Provides functions to deal with modules during compilation time.
It allows a developer to dynamically add, delete and register
attributes, attach documentation and so forth.
After a module is compiled, using many of the functions in
this module will raise errors, since it is out of their scope
to inspect runtime data. Most of the runtime data can be inspected
via the __info__/1 function attached to
each compiled module.
Module attributes
Each module can be decorated with one or more attributes. The following ones
are currently defined by Elixir:
@after_compile
A hook that will be invoked right after the current module is compiled.
Accepts a module or a {module, function_name}. See the "Compile callbacks"
section below.
@after_verify (since v1.14.0)
A hook that will be invoked right after the current module is verified for
undefined functions, deprecations, etc. Accepts a module or a {module, function_name}.
See the "Compile callbacks" section below.
@before_compile
A hook that will be invoked before the module is compiled.
Accepts a module or a {module, function_or_macro_name} tuple.
See the "Compile callbacks" section below.
@behaviour
Note the British spelling!
Behaviours can be referenced by modules to ensure they implement
required specific function signatures defined by @callback.
For example, you could specify a URI.Parser behaviour as follows:
defmodule URI.Parser do
@doc "Defines a default port"
@callback default_port() :: integer
@doc "Parses the given URL"
@callback parse(uri_info :: URI.t()) :: URI.t()
end
And then a module may use it as:
defmodule URI.HTTP do
@behaviour URI.Parser
def default_port(), do: 80
def parse(info), do: info
end
If the behaviour changes or URI.HTTP does not implement
one of the callbacks, a warning will be raised.
For detailed documentation, see the
behaviour typespec documentation.
@impl (since v1.5.0)
To aid in the correct implementation of behaviours, you may optionally declare
@impl for implemented callbacks of a behaviour. This makes callbacks
explicit and can help you to catch errors in your code. The compiler will warn
in these cases:
if you mark a function with @impl when that function is not a callback.
if you don't mark a function with @impl when other functions are marked
with @impl. If you mark one function with @impl, you must mark all
other callbacks for that behaviour as @impl.
@impl works on a per-context basis. If you generate a function through a macro
and mark it with @impl, that won't affect the module where that function is
generated in.
@impl also helps with maintainability by making it clear to other developers
that the function is implementing a callback.
Using @impl, the example above can be rewritten as:
defmodule URI.HTTP do
@behaviour URI.Parser
@impl true
def default_port(), do: 80
@impl true
def parse(info), do: info
end
You may pass either false, true, or a specific behaviour to @impl.
defmodule Foo do
@behaviour Bar
@behaviour Baz
# Will warn if neither Bar nor Baz specify a callback named bar/0.
@impl true
def bar(), do: :ok
# Will warn if Baz does not specify a callback named baz/0.
@impl Baz
def baz(), do: :ok
end
The code is now more readable, as it is now clear which functions are
part of your API and which ones are callback implementations. To reinforce this
idea, @impl true automatically marks the function as @doc false, disabling
documentation unless @doc is explicitly set.
@compile
Defines options for module compilation. This is used to configure
both Elixir and Erlang compilers, as well as any other compilation pass
added by external tools. For example:
defmodule MyModule do
@compile {:inline, my_fun: 1}
def my_fun(arg) do
to_string(arg)
end
end
Multiple uses of @compile will accumulate instead of overriding
previous ones. See the "Compile options" section below.
@deprecated (since v1.6.0)
Provides the deprecation reason for a function. For example:
defmodule Keyword do
@deprecated "Use Kernel.length/1 instead"
def size(keyword) do
length(keyword)
end
end
The Mix compiler automatically looks for calls to deprecated modules
and emit warnings during compilation.
Using the @deprecated attribute will also be reflected in the
documentation of the given function and macro. You can choose between
the @deprecated attribute and the documentation metadata to provide
hard-deprecations (with warnings) and soft-deprecations (without warnings):
This is a soft-deprecation as it simply annotates the documentation
as deprecated:
@doc deprecated: "Use Kernel.length/1 instead"
def size(keyword)
This is a hard-deprecation as it emits warnings and annotates the
documentation as deprecated:
@deprecated "Use Kernel.length/1 instead"
def size(keyword)
Currently @deprecated only supports functions and macros. However
you can use the :deprecated key in the annotation metadata to
annotate the docs of modules, types and callbacks too.
We recommend using this feature with care, especially library authors.
