Union{Function,AbstractString} are often a sign that some design could be cleaner.
It is usually not much help to construct arrays like the following:
a = Vector{Union{Int,AbstractString,Tuple,Array}}(undef, n)In this case Vector{Any}(undef, n) is better. It is also more helpful to the compiler to annotate specific
uses (e.g. a[i]::Int) than to try to pack many alternatives into one type.
Idiomatic Julia code should generally treat a module's exported methods as the interface to its types. An object's fields are generally considered implementation details and user code should only access them directly if this is stated to be the API. This has several benefits:
- Package developers are freer to change the implementation without breaking user code.
- Methods can be passed to higher-order constructs like
map(e.g.map(imag, zs)) rather than[z.im for z in zs]). - Methods can be defined on abstract types.
- Methods can describe a conceptual operation that can be shared across
disparate types (e.g.
real(z)works on Complex numbers or Quaternions).
Julia's dispatch system encourages this style because play(x::MyType) only
defines the play method on that particular type, leaving other types to
have their own implementation.
Similarly, non-exported functions are typically internal and subject to change,
unless the documentations states otherwise. Names sometimes are given a _ prefix
(or suffix) to further suggest that something is "internal" or an
implementation-detail, but it is not a rule.
Counter-examples to this rule include NamedTuple, [RegexMatch](@ref match), [StatStruct](@ref stat).
- modules and type names use capitalization and camel case:
module SparseArrays,struct UnitRange. - functions are lowercase (
maximum,convert) and, when readable, with multiple words squashed together (isequal,haskey). When necessary, use underscores as word separators. Underscores are also used to indicate a combination of concepts (remotecall_fetchas a more efficient implementation offetch(remotecall(...))) or as modifiers. - functions mutating at least one of their arguments end in
!. - conciseness is valued, but avoid abbreviation (
indexinrather thanindxin) as it becomes difficult to remember whether and how particular words are abbreviated.
If a function name requires multiple words, consider whether it might represent more than one concept and might be better split into pieces.
As a general rule, the Base library uses the following order of arguments to functions, as applicable:
-
Function argument. Putting a function argument first permits the use of
doblocks for passing multiline anonymous functions. -
I/O stream. Specifying the
IOobject first permits passing the function to functions such assprint, e.g.sprint(show, x). -
Input being mutated. For example, in [
fill!(x, v)](@ref fill!),xis the object being mutated and it appears before the value to be inserted intox. -
Type. Passing a type typically means that the output will have the given type. In [
parse(Int, "1")](@ref parse), the type comes before the string to parse. There are many such examples where the type appears first, but it's useful to note that in [read(io, String)](@ref read), theIOargument appears before the type, which is in keeping with the order outlined here. -
Input not being mutated. In
fill!(x, v),vis not being mutated and it comes afterx. -
Key. For associative collections, this is the key of the key-value pair(s). For other indexed collections, this is the index.
-
Value. For associative collections, this is the value of the key-value pair(s). In cases like [
fill!(x, v)](@ref fill!), this isv. -
Everything else. Any other arguments.
-
Varargs. This refers to arguments that can be listed indefinitely at the end of a function call. For example, in
Matrix{T}(undef, dims), the dimensions can be given as aTuple, e.g.Matrix{T}(undef, (1,2)), or asVarargs, e.g.Matrix{T}(undef, 1, 2). -
Keyword arguments. In Julia keyword arguments have to come last anyway in function definitions; they're listed here for the sake of completeness.
The vast majority of functions will not take every kind of argument listed above; the numbers merely denote the precedence that should be used for any applicable arguments to a function.
There are of course a few exceptions.
For example, in convert, the type should always come first.
In setindex!, the value comes before the indices so that the indices can be
provided as varargs.
When designing APIs, adhering to this general order as much as possible is likely to give users of your functions a more consistent experience.
It is better to avoid errors than to rely on catching them.
Julia doesn't require parens around conditions in if and while. Write:
if a == binstead of:
if (a == b)Splicing function arguments can be addictive. Instead of [a..., b...], use simply [a; b],
which already concatenates arrays. collect(a) is better than [a...], but since a
is already iterable it is often even better to leave it alone, and not convert it to an array.
When a method T(x) is called on a type T, it is generally expected to return a value of type T.
Defining a [constructor](@ref man-constructors) that returns an unexpected type can lead to confusing and unpredictable behavior:
julia> struct Foo{T}
x::T
end
julia> Base.Float64(foo::Foo) = Foo(Float64(foo.x)) # Do not define methods like this
julia> Float64(Foo(3)) # Should return `Float64`
Foo{Float64}(3.0)
julia> Foo{Int}(x) = Foo{Float64}(x) # Do not define methods like this
julia> Foo{Int}(3) # Should return `Foo{Int}`
Foo{Float64}(3.0)
To maintain code clarity and ensure type consistency, always design constructors to return an instance of the type they are supposed to construct.
