diff options
Diffstat (limited to 'vendor/quote/src/lib.rs')
-rw-r--r-- | vendor/quote/src/lib.rs | 1444 |
1 files changed, 1444 insertions, 0 deletions
diff --git a/vendor/quote/src/lib.rs b/vendor/quote/src/lib.rs new file mode 100644 index 0000000..8b97abd --- /dev/null +++ b/vendor/quote/src/lib.rs @@ -0,0 +1,1444 @@ +//! [![github]](https://github.com/dtolnay/quote) [![crates-io]](https://crates.io/crates/quote) [![docs-rs]](https://docs.rs/quote) +//! +//! [github]: https://img.shields.io/badge/github-8da0cb?style=for-the-badge&labelColor=555555&logo=github +//! [crates-io]: https://img.shields.io/badge/crates.io-fc8d62?style=for-the-badge&labelColor=555555&logo=rust +//! [docs-rs]: https://img.shields.io/badge/docs.rs-66c2a5?style=for-the-badge&labelColor=555555&logo=docs.rs +//! +//! <br> +//! +//! This crate provides the [`quote!`] macro for turning Rust syntax tree data +//! structures into tokens of source code. +//! +//! [`quote!`]: macro.quote.html +//! +//! Procedural macros in Rust receive a stream of tokens as input, execute +//! arbitrary Rust code to determine how to manipulate those tokens, and produce +//! a stream of tokens to hand back to the compiler to compile into the caller's +//! crate. Quasi-quoting is a solution to one piece of that — producing +//! tokens to return to the compiler. +//! +//! The idea of quasi-quoting is that we write *code* that we treat as *data*. +//! Within the `quote!` macro, we can write what looks like code to our text +//! editor or IDE. We get all the benefits of the editor's brace matching, +//! syntax highlighting, indentation, and maybe autocompletion. But rather than +//! compiling that as code into the current crate, we can treat it as data, pass +//! it around, mutate it, and eventually hand it back to the compiler as tokens +//! to compile into the macro caller's crate. +//! +//! This crate is motivated by the procedural macro use case, but is a +//! general-purpose Rust quasi-quoting library and is not specific to procedural +//! macros. +//! +//! ```toml +//! [dependencies] +//! quote = "1.0" +//! ``` +//! +//! <br> +//! +//! # Example +//! +//! The following quasi-quoted block of code is something you might find in [a] +//! procedural macro having to do with data structure serialization. The `#var` +//! syntax performs interpolation of runtime variables into the quoted tokens. +//! Check out the documentation of the [`quote!`] macro for more detail about +//! the syntax. See also the [`quote_spanned!`] macro which is important for +//! implementing hygienic procedural macros. +//! +//! [a]: https://serde.rs/ +//! [`quote_spanned!`]: macro.quote_spanned.html +//! +//! ``` +//! # use quote::quote; +//! # +//! # let generics = ""; +//! # let where_clause = ""; +//! # let field_ty = ""; +//! # let item_ty = ""; +//! # let path = ""; +//! # let value = ""; +//! # +//! let tokens = quote! { +//! struct SerializeWith #generics #where_clause { +//! value: &'a #field_ty, +//! phantom: core::marker::PhantomData<#item_ty>, +//! } +//! +//! impl #generics serde::Serialize for SerializeWith #generics #where_clause { +//! fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> +//! where +//! S: serde::Serializer, +//! { +//! #path(self.value, serializer) +//! } +//! } +//! +//! SerializeWith { +//! value: #value, +//! phantom: core::marker::PhantomData::<#item_ty>, +//! } +//! }; +//! ``` +//! +//! <br> +//! +//! # Non-macro code generators +//! +//! When using `quote` in a build.rs or main.rs and writing the output out to a +//! file, consider having the code generator pass the tokens through +//! [prettyplease] before writing. This way if an error occurs in the generated +//! code it is convenient for a human to read and debug. +//! +//! [prettyplease]: https://github.com/dtolnay/prettyplease + +// Quote types in rustdoc of other crates get linked to