Deprecating code always pushes the burden towards library users. We
also recommend for deprecated functionality to be maintained for long
periods of time, even after deprecation, giving developers plenty of
time to update (except for cases where keeping the deprecated API is
undesired, such as in the presence of security issues).
@doc and @typedoc
Provides documentation for the entity that follows the attribute.
@doc is to be used with a function, macro, callback, or
macrocallback, while @typedoc with a type (public or opaque).
Accepts one of these:
- a string (often a heredoc)
false, which will make the entity invisible to documentation-extraction
tools like ExDoc- a keyword list, since Elixir 1.7.0
For example:
defmodule MyModule do
@typedoc "This type"
@typedoc since: "1.1.0"
@type t :: term
@doc "Hello world"
@doc since: "1.1.0"
def hello do
"world"
end
@doc """
Sums `a` to `b`.
"""
def sum(a, b) do
a + b
end
end
As can be seen in the example above, since Elixir 1.7.0 @doc and @typedoc
also accept a keyword list that serves as a way to provide arbitrary metadata
about the entity. Tools like ExDoc and
IEx may use this information to display annotations. A common use
case is the :since key, which may be used to annotate in which version the
function was introduced.
As illustrated in the example, it is possible to use these attributes
more than once before an entity. However, the compiler will warn if
used twice with binaries as that replaces the documentation text from
the preceding use. Multiple uses with keyword lists will merge the
lists into one.
Note that since the compiler also defines some additional metadata,
there are a few reserved keys that will be ignored and warned if used.
Currently these are: :opaque and :defaults.
Once this module is compiled, this information becomes available via
the Code.fetch_docs/1 function.
@dialyzer
Defines warnings to request or suppress when using :dialyzer.
Accepts an atom, a tuple, or a list of atoms and tuples. For example:
defmodule MyModule do
@dialyzer {:nowarn_function, [my_fun: 1]}
def my_fun(arg) do
M.not_a_function(arg)
end
end
For the list of supported warnings, see :dialyzer module.
Multiple uses of @dialyzer will accumulate instead of overriding
previous ones.
@external_resource
Specifies an external resource for the current module.
Sometimes a module embeds information from an external file. This
attribute allows the module to annotate which external resources
have been used.
Tools may use this information to ensure the module is recompiled
in case any of the external resources change, see for example:
mix compile.elixir.
The specified file path provided is interpreted as relative to
the folder containing the project's mix.exs, which is the
current working directory, not the file where @external_resource
is declared.
If the external resource does not exist, the module still has
a dependency on it, causing the module to be recompiled as soon
as the file is added.
For more control over when a module is recompiled, see
__mix_recompile__?/0.
@file
Changes the filename used in stacktraces for the function or macro that
follows the attribute, such as:
defmodule MyModule do
@doc "Hello world"
@file "hello.ex"
def hello do
"world"
end
end
Note that this is only valid for exceptions/diagnostics that come from the
definition inner scope (which includes its patterns and guards). For example:
defmodule MyModule do # <---- module definition
@file "hello.ex"
defp unused(a) do # <---- function definition
"world" # <---- function scope
end
@file "bye.ex"
def unused(_), do: true
end
If you run this code with the second "unused" definition commented, you will
see that hello.ex is used as the stacktrace when reporting warnings, but if
you uncomment it you'll see that the error will not mention bye.ex, because
it's a module-level error rather than an expression-level error.
@moduledoc
Provides documentation for the current module.
defmodule MyModule do
@moduledoc """
A very useful module.
"""
@moduledoc authors: ["Alice", "Bob"]
end
Accepts a string (often a heredoc) or false where @moduledoc false
will make the module invisible to documentation extraction tools like
ExDoc.
Similarly to @doc also accepts a keyword list to provide metadata
about the module. For more details, see the documentation of @doc
above.
Once this module is compiled, this information becomes available via
the Code.fetch_docs/1 function.
@nifs (since v1.16.0)
A list of functions and their arities which will be overridden
by a native implementation (NIF).
defmodule MyLibrary.MyModule do
@nifs [foo: 1, bar: 2]
def foo(arg1), do: :erlang.nif_error(:not_loaded)
def bar(arg1, arg2), do: :erlang.nif_error(:not_loaded)
end
See the Erlang documentation for more information:
https://www.erlang.org/doc/man/erl_nif
@on_definition
A hook that will be invoked when each function or macro in the current
module is defined. Useful when annotating functions.
