glib_macros/

lib.rs

// Take a look at the license at the top of the repository in the LICENSE file.

mod async_test;
mod boxed_derive;
mod clone;
mod clone_old;
mod closure;
mod closure_old;
mod derived_properties_attribute;
mod downgrade_derive;
mod enum_derive;
mod error_domain_derive;
mod flags_attribute;
mod object_impl_attributes;
mod properties;
mod shared_boxed_derive;
mod value_delegate_derive;
mod variant_derive;

mod utils;

use flags_attribute::AttrInput;
use proc_macro::{TokenStream, TokenTree};
use proc_macro2::Span;
use syn::{parse_macro_input, DeriveInput};
use utils::{parse_nested_meta_items_from_stream, NestedMetaItem};

/// Macro for passing variables as strong or weak references into a closure.
///
/// This macro can be useful in combination with closures, e.g. signal handlers, to reduce the
/// boilerplate required for passing strong or weak references into the closure. It will
/// automatically create the new reference and pass it with the same name into the closure.
///
/// If upgrading the weak reference to a strong reference inside the closure is failing, the
/// closure is immediately returning an optional default return value. If none is provided, `()` is
/// returned.
///
/// **⚠️ IMPORTANT ⚠️**
///
/// `glib` needs to be in scope, so unless it's one of the direct crate dependencies, you need to
/// import it because `clone!` is using it. For example:
///
/// ```rust,ignore
/// use gtk::glib;
/// ```
///
/// ### Debugging
///
/// In case something goes wrong inside the `clone!` macro, we use the [`g_debug`] macro. Meaning
/// that if you want to see these debug messages, you'll have to set the `G_MESSAGES_DEBUG`
/// environment variable when running your code (either in the code directly or when running the
/// binary) to either "all" or [`CLONE_MACRO_LOG_DOMAIN`]:
///
/// [`g_debug`]: ../glib/macro.g_debug.html
/// [`CLONE_MACRO_LOG_DOMAIN`]: ../glib/constant.CLONE_MACRO_LOG_DOMAIN.html
///
/// ```rust,ignore
/// use glib::CLONE_MACRO_LOG_DOMAIN;
///
/// std::env::set_var("G_MESSAGES_DEBUG", CLONE_MACRO_LOG_DOMAIN);
/// std::env::set_var("G_MESSAGES_DEBUG", "all");
/// ```
///
/// Or:
///
/// ```bash
/// $ G_MESSAGES_DEBUG=all ./binary
/// ```
///
/// ### Passing a strong reference
///
/// ```
/// use glib;
/// use glib_macros::clone;
/// use std::rc::Rc;
///
/// let v = Rc::new(1);
/// let closure = clone!(
///     #[strong] v,
///     move |x| {
///         println!("v: {}, x: {}", v, x);
///     },
/// );
///
/// closure(2);
/// ```
///
/// ### Passing a weak reference
///
/// ```
/// use glib;
/// use glib_macros::clone;
/// use std::rc::Rc;
///
/// let u = Rc::new(2);
/// let closure = clone!(
///     #[weak]
///     u,
///     move |x| {
///         println!("u: {}, x: {}", u, x);
///     },
/// );
///
/// closure(3);
/// ```
///
/// #### Allowing a nullable weak reference
///
/// In some cases, even if the weak references can't be retrieved, you might want to still have
/// your closure called. In this case, you need to use `#[weak_allow_none]` instead of `#[weak]`:
///
/// ```
/// use glib;
/// use glib_macros::clone;
/// use std::rc::Rc;
///
/// let closure = {
///     // This `Rc` won't be available in the closure because it's dropped at the end of the
///     // current block
///     let u = Rc::new(2);
///     clone!(
///         #[weak_allow_none]
///         u,
///         move |x| {
///             // We need to use a Debug print for `u` because it'll be an `Option`.
///             println!("u: {:?}, x: {}", u, x);
///             true
///         },
///     )
/// };
///
/// assert_eq!(closure(3), true);
/// ```
///
/// ### Creating owned values from references (`ToOwned`)
///
/// ```
/// use glib;
/// use glib_macros::clone;
///
/// let v = "123";
/// let closure = clone!(
///     #[to_owned] v,
///     move |x| {
///         // v is passed as `String` here
///         println!("v: {}, x: {}", v, x);
///     },
/// );
///
/// closure(2);
/// ```
///
/// ### Renaming variables
///
/// ```
/// use glib;
/// use glib_macros::clone;
/// use std::rc::Rc;
///
/// let v = Rc::new(1);
/// let u = Rc::new(2);
/// let closure = clone!(
///     #[strong(rename_to = y)]
///     v,
///     #[weak] u,
///     move |x| {
///         println!("v as y: {}, u: {}, x: {}", y, u, x);
///     },
/// );
///
/// closure(3);
/// ```
///
/// ### Providing a return value if upgrading a weak reference fails
///
/// By default, `()` is returned if upgrading a weak reference fails. This behaviour can be
/// adjusted in two different ways:
///
/// Either by providing the value yourself using one of
///
///   * `#[upgrade_or]`: Requires an expression that returns a `Copy` value of the expected return type,
///   * `#[upgrade_or_else]`: Requires a closure that returns a value of the expected return type,
///   * `#[upgrade_or_default]`: Requires that the return type implements `Default` and returns that.
///
/// ```
/// use glib;
/// use glib_macros::clone;
/// use std::rc::Rc;
///
/// let v = Rc::new(1);
/// let closure = clone!(
///     #[weak] v,
///     #[upgrade_or]
///     false,
///     move |x| {
///         println!("v: {}, x: {}", v, x);
///         true
///     },
/// );
///
/// // Drop value so that the weak reference can't be upgraded.
/// drop(v);
///
/// assert_eq!(closure(2), false);
/// ```
///
/// Or by using `#[upgrade_or_panic]`: If the value fails to get upgraded, it'll panic.
///
/// ```should_panic
/// # use glib;
/// # use glib_macros::clone;
/// # use std::rc::Rc;
/// # let v = Rc::new(1);
/// let closure = clone!(
///     #[weak] v,
///     #[upgrade_or_panic]
///     move |x| {
///         println!("v: {}, x: {}", v, x);
///         true
///     },
/// );
/// # drop(v);
/// # assert_eq!(closure(2), false);
/// ```
///
/// ### Errors
///
/// Here is a list of errors you might encounter:
///
/// **Missing `#[weak]` or `#[strong]`**:
///
/// ```compile_fail
/// # use glib;
/// # use glib_macros::clone;
/// # use std::rc::Rc;
/// let v = Rc::new(1);
///
/// let closure = clone!(
///     v,
///     move |x| println!("v: {}, x: {}", v, x),
/// );
/// # drop(v);
/// # closure(2);
/// ```
///
/// **Passing `self` as an argument**:
///
/// ```compile_fail
/// # use glib;
/// # use glib_macros::clone;
/// # use std::rc::Rc;
/// #[derive(Debug)]
/// struct Foo;
///
/// impl Foo {
///     fn foo(&self) {
///         let closure = clone!(
///             #[strong] self,
///             move |x| {
///                 println!("self: {:?}", self);
///             },
///         );
///         # closure(2);
///     }
/// }
/// ```
///
/// If you want to use `self` directly, you'll need to rename it:
///
/// ```
/// # use glib;
/// # use glib_macros::clone;
/// # use std::rc::Rc;
/// #[derive(Debug)]
/// struct Foo;
///
/// impl Foo {
///     fn foo(&self) {
///         let closure = clone!(
///             #[strong(rename_to = this)]
///             self,
///             move |x| {
///                 println!("self: {:?}", this);
///             },
///         );
///         # closure(2);
///     }
/// }
/// ```
///
/// **Passing fields directly**
///
/// ```compile_fail
/// # use glib;
/// # use glib_macros::clone;
/// # use std::rc::Rc;
/// #[derive(Debug)]
/// struct Foo {
///     v: Rc<usize>,
/// }
///
/// impl Foo {
///     fn foo(&self) {
///         let closure = clone!(
///             #[strong] self.v,
///             move |x| {
///                 println!("self.v: {:?}", v);
///             },
///         );
///         # closure(2);
///     }
/// }
/// ```
///
/// You can do it by renaming it:
///
/// ```
/// # use glib;
/// # use glib_macros::clone;
/// # use std::rc::Rc;
/// # struct Foo {
/// #     v: Rc<usize>,
/// # }
/// impl Foo {
///     fn foo(&self) {
///         let closure = clone!(
///             #[strong(rename_to = v)]
///             self.v,
///             move |x| {
///                 println!("self.v: {}", v);
///             },
///         );
///         # closure(2);
///     }
/// }
/// ```
#[proc_macro]
pub fn clone(item: TokenStream) -> TokenStream {
    // Check if this is an old-style clone macro invocation.
    // These always start with an '@' punctuation.
    let Some(first) = item.clone().into_iter().next() else {
        return syn::Error::new(Span::call_site(), "expected a closure or async block")
            .to_compile_error()
            .into();
    };