here. +#![doc(html_root_url = "https://docs.rs/quote/1.0.35")] +#![allow( + clippy::doc_markdown, + clippy::missing_errors_doc, + clippy::missing_panics_doc, + clippy::module_name_repetitions, + // false positive https://github.com/rust-lang/rust-clippy/issues/6983 + clippy::wrong_self_convention, +)] + +extern crate alloc; + +#[cfg(feature = "proc-macro")] +extern crate proc_macro; + +mod ext; +mod format; +mod ident_fragment; +mod to_tokens; + +// Not public API. +#[doc(hidden)] +#[path = "runtime.rs"] +pub mod __private; + +pub use crate::ext::TokenStreamExt; +pub use crate::ident_fragment::IdentFragment; +pub use crate::to_tokens::ToTokens; + +// Not public API. +#[doc(hidden)] +pub mod spanned; + +/// The whole point. +/// +/// Performs variable interpolation against the input and produces it as +/// [`proc_macro2::TokenStream`]. +/// +/// Note: for returning tokens to the compiler in a procedural macro, use +/// `.into()` on the result to convert to [`proc_macro::TokenStream`]. +/// +/// [`TokenStream`]: https://docs.rs/proc-macro2/1.0/proc_macro2/struct.TokenStream.html +/// +/// <br> +/// +/// # Interpolation +/// +/// Variable interpolation is done with `#var` (similar to `$var` in +/// `macro_rules!` macros). This grabs the `var` variable that is currently in +/// scope and inserts it in that location in the output tokens. Any type +/// implementing the [`ToTokens`] trait can be interpolated. This includes most +/// Rust primitive types as well as most of the syntax tree types from the [Syn] +/// crate. +/// +/// [`ToTokens`]: trait.ToTokens.html +/// [Syn]: https://github.com/dtolnay/syn +/// +/// Repetition is done using `#(...)*` or `#(...),*` again similar to +/// `macro_rules!`. This iterates through the elements of any variable +/// interpolated within the repetition and inserts a copy of the repetition body +/// for each one. The variables in an interpolation may be a `Vec`, slice, +/// `BTreeSet`, or any `Iterator`. +/// +/// - `#(#var)*` — no separators +/// - `#(#var),*` — the character before the asterisk is used as a separator +/// - `#( struct #var; )*` — the repetition can contain other tokens +/// - `#( #k => println!("{}", #v), )*` — even multiple interpolations +/// +/// <br> +/// +/// # Hygiene +/// +/// Any interpolated tokens preserve the `Span` information provided by their +/// `ToTokens` implementation. Tokens that originate within the `quote!` +/// invocation are spanned with [`Span::call_site()`]. +/// +/// [`Span::call_site()`]: https://docs.rs/proc-macro2/1.0/proc_macro2/struct.Span.html#method.call_site +/// +/// A different span can be provided through the [`quote_spanned!`] macro. +/// +/// [`quote_spanned!`]: macro.quote_spanned.html +/// +/// <br> +/// +/// # Return type +/// +/// The macro evaluates to an expression of type `proc_macro2::TokenStream`. +/// Meanwhile Rust procedural macros are expected to return the type +/// `proc_macro::TokenStream`. +/// +/// The difference between the two types is that `proc_macro` types are entirely +/// specific to procedural macros and cannot ever exist in code outside of a +/// procedural macro, while `proc_macro2` types may exist anywhere including +/// tests and non-macro code like main.rs and build.rs. This is why even the +/// procedural macro ecosystem is largely built around `proc_macro2`, because +/// that ensures the libraries are unit testable and accessible in non-macro +/// contexts. +/// +/// There is a [`From`]-conversion in both directions so returning the output of +/// `quote!` from a procedural macro usually looks like `tokens.into()` or +/// `proc_macro::TokenStream::from(tokens)`. +/// +/// [`From`]: https://doc.rust-lang.org/std/convert/trait.From.html +/// +/// <br> +/// +/// # Examples +/// +/// ### Procedural macro +/// +/// The structure of a basic procedural macro is as follows. Refer to the [Syn] +/// crate for further useful guidance on using `quote!