Accepts a module or a {module, function_name} tuple. The function
must take 6 arguments:
- the module environment
- the kind of the function/macro:
:def, :defp, :defmacro, or :defmacrop - the function/macro name
- the list of quoted arguments
- the list of quoted guards
- the quoted function body
If the function/macro being defined has multiple clauses, the hook will
be called for each clause.
Unlike other hooks, @on_definition will only invoke functions and
never macros. This is to avoid @on_definition callbacks from
redefining functions that have just been defined in favor of more
explicit approaches.
When just a module is provided, the function is assumed to be
__on_definition__/6.
Example
defmodule Hooks do
def on_def(_env, kind, name, args, guards, body) do
IO.puts("Defining #{kind} named #{name} with args:")
IO.inspect(args)
IO.puts("and guards")
IO.inspect(guards)
IO.puts("and body")
IO.puts(Macro.to_string(body))
end
end
defmodule MyModule do
@on_definition {Hooks, :on_def}
def hello(arg) when is_binary(arg) or is_list(arg) do
"Hello" <> to_string(arg)
end
def hello(_) do
:ok
end
end
@on_load
A hook that will be invoked whenever the module is loaded.
Accepts the function name (as an atom) of a function in the current module.
The function must have an arity of 0 (no arguments). If the function does
not return :ok, the loading of the module will be aborted. Its primary
use case is to load NIFs:
defmodule MyModule do
@on_load :load_external_code
def load_external_code do
:erlang.load_nif(~c"path/to/extension.so_or_dll")
end
end
The function given to on_load should avoid calling functions from
other modules. This is because, when running a mix release,
on_load runs extremely early, before any application starts running,
and therefore even systems like the Logger and IO are not yet available.
@vsn
Specify the module version. Accepts any valid Elixir value, for example:
defmodule MyModule do
@vsn "1.0"
end
Struct attributes
@derive - derives an implementation for the given protocol for the
struct defined in the current module
@enforce_keys - ensures the given keys are always set when building
the struct defined in the current module
See defstruct/1 for more information on building and using structs.
Typespec attributes
The following attributes are part of typespecs and are also built-in in
Elixir:
@type - defines a type to be used in @spec@typep - defines a private type to be used in @spec@opaque - defines an opaque type to be used in @spec@spec - provides a specification for a function@callback - provides a specification for a behaviour callback (and generates
a behaviour_info/1 function in the module, see below)@macrocallback - provides a specification for a macro behaviour callback@optional_callbacks - specifies which behaviour callbacks and macro
behaviour callbacks are optional@impl - declares an implementation of a callback function or macro
For detailed documentation, see the typespec documentation.
Custom attributes
In addition to the built-in attributes outlined above, custom attributes may
also be added. Custom attributes are expressed using the @/1 operator followed
by a valid variable name. The value given to the custom attribute must be a valid
Elixir value:
defmodule MyModule do
@custom_attr [some: "stuff"]
end
For more advanced options available when defining custom attributes, see
register_attribute/3.
Compile callbacks
There are three compilation callbacks, invoked in this order:
@before_compile, @after_compile, and @after_verify.
They are described next.
@before_compile
A hook that will be invoked before the module is compiled. This is
often used to change how the current module is being compiled.
Accepts a module or a {module, function_or_macro_name} tuple. The
function/macro must take one argument: the module environment. If
it's a macro, its returned value will be injected at the end of the
module definition before the compilation starts.
When just a module is provided, the function/macro is assumed to be
__before_compile__/1.
Callbacks will run in the order they are registered. Any overridable
definition will be made concrete before the first callback runs.
A definition may be made overridable again in another before compile
callback and it will be made concrete one last time after all callbacks
run.
Note: the callback function/macro must be placed in a separate module
(because when the callback is invoked, the current module does not yet exist).
Example
defmodule A do
defmacro __before_compile__(_env) do
quote do
def hello, do: "world"
end
end
end
defmodule B do
@before_compile A
end
B.hello()
#=> "world"
@after_compile
A hook that will be invoked with the bytecode of the current module.
Although the module has already been compiled, its bytecode may not have
been loaded to memory nor written to disk. For those reasons, prefer to
use @after_verify callbacks or use Code.ensure_compiled!/1 to wait
until the module is fully available for introspection/invocation.
Accepts a module or a {module, function_name} tuple. The function
must take two arguments: the module environment and its bytecode.
When just a module is provided, the function is assumed to be
__after_compile__/2.
Callbacks will run in the order they are registered.
Module functions expecting not yet compiled modules (such as definitions_in/1)
are still available at the time @after_compile is invoked.