    match first {
        TokenTree::Punct(ref p) if p.to_string() == "@" => clone_old::clone_inner(item),
        _ => clone::clone_inner(item),
    }
}

/// Macro for creating a [`Closure`] object. This is a wrapper around [`Closure::new`] that
/// automatically type checks its arguments at run-time.
///
/// A `Closure` takes [`Value`] objects as inputs and output. This macro will automatically convert
/// the inputs to Rust types when invoking its callback, and then will convert the output back to a
/// `Value`. All inputs must implement the [`FromValue`] trait, and outputs must either implement
/// the [`ToValue`] trait or be the unit type `()`. Type-checking of inputs is done at run-time; if
/// incorrect types are passed via [`Closure::invoke`] then the closure will panic. Note that when
/// passing input types derived from [`Object`] or [`Interface`], you must take care to upcast to
/// the exact object or interface type that is being received.
///
/// Similarly to [`clone!`](crate::clone!), this macro can be useful in combination with signal
/// handlers to reduce boilerplate when passing references. Unique to `Closure` objects is the
/// ability to watch an object using the `#[watch]` attribute. Only an [`Object`] value can be
/// passed to `#[watch]`, and only one object can be watched per closure. When an object is watched,
/// a weak reference to the object is held in the closure. When the object is destroyed, the
/// closure will become invalidated: all signal handlers connected to the closure will become
/// disconnected, and any calls to [`Closure::invoke`] on the closure will be silently ignored.
/// Internally, this is accomplished using [`Object::watch_closure`] on the watched object.
///
/// The `#[weak]`, `#[weak_allow_none]`, `#[strong]`, `#[to_owned]` captures are also supported and
/// behave the same as in [`clone!`](crate::clone!), as is aliasing captures via `rename_to`.
/// Similarly, upgrade failure of weak references can be adjusted via `#[upgrade_or]`,
/// `#[upgrade_or_else]`, `#[upgrade_or_default]` and `#[upgrade_or_panic]`.
///
/// Notably, these captures are able to reference `Rc` and `Arc` values in addition to `Object`
/// values.
///
/// [`Closure`]: ../glib/closure/struct.Closure.html
/// [`Closure::new`]: ../glib/closure/struct.Closure.html#method.new
/// [`Closure::new_local`]: ../glib/closure/struct.Closure.html#method.new_local
/// [`Closure::invoke`]: ../glib/closure/struct.Closure.html#method.invoke
/// [`Value`]: ../glib/value/struct.Value.html
/// [`FromValue`]: ../glib/value/trait.FromValue.html
/// [`ToValue`]: ../glib/value/trait.ToValue.html
/// [`Interface`]: ../glib/object/struct.Interface.html
/// [`Object`]: ../glib/object/struct.Object.html
/// [`Object::watch_closure`]: ../glib/object/trait.ObjectExt.html#tymethod.watch_closure
/// **⚠️ IMPORTANT ⚠️**
///
/// `glib` needs to be in scope, so unless it's one of the direct crate dependencies, you need to
/// import it because `closure!` is using it. For example:
///
/// ```rust,ignore
/// use gtk::glib;
/// ```
///
/// ### Using as a closure object
///
/// ```
/// use glib_macros::closure;
///
/// let concat_str = closure!(|s: &str| s.to_owned() + " World");
/// let result = concat_str.invoke::<String>(&[&"Hello"]);
/// assert_eq!(result, "Hello World");
/// ```
///
/// ### Connecting to a signal
///
/// For wrapping closures that can't be sent across threads, the
/// [`closure_local!`](crate::closure_local!) macro can be used. It has the same syntax as
/// `closure!`, but instead uses [`Closure::new_local`] internally.
///
/// ```
/// use glib;
/// use glib::prelude::*;
/// use glib_macros::closure_local;
///
/// let obj = glib::Object::new::<glib::Object>();
/// obj.connect_closure(
///     "notify", false,
///     closure_local!(|_obj: glib::Object, pspec: glib::ParamSpec| {
///         println!("property notify: {}", pspec.name());
///     }));
/// ```
///
/// ### Object Watching
///
/// ```
/// use glib;
/// use glib::prelude::*;
/// use glib_macros::closure_local;
///
/// let closure = {
///     let obj = glib::Object::new::<glib::Object>();
///     let closure = closure_local!(
///         #[watch] obj,
///         move || {
///             obj.type_().name()
///         },
///     );
///     assert_eq!(closure.invoke::<String>(&[]), "GObject");
///     closure
/// };
/// // `obj` is dropped, closure invalidated so it always does nothing and returns None
/// closure.invoke::<()>(&[]);
/// ```
///
/// `#[watch]` has special behavior when connected to a signal:
///
/// ```
/// use glib;
/// use glib::prelude::*;
/// use glib_macros::closure_local;
///
/// let obj = glib::Object::new::<glib::Object>();
/// {
///     let other = glib::Object::new::<glib::Object>();
///     obj.connect_closure(
///         "notify", false,
///         closure_local!(
///             #[watch(rename_to = b)]
///             other,
///             move |a: glib::Object, pspec: glib::ParamSpec| {
///                 let value = a.property_value(pspec.name());
///                 b.set_property(pspec.name(), &value);
///             },
///         ),
///     );
///     // The signal handler will disconnect automatically at the end of this
///     // block when `other` is dropped.
/// }
/// ```
///
/// ### Weak and Strong References
///
/// ```
/// use glib;
/// use glib::prelude::*;
/// use glib_macros::closure;
/// use std::sync::Arc;
///
/// let closure = {
///     let a = Arc::new(String::from("Hello"));
///     let b = Arc::new(String::from("World"));
///     let c = "!";
///     let closure = closure!(
///         #[strong] a,
///         #[weak_allow_none]
///         b,
///         #[to_owned]
///         c,
///         move || {
///             // `a` is Arc<String>, `b` is Option<Arc<String>>, `c` is a `String`
///             format!("{} {}{}", a, b.as_ref().map(|b| b.as_str()).unwrap_or_else(|| "Moon"), c)
///         },
///     );
///     assert_eq!(closure.invoke::<String>(&[]), "Hello World!");
///     closure
/// };
/// // `a`, `c` still kept alive, `b` is dropped
/// assert_eq!(closure.invoke::<String>(&[]), "Hello Moon!");
/// ```
#[proc_macro]
pub fn closure(item: TokenStream) -> TokenStream {
    // Check if this is an old-style closure macro invocation.
    // These always start with an '@' punctuation.
    let Some(first) = item.clone().into_iter().next() else {
        return syn::Error::new(Span::call_site(), "expected a closure")
            .to_compile_error()
            .into();
    };

    match first {
        TokenTree::Punct(ref p) if p.to_string() == "@" => closure_old::closure_inner(item, "new"),
        _ => closure::closure_inner(item, "new"),
    }
}