` as part of a procedural +/// macro. +/// +/// [Syn]: https://github.com/dtolnay/syn +/// +/// ``` +/// # #[cfg(any())] +/// extern crate proc_macro; +/// # extern crate proc_macro2; +/// +/// # #[cfg(any())] +/// use proc_macro::TokenStream; +/// # use proc_macro2::TokenStream; +/// use quote::quote; +/// +/// # const IGNORE_TOKENS: &'static str = stringify! { +/// #[proc_macro_derive(HeapSize)] +/// # }; +/// pub fn derive_heap_size(input: TokenStream) -> TokenStream { +/// // Parse the input and figure out what implementation to generate... +/// # const IGNORE_TOKENS: &'static str = stringify! { +/// let name = /* ... */; +/// let expr = /* ... */; +/// # }; +/// # +/// # let name = 0; +/// # let expr = 0; +/// +/// let expanded = quote! { +/// // The generated impl. +/// impl heapsize::HeapSize for #name { +/// fn heap_size_of_children(&self) -> usize { +/// #expr +/// } +/// } +/// }; +/// +/// // Hand the output tokens back to the compiler. +/// TokenStream::from(expanded) +/// } +/// ``` +/// +/// <p><br></p> +/// +/// ### Combining quoted fragments +/// +/// Usually you don't end up constructing an entire final `TokenStream` in one +/// piece. Different parts may come from different helper functions. The tokens +/// produced by `quote!` themselves implement `ToTokens` and so can be +/// interpolated into later `quote!` invocations to build up a final result. +/// +/// ``` +/// # use quote::quote; +/// # +/// let type_definition = quote! {...}; +/// let methods = quote! {...}; +/// +/// let tokens = quote! { +/// #type_definition +/// #methods +/// }; +/// ``` +/// +/// <p><br></p> +/// +/// ### Constructing identifiers +/// +/// Suppose we have an identifier `ident` which came from somewhere in a macro +/// input and we need to modify it in some way for the macro output. Let's +/// consider prepending the identifier with an underscore. +/// +/// Simply interpolating the identifier next to an underscore will not have the +/// behavior of concatenating them. The underscore and the identifier will +/// continue to be two separate tokens as if you had written `_ x`. +/// +/// ``` +/// # use proc_macro2::{self as syn, Span}; +/// # use quote::quote; +/// # +/// # let ident = syn::Ident::new("i", Span::call_site()); +/// # +/// // incorrect +/// quote! { +/// let mut _#ident = 0; +/// } +/// # ; +/// ``` +/// +/// The solution is to build a new identifier token with the correct value. As +/// this is such a common case, the [`format_ident!`] macro provides a +/// convenient utility for doing so correctly. +/// +/// ``` +/// # use proc_macro2::{Ident, Span}; +/// # use quote::{format_ident, quote}; +/// # +/// # let ident = Ident::new("i", Span::call_site()); +/// # +/// let varname = format_ident!("_{}", ident); +/// quote! { +/// let mut #varname = 0; +/// } +/// # ; +/// ``` +/// +/// Alternatively, the APIs provided by Syn and proc-macro2 can be used to +/// directly build the identifier. This is roughly equivalent to the above, but +/// will not handle `ident` being a raw identifier. +/// +/// ``` +/// # use proc_macro2::{self as syn, Span}; +/// # use quote::quote; +/// # +/// # let ident = syn::Ident::new("i", Span::call_site()); +/// # +/// let concatenated = format!("_{}", ident); +/// let varname = syn::Ident::new(&concatenated, ident.span()); +/// quote! { +/// let mut #varname = 0; +/// } +/// # ; +/// ``` +/// +/// <p><br></p> +/// +/// ### Making method calls +/// +/// Let's say our macro requires some type specified in the macro input to have +/// a constructor called `new`. We have the type in a variable called +/// `field_type` of type `syn::Type` and want to invoke the constructor. +/// +/// ``` +/// # use quote::quote; +/// # +/// # let field_type = quote!