Example
defmodule MyModule do
@after_compile __MODULE__
def __after_compile__(env, _bytecode) do
IO.inspect(env)
end
end
@after_verify
A hook that will be invoked right after the current module is verified for
undefined functions, deprecations, etc. A module is always verified after
it is compiled. In Mix projects, a module is also verified when any of its
runtime dependencies change. Therefore this is useful to perform verification
of the current module while avoiding compile-time dependencies. Given the
callback is invoked under different scenarios, Elixir provides no guarantees
of when in the compilation cycle nor in which process the callback runs.
Furthermore, after verification callbacks are not expected to raise.
Given they run after the code is compiled, artifacts have already been
written to disk, and therefore raising does not effectively halt compilation
and may leave unused artifacts on disk. If you must raise, use @after_compile
or other callback. Given modules have already been compiled, functions in
this module, such as get_attribute/2, which expect modules to not have been
yet compiled, do not work on @after_verify callback.
Accepts a module or a {module, function_name} tuple. The function
must take one argument: the module name. When just a module is provided,
the function is assumed to be __after_verify__/1.
Callbacks will run in the order they are registered.
Example
defmodule MyModule do
@after_verify __MODULE__
def __after_verify__(module) do
IO.inspect(module)
:ok
end
end
Compile options
The @compile attribute accepts different options that are used by both
Elixir and Erlang compilers. Some of the common use cases are documented
below:
@compile :debug_info - includes :debug_info regardless of the
corresponding setting in Code.get_compiler_option/1
@compile {:debug_info, false} - disables :debug_info regardless
of the corresponding setting in Code.get_compiler_option/1. Note
disabling :debug_info is not recommended as it removes the ability
of the Elixir compiler and other tools to static analyse the code.
If you want to remove the :debug_info while deploying, tools like
mix release already do such by default.
@compile {:inline, some_fun: 2, other_fun: 3} - inlines the given
name/arity pairs. Inlining is applied locally, calls from another
module are not affected by this option
@compile {:autoload, true} - configures if modules are automatically
loaded after definition. It defaults to false when compiling modules
to .beam files in disk (as the modules are then lazily loaded from
disk). If modules are not compiled to disk, then they are always loaded,
regardless of this flag
@compile {:no_warn_undefined, Mod} or
@compile {:no_warn_undefined, {Mod, fun, arity}} - does not warn if
the given module or the given Mod.fun/arity are not defined
Generated functions
Sometimes the compiler will generate public functions within modules. These
are documented below.
behaviour_info/1
This function is generated for modules that define a behaviour, that is,
that have one or more @callback definitions. The signature for this function,
expressed as a spec, is:
@spec behaviour_info(:callbacks) :: [function_info]
when function_info: {function_name :: atom(), arity :: non_neg_integer()}
@spec behaviour_info(:optional_callbacks) :: [function_info]
when function_info: {function_name :: atom(), arity :: non_neg_integer()}
behaviour_info(:callbacks) includes optional callbacks.
For example:
iex> Enum.sort(GenServer.behaviour_info(:callbacks))
[
code_change: 3,
format_status: 1,
format_status: 2,
handle_call: 3,
handle_cast: 2,
handle_continue: 2,
handle_info: 2,
init: 1,
terminate: 2
]
module_info/0
This function is generated for all modules. It returns all the attributes
returned by module_info/1 (see below), but as a single keyword list. See also the
Erlang documentation.
module_info/1
This function is generated for all modules and returns
information about the module. The signature for this function,
expressed as a spec, is:
@spec module_info(:module) :: module() # Returns the module itself
@spec module_info(:attributes) :: keyword()
@spec module_info(:compile) :: keyword()
@spec module_info(:md5) :: binary()
@spec module_info(:nifs) :: [function_info]
when function_info: {function_name :: atom(), arity :: non_neg_integer()}
@spec module_info(:exports) :: [function_info]
when function_info: {function_name :: atom(), arity :: non_neg_integer()}
@spec module_info(:functions) :: [function_info]
when function_info: {function_name :: atom(), arity :: non_neg_integer()}
For example:
iex> URI.module_info(:module)
URI
iex> {:decode_www_form, 1} in URI.module_info(:exports)
true
For more information about module_info/1, also check out the Erlang
documentation.
__info__/1
This function is generated for all modules. It's similar to module_info/1 but
includes some additional Elixir-specific information, such as struct and macro
information. For documentation, see Module.__info__/1.