/// The same as [`closure!`](crate::closure!) but uses [`Closure::new_local`] as a constructor.
/// This is useful for closures which can't be sent across threads. See the documentation of
/// [`closure!`](crate::closure!) for details.
///
/// [`Closure::new_local`]: ../glib/closure/struct.Closure.html#method.new_local
#[proc_macro]
pub fn closure_local(item: TokenStream) -> TokenStream {
    // Check if this is an old-style closure macro invocation.
    // These always start with an '@' punctuation.
    let Some(first) = item.clone().into_iter().next() else {
        return syn::Error::new(Span::call_site(), "expected a closure")
            .to_compile_error()
            .into();
    };

    match first {
        TokenTree::Punct(ref p) if p.to_string() == "@" => {
            closure_old::closure_inner(item, "new_local")
        }
        _ => closure::closure_inner(item, "new_local"),
    }
}

/// Derive macro to register a Rust enum in the GLib type system and derive the
/// [`glib::Value`] traits.
///
/// # Example
///
/// ```
/// use glib::prelude::*;
/// use glib::subclass::prelude::*;
///
/// #[derive(Debug, Copy, Clone, PartialEq, Eq, glib::Enum)]
/// #[enum_type(name = "MyEnum")]
/// enum MyEnum {
///     Val,
///     #[enum_value(name = "My Val")]
///     ValWithCustomName,
///     #[enum_value(name = "My Other Val", nick = "other")]
///     ValWithCustomNameAndNick,
/// }
/// ```
///
/// An enum can be registered as a dynamic type by setting the derive macro
/// helper attribute `enum_dynamic`:
///
/// ```ignore
/// use glib::prelude::*;
/// use glib::subclass::prelude::*;
///
/// #[derive(Debug, Copy, Clone, PartialEq, Eq, glib::Enum)]
/// #[enum_type(name = "MyEnum")]
/// #[enum_dynamic]
/// enum MyEnum {
///     ...
/// }
/// ```
///
/// As a dynamic type, an enum must be explicitly registered when the system
/// loads the implementation (see [`TypePlugin`] and [`TypeModule`]).
/// Therefore, whereas an enum can be registered only once as a static type,
/// it can be registered several times as a dynamic type.
///
/// An enum registered as a dynamic type is never unregistered. The system
/// calls [`TypePluginExt::unuse`] to unload the implementation. If the
/// [`TypePlugin`] subclass is a [`TypeModule`], the enum registered as a
/// dynamic type is marked as unloaded and must be registered again when the
/// module is reloaded.
///
/// The derive macro helper attribute `enum_dynamic` provides two behaviors
/// when registering an enum as a dynamic type:
///
/// - lazy registration: by default an enum is registered as a dynamic type
///   when the system loads the implementation (e.g. when the module is loaded).
///   Optionally setting `lazy_registration` to `true` postpones registration on
///   the first use (when `static_type()` is called for the first time):
///
/// ```ignore
/// #[derive(Debug, Copy, Clone, PartialEq, Eq, glib::Enum)]
/// #[enum_type(name = "MyEnum")]
/// #[enum_dynamic(lazy_registration = true)]
/// enum MyEnum {
///     ...
/// }
/// ```
///
/// - registration within [`TypeModule`] subclass or within [`TypePlugin`]
///   subclass: an enum is usually registered as a dynamic type within a
///   [`TypeModule`] subclass:
///
/// ```ignore
/// #[derive(Debug, Copy, Clone, PartialEq, Eq, glib::Enum)]
/// #[enum_type(name = "MyModuleEnum")]
/// #[enum_dynamic]
/// enum MyModuleEnum {
///     ...
/// }
/// ...
/// #[derive(Default)]
/// pub struct MyModule;
/// ...
/// impl TypeModuleImpl for MyModule {
///     fn load(&self) -> bool {
///         // registers enums as dynamic types.
///         let my_module = self.obj();
///         let type_module: &glib::TypeModule = my_module.upcast_ref();
///         MyModuleEnum::on_implementation_load(type_module)
///     }
///     ...
/// }
/// ```
///
/// Optionally setting `plugin_type` allows to register an enum as a dynamic
/// type within a [`TypePlugin`] subclass that is not a [`TypeModule`]:
///
/// ```ignore
/// #[derive(Debug, Copy, Clone, PartialEq, Eq, glib::Enum)]
/// #[enum_type(name = "MyPluginEnum")]
/// #[enum_dynamic(plugin_type = MyPlugin)]
/// enum MyPluginEnum {
///     ...
/// }
/// ...
/// #[derive(Default)]
/// pub struct MyPlugin;
/// ...
/// impl TypePluginImpl for MyPlugin {
///     fn use_plugin(&self) {
///         // register enums as dynamic types.
///         let my_plugin = self.obj();
///         MyPluginEnum::on_implementation_load(my_plugin.as_ref());
///     }
///     ...
/// }
/// ```
///
/// [`glib::Value`]: ../glib/value/struct.Value.html
/// [`TypePlugin`]: ../glib/gobject/type_plugin/struct.TypePlugin.html
/// [`TypeModule`]: ../glib/gobject/type_module/struct.TypeModule.html
/// [`TypePluginExt::unuse`]: ../glib/gobject/type_plugin/trait.TypePluginExt.
#[proc_macro_derive(Enum, attributes(enum_type, enum_dynamic, enum_value))]
pub fn enum_derive(input: TokenStream) -> TokenStream {
    let input = parse_macro_input!(input as DeriveInput);
    enum_derive::impl_enum(&input)
        .unwrap_or_else(syn::Error::into_compile_error)
        .into()
}