(...); +/// # +/// // incorrect +/// quote! { +/// let value = #field_type::new(); +/// } +/// # ; +/// ``` +/// +/// This works only sometimes. If `field_type` is `String`, the expanded code +/// contains `String::new()` which is fine. But if `field_type` is something +/// like `Vec<i32>` then the expanded code is `Vec<i32>::new()` which is invalid +/// syntax. Ordinarily in handwritten Rust we would write `Vec::<i32>::new()` +/// but for macros often the following is more convenient. +/// +/// ``` +/// # use quote::quote; +/// # +/// # let field_type = quote!(...); +/// # +/// quote! { +/// let value = <#field_type>::new(); +/// } +/// # ; +/// ``` +/// +/// This expands to `<Vec<i32>>::new()` which behaves correctly. +/// +/// A similar pattern is appropriate for trait methods. +/// +/// ``` +/// # use quote::quote; +/// # +/// # let field_type = quote!(...); +/// # +/// quote! { +/// let value = <#field_type as core::default::Default>::default(); +/// } +/// # ; +/// ``` +/// +/// <p><br></p> +/// +/// ### Interpolating text inside of doc comments +/// +/// Neither doc comments nor string literals get interpolation behavior in +/// quote: +/// +/// ```compile_fail +/// quote! { +/// /// try to interpolate: #ident +/// /// +/// /// ... +/// } +/// ``` +/// +/// ```compile_fail +/// quote! { +/// #[doc = "try to interpolate: #ident"] +/// } +/// ``` +/// +/// Instead the best way to build doc comments that involve variables is by +/// formatting the doc string literal outside of quote. +/// +/// ```rust +/// # use proc_macro2::{Ident, Span}; +/// # use quote::quote; +/// # +/// # const IGNORE: &str = stringify! { +/// let msg = format!(...); +/// # }; +/// # +/// # let ident = Ident::new("var", Span::call_site()); +/// # let msg = format!("try to interpolate: {}", ident); +/// quote! { +/// #[doc = #msg] +/// /// +/// /// ... +/// } +/// # ; +/// ``` +/// +/// <p><br></p> +/// +/// ### Indexing into a tuple struct +/// +/// When interpolating indices of a tuple or tuple struct, we need them not to +/// appears suffixed as integer literals by interpolating them as [`syn::Index`] +/// instead. +/// +/// [`syn::Index`]: https://docs.rs/syn/2.0/syn/struct.Index.html +/// +/// ```compile_fail +/// let i = 0usize..self.fields.len(); +/// +/// // expands to 0 + self.0usize.heap_size() + self.1usize.heap_size() + ... +/// // which is not valid syntax +/// quote! { +/// 0 #( + self.#i.heap_size() )* +/// } +/// ``` +/// +/// ``` +/// # use proc_macro2::{Ident, TokenStream}; +/// # use quote::quote; +/// # +/// # mod syn { +/// # use proc_macro2::{Literal, TokenStream}; +/// # use quote::{ToTokens, TokenStreamExt}; +/// # +/// # pub struct Index(usize); +/// # +/// # impl From<usize> for Index { +/// # fn from(i: usize) -> Self { +/// # Index(i) +/// # } +/// # } +/// # +/// # impl ToTokens for Index { +/// # fn to_tokens(&self, tokens: &mut TokenStream) { +/// # tokens.append(Literal::usize_unsuffixed(self.0)); +/// # } +/// # } +/// # } +/// # +/// # struct Struct { +/// # fields: Vec<Ident>, +/// # } +/// # +/// # impl Struct { +/// # fn example(&self) -> TokenStream { +/// let i = (0..self.fields.len()).map(syn::Index::from); +/// +/// // expands to 0 + self.0.heap_size() + self.1.heap_size() + ... +/// quote! { +/// 0 #( + self.#i.heap_size() )* +/// } +/// # } +/// # } +/// ``` +#[cfg(doc)] +#[macro_export] +macro_rules! quote { + ($($tt:tt)*) => { + ... + }; +} + +#[cfg(not(doc))] +#[macro_export] +macro_rules! quote { + () => { + $crate::__private::TokenStream::new() + }; + + // Special case rule for a single tt, for performance. + ($tt:tt) => {{ + let mut _s = $crate::__private::TokenStream::new(); + $crate::quote_token!{$tt _s} + _s + }}; + + // Special case rules for two tts, for performance. + (# $var:ident) => {{ + let mut _s = $crate::__private::TokenStream::new(); + $crate::ToTokens::to_tokens(&$var, &mut _s); + _s + }}; + ($tt1:tt $tt2:tt) => {{ + let mut _s = $crate::__private::TokenStream::new(); + $crate::quote_token!