A function signature:
foo(x::T) where {T<:Real} = ...should be written as:
foo(x::Real) = ...instead, especially if T is not used in the function body. Even if T is used, it can be replaced
with typeof(x) if convenient. There is no performance difference. Note that this is
not a general caution against static parameters, just against uses where they are not needed.
Note also that container types, specifically may need type parameters in function calls. See the FAQ Avoid fields with abstract containers for more information.
Sets of definitions like the following are confusing:
foo(::Type{MyType}) = ...
foo(::MyType) = foo(MyType)Decide whether the concept in question will be written as MyType or MyType(), and stick to
it.
The preferred style is to use instances by default, and only add methods involving Type{MyType}
later if they become necessary to solve some problems.
If a type is effectively an enumeration, it should be defined as a single (ideally immutable struct or primitive) type, with the enumeration values being instances of it. Constructors and conversions can check whether values are valid. This design is preferred over making the enumeration an abstract type, with the "values" as subtypes.
Be aware of when a macro could really be a function instead.
Calling eval inside a macro is a particularly dangerous warning sign; it means the
macro will only work when called at the top level. If such a macro is written as a function instead,
it will naturally have access to the run-time values it needs.
If you have a type that uses a native pointer:
mutable struct NativeType
p::Ptr{UInt8}
...
enddon't write definitions like the following:
getindex(x::NativeType, i) = unsafe_load(x.p, i)The problem is that users of this type can write x[i] without realizing that the operation is
unsafe, and then be susceptible to memory bugs.
Such a function should either check the operation to ensure it is safe, or have unsafe somewhere
in its name to alert callers.
It is possible to write definitions like the following:
show(io::IO, v::Vector{MyType}) = ...This would provide custom showing of vectors with a specific new element type. While tempting,
this should be avoided. The trouble is that users will expect a well-known type like Vector()
to behave in a certain way, and overly customizing its behavior can make it harder to work with.
"Type piracy" refers to the practice of extending or redefining methods in Base
or other packages on types that you have not defined. In extreme cases, you can crash Julia
(e.g. if your method extension or redefinition causes invalid input to be passed to a
ccall). Type piracy can complicate reasoning about code, and may introduce
incompatibilities that are hard to predict and diagnose.
As an example, suppose you wanted to define multiplication on symbols in a module:
module A
import Base.*
*(x::Symbol, y::Symbol) = Symbol(x,y)
endThe problem is that now any other module that uses Base.* will also see this definition.
Since Symbol is defined in Base and is used by other modules, this can change the
behavior of unrelated code unexpectedly. There are several alternatives here, including
using a different function name, or wrapping the Symbols in another type that you define.
Sometimes, coupled packages may engage in type piracy to separate features from definitions, especially when the packages were designed by collaborating authors, and when the definitions are reusable. For example, one package might provide some types useful for working with colors; another package could define methods for those types that enable conversions between color spaces. Another example might be a package that acts as a thin wrapper for some C code, which another package might then pirate to implement a higher-level, Julia-friendly API.
You generally want to use isa and <: for testing types,
not ==. Checking types for exact equality typically only makes sense when comparing to a known
concrete type (e.g. T == Float64), or if you really, really know what you're doing.
Since higher-order functions are often called with anonymous functions, it is easy to conclude
that this is desirable or even necessary. But any function can be passed directly, without being
"wrapped" in an anonymous function. Instead of writing map(x->f(x), a), write map(f, a).
If you write generic code which handles numbers, and which can be expected to run with many different numeric type arguments, try using literals of a numeric type that will affect the arguments as little as possible through promotion.
For example,
julia> f(x) = 2.0 * x
f (generic function with 1 method)
julia> f(1//2)
1.0
julia> f(1/2)
1.0
julia> f(1)
2.0
while
julia> g(x) = 2 * x
g (generic function with 1 method)
julia> g(1//2)
1//1
julia> g(1/2)
1.0
julia> g(1)
2
As you can see, the second version, where we used an Int literal, preserved the type of the
input argument, while the first didn't. This is because e.g. promote_type(Int, Float64) == Float64,
and promotion happens with the multiplication. Similarly, Rational literals are less type disruptive
than Float64 literals, but more disruptive than Ints:
julia> h(x) = 2//1 * x
h (generic function with 1 method)
julia> h(1//2)
1//1
julia> h(1/2)
1.0
julia> h(1)
2//1
Thus, use Int literals when possible, with Rational{Int} for literal non-integer numbers,
in order to make it easier to use your code.