/// Attribute macro for defining flags using the `bitflags` crate.
/// This macro will also define a `GFlags::type_` function and
/// the [`glib::Value`] traits.
///
/// The expected `GType` name has to be passed as macro attribute.
/// The name and nick of each flag can also be optionally defined.
/// Default name is the flag identifier in CamelCase and default nick
/// is the identifier in kebab-case.
/// Combined flags should not be registered with the `GType` system
/// and so need to be tagged with the `#[flags_value(skip)]` attribute.
///
/// # Example
///
/// ```
/// use glib::prelude::*;
/// use glib::subclass::prelude::*;
///
/// #[glib::flags(name = "MyFlags")]
/// enum MyFlags {
///     #[flags_value(name = "Flag A", nick = "nick-a")]
///     A = 0b00000001,
///     #[flags_value(name = "Flag B")]
///     B = 0b00000010,
///     #[flags_value(skip)]
///     AB = Self::A.bits() | Self::B.bits(),
///     C = 0b00000100,
/// }
/// ```
///
/// The flags can be registered as a dynamic type by setting the macro helper
/// attribute `flags_dynamic`:
/// ```ignore
/// use glib::prelude::*;
/// use glib::subclass::prelude::*;
///
/// #[glib::flags(name = "MyFlags")]
/// #[flags_dynamic]
/// enum MyFlags {
///     ...
/// }
/// ```
///
/// As a dynamic type, the flags must be explicitly registered when the system
/// loads the implementation (see [`TypePlugin`] and [`TypeModule`]).
/// Therefore, whereas the flags can be registered only once as a static type,
/// they can be registered several times as a dynamic type.
///
/// The flags registered as a dynamic type are never unregistered. The system
/// calls [`TypePluginExt::unuse`] to unload the implementation. If the
/// [`TypePlugin`] subclass is a [`TypeModule`], the flags registered as a
/// dynamic type are marked as unloaded and must be registered again when the
/// module is reloaded.
///
/// The macro helper attribute `flags_dynamic` provides two behaviors when
/// registering the flags as a dynamic type:
///
/// - lazy registration: by default the flags are registered as a dynamic type
///   when the system loads the implementation (e.g. when the module is loaded).
///   Optionally setting `lazy_registration` to `true` postpones registration on
///   the first use (when `static_type()` is called for the first time):
///
/// ```ignore
/// #[glib::flags(name = "MyFlags")]
/// #[flags_dynamic(lazy_registration = true)]
/// enum MyFlags {
///     ...
/// }
/// ```
///
/// - registration within [`TypeModule`] subclass or within [`TypePlugin`]
///   subclass: the flags are usually registered as a dynamic type within a
///   [`TypeModule`] subclass:
///
/// ```ignore
/// #[glib::flags(name = "MyModuleFlags")]
/// #[flags_dynamic]
/// enum MyModuleFlags {
///     ...
/// }
/// ...
/// #[derive(Default)]
/// pub struct MyModule;
/// ...
/// impl TypeModuleImpl for MyModule {
///     fn load(&self) -> bool {
///         // registers flags as dynamic types.
///         let my_module = self.obj();
///         let type_module: &glib::TypeModule = my_module.upcast_ref();
///         MyModuleFlags::on_implementation_load(type_module)
///     }
///     ...
/// }
/// ```
///
/// Optionally setting `plugin_type` allows to register the flags as a dynamic
/// type within a [`TypePlugin`] subclass that is not a [`TypeModule`]:
/// ```ignore
/// #[glib::flags(name = "MyModuleFlags")]
/// #[flags_dynamic(plugin_type = MyPlugin)]
/// enum MyModuleFlags {
///     ...
/// }
/// ...
/// #[derive(Default)]
/// pub struct MyPlugin;
/// ...
/// impl TypePluginImpl for MyPlugin {
///     fn use_plugin(&self) {
///         // register flags as dynamic types.
///         let my_plugin = self.obj();
///         MyPluginFlags::on_implementation_load(my_plugin.as_ref());
///     }
///     ...
/// }
/// ```
///
/// [`glib::Value`]: ../glib/value/struct.Value.html
/// [`TypePlugin`]: ../glib/gobject/type_plugin/struct.TypePlugin.html
/// [`TypeModule`]: ../glib/gobject/type_module/struct.TypeModule.html
/// [`TypePluginExt::unuse`]: ../glib/gobject/type_plugin/trait.TypePluginExt.
#[proc_macro_attribute]
pub fn flags(attr: TokenStream, item: TokenStream) -> TokenStream {
    let mut name = NestedMetaItem::<syn::LitStr>::new("name")
        .required()
        .value_required();
    let mut allow_name_conflict_attr =
        NestedMetaItem::<syn::LitBool>::new("allow_name_conflict").value_optional();

    if let Err(e) = parse_nested_meta_items_from_stream(
        attr.into(),
        &mut [&mut name, &mut allow_name_conflict_attr],
    ) {
        return e.to_compile_error().into();
    }

    let allow_name_conflict = allow_name_conflict_attr.found
        || allow_name_conflict_attr
            .value
            .map(|b| b.value())
            .unwrap_or(false);

    let attr_meta = AttrInput {
        enum_name: name.value.unwrap(),
        allow_name_conflict,
    };

    syn::parse::<syn::ItemEnum>(item)
        .map_err(|_| syn::Error::new(Span::call_site(), flags_attribute::WRONG_PLACE_MSG))
        .map(|mut input| flags_attribute::impl_flags(attr_meta, &mut input))
        .unwrap_or_else(syn::Error::into_compile_error)
        .into()
}

/// Derive macro for defining a GLib error domain and its associated
/// [`ErrorDomain`] trait.
///
/// # Example
///
/// ```
/// use glib::prelude::*;
/// use glib::subclass::prelude::*;
///
/// #[derive(Debug, Copy, Clone, glib::ErrorDomain)]
/// #[error_domain(name = "ex-foo")]
/// enum Foo {
///     Blah,
///     Baaz,
/// }
/// ```
///
/// [`ErrorDomain`]: ../glib/error/trait.ErrorDomain.html
#[proc_macro_derive(ErrorDomain, attributes(error_domain))]
pub fn error_domain_derive(input: TokenStream) -> TokenStream {
    let input = parse_macro_input!(input as DeriveInput);
    error_domain_derive::impl_error_domain(&input)
        .unwrap_or_else(syn::Error::into_compile_error)
        .into()
}

/// Derive macro for defining a [`BoxedType`]`::type_` function and
/// the [`glib::Value`] traits. Optionally, the type can be marked as
/// `nullable` to get an implementation of `glib::value::ToValueOptional`.
///
/// # Example
///
/// ```
/// use glib::prelude::*;
/// use glib::subclass::prelude::*;
///
/// #[derive(Clone, Debug, PartialEq, Eq, glib::Boxed)]
/// #[boxed_type(name = "MyBoxed")]
/// struct MyBoxed(String);
///
/// #[derive(Clone, Debug, PartialEq, Eq, glib::Boxed)]
/// #[boxed_type(name = "MyNullableBoxed", nullable)]
/// struct MyNullableBoxed(String);
/// ```
///
/// [`BoxedType`]: ../glib/subclass/boxed/trait.BoxedType.html
/// [`glib::Value`]: ../glib/value/struct.Value.html
#[proc_macro_derive(Boxed, attributes(boxed_type))]
pub fn boxed_derive(input: TokenStream) -> TokenStream {
    let input = parse_macro_input!(input as DeriveInput);
    boxed_derive::impl_boxed(&input)
        .unwrap_or_else(syn::Error::into_compile_error)
        .into()
}

/// Derive macro for defining a [`SharedType`]`::get_type` function and
/// the [`glib::Value`] traits. Optionally, the type can be marked as
/// `nullable` to get an implementation of `glib::value::ToValueOptional`.
///
/// # Example
///
/// ```
/// use glib::prelude::*;
/// use glib::subclass::prelude::*;
///
/// #[derive(Clone, Debug, PartialEq, Eq)]
/// struct MySharedInner {
///   foo: String,
/// }
///
/// #[derive(Clone, Debug, PartialEq, Eq, glib::SharedBoxed)]
/// #[shared_boxed_type(name = "MySharedBoxed")]
/// struct MySharedBoxed(std::sync::Arc<MySharedInner>);
///
/// #[derive(Clone, Debug, PartialEq, Eq, glib::SharedBoxed)]
/// #[shared_boxed_type(name = "MyNullableSharedBoxed", nullable)]
/// struct MyNullableSharedBoxed(std::sync::Arc<MySharedInner>);
/// ```
///
/// [`SharedType`]: ../glib/subclass/shared/trait.SharedType.html
/// [`glib::Value`]: ../glib/value/struct.Value.html
#[proc_macro_derive(SharedBoxed, attributes(shared_boxed_type))]
pub fn shared_boxed_derive(input: TokenStream) -> TokenStream {
    let input = parse_macro_input!(input as DeriveInput);
    shared_boxed_derive::impl_shared_boxed(&input)
        .unwrap_or_else(syn::Error::into_compile_error)
        .into()
}