{$tt1 _s} + $crate::quote_token!{$tt2 _s} + _s + }}; + + // Rule for any other number of tokens. + ($($tt:tt)*) => {{ + let mut _s = $crate::__private::TokenStream::new(); + $crate::quote_each_token!{_s $($tt)*} + _s + }}; +} + +/// Same as `quote!`, but applies a given span to all tokens originating within +/// the macro invocation. +/// +/// <br> +/// +/// # Syntax +/// +/// A span expression of type [`Span`], followed by `=>`, followed by the tokens +/// to quote. The span expression should be brief — use a variable for +/// anything more than a few characters. There should be no space before the +/// `=>` token. +/// +/// [`Span`]: https://docs.rs/proc-macro2/1.0/proc_macro2/struct.Span.html +/// +/// ``` +/// # use proc_macro2::Span; +/// # use quote::quote_spanned; +/// # +/// # const IGNORE_TOKENS: &'static str = stringify! { +/// let span = /* ... */; +/// # }; +/// # let span = Span::call_site(); +/// # let init = 0; +/// +/// // On one line, use parentheses. +/// let tokens = quote_spanned!(span=> Box::into_raw(Box::new(#init))); +/// +/// // On multiple lines, place the span at the top and use braces. +/// let tokens = quote_spanned! {span=> +/// Box::into_raw(Box::new(#init)) +/// }; +/// ``` +/// +/// The lack of space before the `=>` should look jarring to Rust programmers +/// and this is intentional. The formatting is designed to be visibly +/// off-balance and draw the eye a particular way, due to the span expression +/// being evaluated in the context of the procedural macro and the remaining +/// tokens being evaluated in the generated code. +/// +/// <br> +/// +/// # Hygiene +/// +/// Any interpolated tokens preserve the `Span` information provided by their +/// `ToTokens` implementation. Tokens that originate within the `quote_spanned!` +/// invocation are spanned with the given span argument. +/// +/// <br> +/// +/// # Example +/// +/// The following procedural macro code uses `quote_spanned!` to assert that a +/// particular Rust type implements the [`Sync`] trait so that references can be +/// safely shared between threads. +/// +/// [`Sync`]: https://doc.rust-lang.org/std/marker/trait.Sync.html +/// +/// ``` +/// # use quote::{quote_spanned, TokenStreamExt, ToTokens}; +/// # use proc_macro2::{Span, TokenStream}; +/// # +/// # struct Type; +/// # +/// # impl Type { +/// # fn span(&self) -> Span { +/// # Span::call_site() +/// # } +/// # } +/// # +/// # impl ToTokens for Type { +/// # fn to_tokens(&self, _tokens: &mut TokenStream) {} +/// # } +/// # +/// # let ty = Type; +/// # let call_site = Span::call_site(); +/// # +/// let ty_span = ty.span(); +/// let assert_sync = quote_spanned! {ty_span=> +/// struct _AssertSync where #ty: Sync; +/// }; +/// ``` +/// +/// If the assertion fails, the user will see an error like the following. The +/// input span of their type is highlighted in the error. +/// +/// ```text +/// error[E0277]: the trait bound `*const (): std::marker::Sync` is not satisfied +/// --> src/main.rs:10:21 +/// | +/// 10 | static ref PTR: *const () = &(); +/// | ^^^^^^^^^ `*const ()` cannot be shared between threads safely +/// ``` +/// +/// In this example it is important for the where-clause to be spanned with the +/// line/column information of the user's input type so that error messages are +/// placed appropriately by the compiler. +#[cfg(doc)] +#[macro_export] +macro_rules! quote_spanned { + ($span:expr=> $($tt:tt)*) => { + ... + }; +} + +#[cfg(not(doc))] +#[macro_export] +macro_rules! quote_spanned { + ($span:expr=>) => {{ + let _: $crate::__private::Span = $crate::__private::get_span($span).__into_span(); + $crate::__private::TokenStream::new() + }}; + + // Special case rule for a single tt, for performance. + ($span:expr=> $tt:tt) => {{ + let mut _s = $crate::__private::TokenStream::new(); + let _span: $crate::__private::Span = $crate::__private::get_span($span).