/// Macro for boilerplate of [`ObjectSubclass`] implementations.
///
/// This adds implementations for the `type_data()` and `type_()` methods,
/// which should probably never be defined differently.
///
/// It provides default values for the `Instance`, `Class`, and `Interfaces`
/// type parameters. If these are present, the macro will use the provided value
/// instead of the default.
///
/// Usually the defaults for `Instance` and `Class` will work. `Interfaces` is
/// necessary for types that implement interfaces.
///
/// ```ignore
/// type Instance = glib::subclass::basic::InstanceStruct<Self>;
/// type Class = glib::subclass::basic::ClassStruct<Self>;
/// type Interfaces = ();
/// ```
///
/// If no `new()` or `with_class()` method is provided, the macro adds a `new()`
/// implementation calling `Default::default()`. So the type needs to implement
/// `Default`, or this should be overridden.
///
/// ```ignore
/// fn new() -> Self {
///     Default::default()
/// }
/// ```
///
/// An object subclass can be registered as a dynamic type by setting the macro
/// helper attribute `object_class_dynamic`:
///
/// ```ignore
/// #[derive(Default)]
/// pub struct MyType;
///
/// #[glib::object_subclass]
/// #[object_subclass_dynamic]
/// impl ObjectSubclass for MyType { ... }
/// ```
///
/// As a dynamic type, an object subclass must be explicitly registered when
/// the system loads the implementation (see [`TypePlugin`] and [`TypeModule`]).
/// Therefore, whereas an object subclass can be registered only once as a
/// static type, it can be registered several times as a dynamic type.
///
/// An object subclass registered as a dynamic type is never unregistered. The
/// system calls [`TypePluginExt::unuse`] to unload the implementation. If the
/// [`TypePlugin`] subclass is a [`TypeModule`], the object subclass registered
/// as a dynamic type is marked as unloaded and must be registered again when
/// the module is reloaded.
///
/// The macro helper attribute `object_class_dynamic` provides two behaviors
/// when registering an object subclass as a dynamic type:
///
/// - lazy registration: by default an object subclass is registered as a
///   dynamic type when the system loads the implementation (e.g. when the module
///   is loaded). Optionally setting `lazy_registration` to `true` postpones
///   registration on the first use (when `static_type()` is called for the first
///   time):
///
/// ```ignore
/// #[derive(Default)]
/// pub struct MyType;
///
/// #[glib::object_subclass]
/// #[object_subclass_dynamic(lazy_registration = true)]
/// impl ObjectSubclass for MyType { ... }
/// ```
///
/// - registration within [`TypeModule`] subclass or within [`TypePlugin`]
///   subclass: an object subclass is usually registered as a dynamic type within
///   a [`TypeModule`] subclass:
///
/// ```ignore
/// #[derive(Default)]
/// pub struct MyModuleType;
///
/// #[glib::object_subclass]
/// #[object_subclass_dynamic]
/// impl ObjectSubclass for MyModuleType { ... }
/// ...
/// #[derive(Default)]
/// pub struct MyModule;
/// ...
/// impl TypeModuleImpl for MyModule {
///     fn load(&self) -> bool {
///         // registers object subclasses as dynamic types.
///         let my_module = self.obj();
///         let type_module: &glib::TypeModule = my_module.upcast_ref();
///         MyModuleType::on_implementation_load(type_module)
///     }
///     ...
/// }
/// ```
///
/// Optionally setting `plugin_type` allows to register an object subclass as a
/// dynamic type within a [`TypePlugin`] subclass that is not a [`TypeModule`]:
///
/// ```ignore
/// #[derive(Default)]
/// pub struct MyPluginType;
///
/// #[glib::object_subclass]
/// #[object_subclass_dynamic(plugin_type = MyPlugin)]
/// impl ObjectSubclass for MyPluginType { ... }
/// ...
/// #[derive(Default)]
/// pub struct MyPlugin;
/// ...
/// impl TypePluginImpl for MyPlugin {
///     fn use_plugin(&self) {
///         // register object subclasses as dynamic types.
///         let my_plugin = self.obj();
///         MyPluginType::on_implementation_load(my_plugin.as_ref());
///     }
///     ...
/// }
/// ```
///
/// [`ObjectSubclass`]: ../glib/subclass/types/trait.ObjectSubclass.html
/// [`TypePlugin`]: ../glib/gobject/type_plugin/struct.TypePlugin.html
/// [`TypeModule`]: ../glib/gobject/type_module/struct.TypeModule.html
/// [`TypePluginExt::unuse`]: ../glib/gobject/type_plugin/trait.TypePluginExt.html#method.unuse
#[proc_macro_attribute]
pub fn object_subclass(_attr: TokenStream, item: TokenStream) -> TokenStream {
    let input = parse_macro_input!(item with object_impl_attributes::Input::parse_subclass);
    object_impl_attributes::subclass::impl_object_subclass(input).into()
}

/// Macro for boilerplate of [`ObjectInterface`] implementations.
///
/// This adds implementations for the `get_type()` method, which should probably never be defined
/// differently.
///
/// It provides default values for the `Prerequisites` type parameter. If this is present, the macro
/// will use the provided value instead of the default.
///
/// `Prerequisites` are interfaces for types that require a specific base class or interfaces.
///
/// ```ignore
/// type Prerequisites = ();
/// ```
///
/// An object interface can be registered as a dynamic type by setting the
/// macro helper attribute `object_interface_dynamic`:
/// ```ignore
/// pub struct MyInterface {
///     parent: glib::gobject_ffi::GTypeInterface,
/// }
/// #[glib::object_interface]
/// #[object_interface_dynamic]
/// unsafe impl ObjectInterface for MyInterface { ... }
/// ```
///
/// As a dynamic type, an object interface must be explicitly registered when
/// the system loads the implementation (see [`TypePlugin`] and [`TypeModule`]).
/// Therefore, whereas an object interface can be registered only once as a
/// static type, it can be registered several times as a dynamic type.
///
/// An object interface registered as a dynamic type is never unregistered. The
/// system calls [`TypePluginExt::unuse`] to unload the implementation. If the
/// [`TypePlugin`] subclass is a [`TypeModule`], the object interface
/// registered as a dynamic type is marked as unloaded and must be registered
/// again when the module is reloaded.
///
/// The macro helper attribute `object_interface_dynamic` provides two
/// behaviors when registering an object interface as a dynamic type:
///
/// - lazy registration: by default an object interface is registered as a
///   dynamic type when the system loads the implementation (e.g. when the module
///   is loaded). Optionally setting `lazy_registration` to `true` postpones
///   registration on the first use (when `type_()` is called for the first time):
///
/// ```ignore
/// pub struct MyInterface {
///     parent: glib::gobject_ffi::GTypeInterface,
/// }
/// #[glib::object_interface]
/// #[object_interface_dynamic(lazy_registration = true)]
/// unsafe impl ObjectInterface for MyInterface { ... }
/// ```
///
/// - registration within [`TypeModule`] subclass or within [`TypePlugin`]
///   subclass: an object interface is usually registered as a dynamic type
///   within a [`TypeModule`] subclass:
///
/// ```ignore
/// pub struct MyModuleInterface {
///     parent: glib::gobject_ffi::GTypeInterface,
/// }
/// #[glib::object_interface]
/// #[object_interface_dynamic]
/// unsafe impl ObjectInterface for MyModuleInterface { ... }
/// ...
/// #[derive(Default)]
/// pub struct MyModule;
/// ...
/// impl TypeModuleImpl for MyModule {
///     fn load(&self) -> bool {
///         // registers object interfaces as dynamic types.
///         let my_module = self.obj();
///         let type_module: &glib::TypeModule = my_module.upcast_ref();
///         MyModuleInterface::on_implementation_load(type_module)
///     }
///     ...
/// }
/// ```
///
/// Optionally setting `plugin_type` allows to register an object interface as
/// a dynamic type within a [`TypePlugin`] subclass that is not a [`TypeModule`]:
///
/// ```ignore
/// pub struct MyPluginInterface {
///     parent: glib::gobject_ffi::GTypeInterface,
/// }
/// #[glib::object_interface]
/// #[object_interface_dynamic(plugin_type = MyPlugin)]
/// unsafe impl ObjectInterface for MyPluginInterface { ... }
/// ...
/// #[derive(Default)]
/// pub struct MyPlugin;
/// ...
/// impl TypePluginImpl for MyPlugin {
///     fn use_plugin(&self) {
///         // register object interfaces as dynamic types.
///         let my_plugin = self.obj();
///         MyPluginInterface::on_implementation_load(my_plugin.as_ref());
///     }
///     ...
/// }
/// ```
///
/// [`ObjectInterface`]: ../glib/subclass/interface/trait.ObjectInterface.html
/// [`TypePlugin`]: ../glib/gobject/type_plugin/struct.TypePlugin.html
/// [`TypeModule`]: ../glib/gobject/type_module/struct.TypeModule.html
/// [`TypePluginExt::unuse`]: ../glib/gobject/type_plugin/trait.TypePluginExt.html#method.unuse///
#[proc_macro_attribute]
pub fn object_interface(_attr: TokenStream, item: TokenStream) -> TokenStream {
    let input = parse_macro_input!(item with object_impl_attributes::Input::parse_interface);
    object_impl_attributes::interface::impl_object_interface(input).into()
}

/// Macro for deriving implementations of [`glib::clone::Downgrade`] and
/// [`glib::clone::Upgrade`] traits and a weak type.
///
/// # Examples
///
/// ## New Type Idiom
///
/// ```rust,ignore
/// #[derive(glib::Downgrade)]
/// pub struct FancyLabel(gtk::Label);
///
/// impl FancyLabel {
///     pub fn new(label: &str) -> Self {
///         Self(gtk::LabelBuilder::new().label(label).build())
///     }
///
///     pub fn flip(&self) {
///         self.0.set_angle(180.0 - self.0.angle());
///     }
/// }
///
/// let fancy_label = FancyLabel::new("Look at me!");
/// let button = gtk::ButtonBuilder::new().label("Click me!").build();
/// button.connect_clicked(
///     clone!(
///         #[weak]
///         fancy_label,
///         move || fancy_label.flip(),
///     ),
/// );
/// ```
///
/// ## Generic New Type
///
/// ```rust,ignore
/// #[derive(glib::Downgrade)]
/// pub struct TypedEntry<T>(gtk::Entry, std::marker::PhantomData<T>);
///
/// impl<T: ToString + FromStr> for TypedEntry<T> {
///     // ...
/// }
/// ```
///
/// ## Structures and Enums
///
/// ```rust,ignore
/// #[derive(Clone, glib::Downgrade)]
/// pub struct ControlButtons {
///     pub up: gtk::Button,
///     pub down: gtk::Button,
///     pub left: gtk::Button,
///     pub right: gtk::Button,
/// }
///
/// #[derive(Clone, glib::Downgrade)]
/// pub enum DirectionButton {
///     Left(gtk::Button),
///     Right(gtk::Button),
///     Up(gtk::Button),
///     Down(gtk::Button),
/// }
/// ```
///
/// [`glib::clone::Downgrade`]: ../glib/clone/trait.Downgrade.html
/// [`glib::clone::Upgrade`]: ../glib/clone/trait.Upgrade.html
#[proc_macro_derive(Downgrade)]
pub fn downgrade(input: TokenStream) -> TokenStream {
    let input = parse_macro_input!(input as DeriveInput);
    downgrade_derive::impl_downgrade(input)
}

/// Derive macro for serializing/deserializing custom structs/enums as [`glib::Variant`]s.
///
/// # Example
///
/// ```
/// use glib::prelude::*;
///
/// #[derive(Debug, PartialEq, Eq, glib::Variant)]
/// struct Foo {
///     some_string: String,
///     some_int: i32,
/// }
///
/// let v = Foo { some_string: String::from("bar"), some_int: 1 };
/// let var = v.to_variant();
/// assert_eq!(var.get::<Foo>(), Some(v));
/// ```
///
/// When storing `Vec`s of fixed size types it is a good idea to wrap these in
/// `glib::FixedSizeVariantArray` as serialization/deserialization will be more efficient.
///
/// # Example
///
/// ```
/// use glib::prelude::*;
///
/// #[derive(Debug, PartialEq, Eq, glib::Variant)]
/// struct Foo {
///     some_vec: glib::FixedSizeVariantArray<Vec<u32>, u32>,
///     some_int: i32,
/// }
///
/// let v = Foo { some_vec: vec![1u32, 2u32].into(), some_int: 1 };
/// let var = v.to_variant();
/// assert_eq!(var.get::<Foo>(), Some(v));
/// ```
///
/// Enums are serialized as a tuple `(sv)` with the first value as a [kebab case] string for the
/// enum variant, or just `s` if this is a C-style enum. Some additional attributes are supported
/// for enums:
/// - `#[variant_enum(repr)]` to serialize the enum variant as an integer type instead of `s`.  The
///   `#[repr]` attribute must also be specified on the enum with a sized integer type, and the type
///   must implement `Copy`.
/// - `#[variant_enum(enum)]` uses [`EnumClass`] to serialize/deserialize as nicks. Meant for use
///   with [`glib::Enum`](Enum).
/// - `#[variant_enum(flags)]` uses [`FlagsClass`] to serialize/deserialize as nicks. Meant for use
///   with [`glib::flags`](macro@flags).
/// - `#[variant_enum(enum, repr)]` serializes as `i32`. Meant for use with [`glib::Enum`](Enum).
///   The type must also implement `Copy`.
/// - `#[variant_enum(flags, repr)]` serializes as `u32`. Meant for use with
///   [`glib::flags`](macro@flags).
///
/// # Example
///
/// ```
/// use glib::prelude::*;
///
/// #[derive(Debug, PartialEq, Eq, glib::Variant)]
/// enum Foo {
///     MyA,
///     MyB(i32),
///     MyC { some_int: u32, some_string: String }
/// }
///
/// let v = Foo::MyC { some_int: 1, some_string: String::from("bar") };
/// let var = v.to_variant();
/// assert_eq!(var.child_value(0).str(), Some("my-c"));
/// assert_eq!(var.get::<Foo>(), Some(v));
///
/// #[derive(Debug, Copy, Clone, PartialEq, Eq, glib::Variant)]
/// #[variant_enum(repr)]
/// #[repr(u8)]
/// enum Bar {
///     A,
///     B = 3,
///     C = 7
/// }
///
/// let v = Bar::B;
/// let var = v.to_variant();
/// assert_eq!(var.get::<u8>(), Some(3));
/// assert_eq!(var.get::<Bar>(), Some(v));
///
/// #[derive(Debug, Copy, Clone, PartialEq, Eq, glib::Enum, glib::Variant)]
/// #[variant_enum(enum)]
/// #[enum_type(name = "MyEnum")]
/// enum MyEnum {
///     Val,
///     #[enum_value(name = "My Val")]
///     ValWithCustomName,
///     #[enum_value(name = "My Other Val", nick = "other")]
///     ValWithCustomNameAndNick,
/// }
///
/// let v = MyEnum::ValWithCustomNameAndNick;
/// let var = v.to_variant();
/// assert_eq!(var.str(), Some("other"));
/// assert_eq!(var.get::<MyEnum>(), Some(v));
/// ```
///
/// [`glib::Variant`]: ../glib/variant/struct.Variant.html
/// [`EnumClass`]: ../glib/struct.EnumClass.html
/// [`FlagsClass`]: ../glib/struct.FlagsClass.html
/// [kebab case]: https://docs.rs/heck/0.4.0/heck/trait.ToKebabCase.html
#[proc_macro_derive(Variant, attributes(variant_enum))]
pub fn variant_derive(input: TokenStream) -> TokenStream {
    let input = parse_macro_input!(input as DeriveInput);
    variant_derive::impl_variant(input)
        .unwrap_or_else(syn::Error::into_compile_error)
        .into()
}
#[proc_macro]
pub fn cstr_bytes(item: TokenStream) -> TokenStream {
    syn::parse::Parser::parse2(
        |stream: syn::parse::ParseStream<'_>| {
            let literal = stream.parse::<syn::LitStr>()?;
            stream.parse::<syn::parse::Nothing>()?;
            let bytes = std::ffi::CString::new(literal.value())
                .map_err(|e| syn::Error::new_spanned(&literal, format!("{e}")))?
                .into_bytes_with_nul();
            let bytes = proc_macro2::Literal::byte_string(&bytes);
            Ok(quote::quote! { #bytes }.into())
        },
        item.into(),
    )
    .unwrap_or_else(|e| e.into_compile_error().into())
}

/// This macro enables you to derive object properties in a quick way.
///
/// # Supported `#[property]` attributes
/// | Attribute | Description | Default | Example |
/// | --- | --- | --- | --- |
/// | `name = "literal"` | The name of the property | field ident where `_` (leading and trailing `_` are trimmed) is replaced into `-` | `#[property(name = "prop-name")]` |
/// | `type = expr` | The type of the property | inferred | `#[property(type = i32)]` |
/// | `get [= expr]` | Specify that the property is readable and use [`PropertyGet::get`] [or optionally set a custom internal getter] | | `#[property(get)]`, `#[property(get = get_prop)]`, or `[property(get = \|_\| 2)]` |
/// | `set [= expr]` | Specify that the property is writable and use [`PropertySet::set`] [or optionally set a custom internal setter] | | `#[property(set)]`, `#[property(set = set_prop)]`, or `[property(set = \|_, val\| {})]` |
/// | `override_class = expr` | The type of class of which to override the property from | | `#[property(override_class = SomeClass)]` |
/// | `override_interface = expr` | The type of interface of which to override the property from | | `#[property(override_interface = SomeInterface)]` |
/// | `nullable` | Whether to use `Option<T>` in the generated setter method |  | `#[property(nullable)]` |
/// | `member = ident` | Field of the nested type where property is retrieved and set | | `#[property(member = author)]` |
/// | `construct` | Specify that the property is construct property. Ensures that the property is always set during construction (if not explicitly then the default value is used). The use of a custom internal setter is supported. | | `#[property(get, construct)]` or `#[property(get, set = set_prop, construct)]` |
/// | `construct_only` | Specify that the property is construct only. This will not generate a public setter and only allow the property to be set during object construction. The use of a custom internal setter is supported. | | `#[property(get, construct_only)]` or `#[property(get, set = set_prop, construct_only)]` |
/// | `builder(<required-params>)[.ident]*` | Used to input required params or add optional Param Spec builder fields | | `#[property(builder(SomeEnum::default()))]`, `#[builder().default_value(1).minimum(0).maximum(5)]`, etc.  |
/// | `default` | Sets the `default_value` field of the Param Spec builder | | `#[property(default = 1)]` |
/// | `<optional-pspec-builder-fields> = expr` | Used to add optional Param Spec builder fields | | `#[property(minimum = 0)` , `#[property(minimum = 0, maximum = 1)]`, etc. |
/// | `<optional-pspec-builder-fields>` | Used to add optional Param Spec builder fields | | `#[property(explicit_notify)]` , `#[property(construct_only)]`, etc. |
///
/// ## Using Rust keywords as property names
/// You might hit a roadblock when declaring properties with this macro because you want to use a name that happens to be a Rust keyword. This may happen with names like `loop`, which is a pretty common name when creating things like animation handlers.
/// To use those names, you can make use of the raw identifier feature of Rust. Simply prefix the identifier name with `r#` in the struct declaration. Internally, those `r#`s are stripped so you can use its expected name in [`ObjectExt::property`] or within GtkBuilder template files.
///
/// # Generated methods
/// The following methods are generated on the wrapper type specified on `#[properties(wrapper_type = ...)]`:
/// * `$property()`, when the property is readable
/// * `set_$property()`, when the property is writable and not construct-only
/// * `connect_$property_notify()`
/// * `notify_$property()`
///
/// ## Extension trait
/// You can choose to move the method definitions to a trait by using `#[properties(wrapper_type = super::MyType, ext_trait = MyTypePropertiesExt)]`.
/// The trait name is optional, and defaults to `MyTypePropertiesExt`, where `MyType` is extracted from the wrapper type.
/// Note: The trait is defined in the same module where the `#[derive(Properties)]` call happens, and is implemented on the wrapper type.
///
/// Notice: You can't reimplement the generated methods on the wrapper type, unless you move them to a trait.
/// You can change the behavior of the generated getter/setter methods by using a custom internal getter/setter.
///
/// # Internal getters and setters
/// By default, they are generated for you. However, you can use a custom getter/setter
/// by assigning an expression to `get`/`set` `#[property]` attributes: `#[property(get = |_| 2, set)]` or `#[property(get, set = custom_setter_func)]`.
///
/// # Supported types
/// Every type implementing the trait [`Property`] is supported.
/// The type `Option<T>` is supported as a property only if `Option<T>` implements [`ToValueOptional`].
/// Optional types also require the `nullable` attribute: without it, the generated setter on the wrapper type
/// will take `T` instead of `Option<T>`, preventing the user from ever calling the setter with a `None` value.
///
/// ## Adding support for custom types
/// ### Types wrapping an existing <code>T: [ToValue] + [HasParamSpec]</code>
/// If you have declared a newtype as
/// ```rust
/// struct MyInt(i32);
/// ```
/// you can use it as a property by deriving [`ValueDelegate`].
///
/// ### Types with inner mutability
/// The trait [`Property`] must be implemented.
/// The traits [`PropertyGet`] and [`PropertySet`] should be implemented to enable the Properties macro
/// to generate a default internal getter/setter.
/// If possible, implementing [`PropertySetNested`] is preferred over `PropertySet`, because it
/// enables this macro to access the contained type and provide access to its fields,
/// using the `member = $structfield` syntax.
///
/// ### Types without [`HasParamSpec`][HasParamSpec]
/// If you have encountered a type <code>T: [ToValue]</code>, inside the gtk-rs crate, which doesn't implement [`HasParamSpec`][HasParamSpec],
/// then it's a bug and you should report it.
/// If you need to support a `ToValue` type with a [`ParamSpec`] not provided by gtk-rs, then you need to
/// implement `HasParamSpec` on that type.
///
/// # Example
/// ```
/// use std::cell::RefCell;
/// use glib::prelude::*;
/// use glib::subclass::prelude::*;
/// use glib_macros::Properties;
///
/// #[derive(Default, Clone)]
/// struct Author {
///     name: String,
///     nick: String,
/// }
///
/// pub mod imp {
///     use std::rc::Rc;
///
///     use super::*;
///
///     #[derive(Properties, Default)]
///     #[properties(wrapper_type = super::Foo)]
///     pub struct Foo {
///         #[property(get, set = Self::set_fizz)]
///         fizz: RefCell<String>,
///         #[property(name = "author-name", get, set, type = String, member = name)]
///         #[property(name = "author-nick", get, set, type = String, member = nick)]
///         author: RefCell<Author>,
///         #[property(get, set, explicit_notify, lax_validation)]
///         custom_flags: RefCell<String>,
///         #[property(get, set, minimum = 0, maximum = 3)]
///         numeric_builder: RefCell<u32>,
///         #[property(get, set, builder('c'))]
///         builder_with_required_param: RefCell<char>,
///         #[property(get, set, nullable)]
///         optional: RefCell<Option<String>>,
///         #[property(get, set)]
///         smart_pointer: Rc<RefCell<String>>,
///     }
///     
///     #[glib::derived_properties]
///     impl ObjectImpl for Foo {}
///
///     #[glib::object_subclass]
///     impl ObjectSubclass for Foo {
///         const NAME: &'static str = "MyFoo";
///         type Type = super::Foo;
///     }
///
///     impl Foo {
///         fn set_fizz(&self, value: String) {
///             *self.fizz.borrow_mut() = format!("custom set: {}", value);
///         }
///     }
/// }
///
/// glib::wrapper! {
///     pub struct Foo(ObjectSubclass<imp::Foo>);
/// }
///
/// fn main() {
///   let myfoo: Foo = glib::object::Object::new();
///
///   myfoo.set_fizz("test value");
///   assert_eq!(myfoo.fizz(), "custom set: test value".to_string());
/// }
/// ```
///
/// [`Property`]: ../glib/property/trait.Property.html
/// [`PropertyGet`]: ../glib/property/trait.PropertyGet.html
/// [`PropertyGet::get`]: ../glib/property/trait.PropertyGet.html#tymethod.get
/// [`PropertySet`]: ../glib/property/trait.PropertySet.html
/// [`PropertySet::set`]: ../glib/property/trait.PropertySet.html#tymethod.set
/// [`PropertySetNested`]: ../glib/property/trait.PropertySetNested.html
/// [`ObjectExt::property`]: ../glib/object/trait.ObjectExt.html#tymethod.property
/// [HasParamSpec]: ../glib/trait.HasParamSpec.html
/// [`ParamSpec`]: ../glib/struct.ParamSpec.html
/// [`ToValueOptional`]: ../glib/value/trait.ToValueOptional.html
/// [ToValue]: ../glib/value/trait.ToValue.html
#[allow(clippy::needless_doctest_main)]
#[proc_macro_derive(Properties, attributes(properties, property))]
pub fn derive_props(input: TokenStream) -> TokenStream {
    let input = parse_macro_input!(input as properties::PropsMacroInput);
    properties::impl_derive_props(input)
}

/// When applied to `ObjectImpl`
/// ```ignore
/// #[glib::derived_properties]
/// impl ObjectImpl for CustomObject
/// ```
/// this macro generates
/// ```ignore
/// impl ObjectImpl for CustomObject {
///     fn properties() -> &'static [glib::ParamSpec] {
///         Self::derived_properties()
///     }
///     fn set_property(&self, id: usize, value: &glib::Value, pspec: &glib::ParamSpec) {
///         self.derived_set_property(id, value, pspec)
///     }
///     fn property(&self, id: usize, pspec: &glib::ParamSpec) -> glib::Value {
///         self.derived_property(id, pspec)
///     }
/// }
/// ```
#[proc_macro_attribute]
pub fn derived_properties(_attr: TokenStream, item: TokenStream) -> TokenStream {
    syn::parse::<syn::ItemImpl>(item)
        .map_err(|_| {
            syn::Error::new(
                Span::call_site(),
                derived_properties_attribute::WRONG_PLACE_MSG,
            )
        })
        .and_then(|input| derived_properties_attribute::impl_derived_properties(&input))
        .unwrap_or_else(syn::Error::into_compile_error)
        .into()
}

/// # Example
/// ```
/// use glib::prelude::*;
/// use glib::ValueDelegate;
///
/// #[derive(ValueDelegate, Debug, PartialEq)]
/// struct MyInt(i32);
///
/// let myv = MyInt(2);
/// let convertedv = myv.to_value();
/// assert_eq!(convertedv.get::<MyInt>(), Ok(myv));
///
///
/// #[derive(ValueDelegate, Debug, PartialEq)]
/// #[value_delegate(from = u32)]
/// enum MyEnum {
///     Zero,
///     NotZero(u32)
/// }
///
/// impl From<u32> for MyEnum {
///     fn from(v: u32) -> Self {
///         match v {
///             0 => MyEnum::Zero,
///             x => MyEnum::NotZero(x)
///         }
///     }
/// }
/// impl<'a> From<&'a MyEnum> for u32 {
///     fn from(v: &'a MyEnum) -> Self {
///         match v {
///             MyEnum::Zero => 0,
///             MyEnum::NotZero(x) => *x
///         }
///     }
/// }
/// impl From<MyEnum> for u32 {
///     fn from(v: MyEnum) -> Self {
///         match v {
///             MyEnum::Zero => 0,
///             MyEnum::NotZero(x) => x
///         }
///     }
/// }
///
/// let myv = MyEnum::NotZero(34);
/// let convertedv = myv.to_value();
/// assert_eq!(convertedv.get::<MyEnum>(), Ok(myv));
///
///
/// // If you want your type to be usable inside an `Option`, you can derive `ToValueOptional`
/// // by adding `nullable` as follows
/// #[derive(ValueDelegate, Debug, PartialEq)]
/// #[value_delegate(nullable)]
/// struct MyString(String);
///
/// let myv = Some(MyString("Hello world".to_string()));
/// let convertedv = myv.to_value();
/// assert_eq!(convertedv.get::<Option<MyString>>(), Ok(myv));
/// let convertedv = None::<MyString>.to_value();
/// assert_eq!(convertedv.get::<Option<MyString>>(), Ok(None::<MyString>));
/// ```
#[proc_macro_derive(ValueDelegate, attributes(value_delegate))]
pub fn derive_value_delegate(input: TokenStream) -> TokenStream {
    let input = parse_macro_input!(input as value_delegate_derive::ValueDelegateInput);
    value_delegate_derive::impl_value_delegate(input).unwrap()
}

/// An attribute macro for writing asynchronous test functions.
///
/// This macro is designed to wrap an asynchronous test function and ensure that
/// it runs within a `glib::MainContext`. It helps in writing async tests that
/// require the use of an event loop for the asynchronous execution.
///
/// # Example
///
/// ```
/// #[glib::async_test]
/// async fn my_async_test() {
///     // Test code that runs asynchronously
/// }
/// ```
#[proc_macro_attribute]
pub fn async_test(args: TokenStream, item: TokenStream) -> TokenStream {
    async_test::async_test(args, item)
}