__into_span(); + $crate::quote_token_spanned!{$tt _s _span} + _s + }}; + + // Special case rules for two tts, for performance. + ($span:expr=> # $var:ident) => {{ + let mut _s = $crate::__private::TokenStream::new(); + let _: $crate::__private::Span = $crate::__private::get_span($span).__into_span(); + $crate::ToTokens::to_tokens(&$var, &mut _s); + _s + }}; + ($span:expr=> $tt1:tt $tt2:tt) => {{ + let mut _s = $crate::__private::TokenStream::new(); + let _span: $crate::__private::Span = $crate::__private::get_span($span).__into_span(); + $crate::quote_token_spanned!{$tt1 _s _span} + $crate::quote_token_spanned!{$tt2 _s _span} + _s + }}; + + // Rule for any other number of tokens. + ($span:expr=> $($tt:tt)*) => {{ + let mut _s = $crate::__private::TokenStream::new(); + let _span: $crate::__private::Span = $crate::__private::get_span($span).__into_span(); + $crate::quote_each_token_spanned!{_s _span $($tt)*} + _s + }}; +} + +// Extract the names of all #metavariables and pass them to the $call macro. +// +// in: pounded_var_names!(then!(...) a #b c #( #d )* #e) +// out: then!(... b); +// then!(... d); +// then!(... e); +#[macro_export] +#[doc(hidden)] +macro_rules! pounded_var_names { + ($call:ident! $extra:tt $($tts:tt)*) => { + $crate::pounded_var_names_with_context!{$call! $extra + (@ $($tts)*) + ($($tts)* @) + } + }; +} + +#[macro_export] +#[doc(hidden)] +macro_rules! pounded_var_names_with_context { + ($call:ident! $extra:tt ($($b1:tt)*) ($($curr:tt)*)) => { + $( + $crate::pounded_var_with_context!{$call! $extra $b1 $curr} + )* + }; +} + +#[macro_export] +#[doc(hidden)] +macro_rules! pounded_var_with_context { + ($call:ident! $extra:tt $b1:tt ( $($inner:tt)* )) => { + $crate::pounded_var_names!{$call! $extra $($inner)*} + }; + + ($call:ident! $extra:tt $b1:tt [ $($inner:tt)* ]) => { + $crate::pounded_var_names!{$call! $extra $($inner)*} + }; + + ($call:ident! $extra:tt $b1:tt { $($inner:tt)* }) => { + $crate::pounded_var_names!{$call! $extra $($inner)*} + }; + + ($call:ident!($($extra:tt)*) # $var:ident) => { + $crate::$call!($($extra)* $var); + }; + + ($call:ident! $extra:tt $b1:tt $curr:tt) => {}; +} + +#[macro_export] +#[doc(hidden)] +macro_rules! quote_bind_into_iter { + ($has_iter:ident $var:ident) => { + // `mut` may be unused if $var occurs multiple times in the list. + #[allow(unused_mut)] + let (mut $var, i) = $var.quote_into_iter(); + let $has_iter = $has_iter | i; + }; +} + +#[macro_export] +#[doc(hidden)] +macro_rules! quote_bind_next_or_break { + ($var:ident) => { + let $var = match $var.next() { + Some(_x) => $crate::__private::RepInterp(_x), + None => break, + }; + }; +} + +// The obvious way to write this macro is as a tt muncher. This implementation +// does something more complex for two reasons. +// +// - With a tt muncher it's easy to hit Rust's built-in recursion_limit, which +// this implementation avoids because it isn't tail recursive. +// +// - Compile times for a tt muncher are quadratic relative to the length of +// the input. This implementation is linear, so it will be faster +// (potentially much faster) for big inputs. However, the constant factors +// of this implementation are higher than that of a tt muncher, so it is +// somewhat slower than a tt muncher if there are many invocations with +// short inputs. +// +// An invocation like this: +// +// quote_each_token!(_s a b c d e f g h i j); +// +// expands to this: +// +// quote_tokens_with_context!(_s +// (@ @ @ @ @ @ a b c d e f g h i j) +// (@ @ @ @ @ a b c d e f g h i j @) +// (@ @ @ @ a b c d e f g h i j @ @) +// (@ @ @ (a) (b) (c) (d) (e) (f) (g) (h) (i) (j) @ @ @) +// (@ @ a b c d e f g h i j @ @ @ @) +// (@ a b c d e f g h i j @ @ @ @ @) +// (a b c d e f g h i j @ @ @ @ @ @) +// ); +// +// which gets transposed and expanded to this: +// +// quote_token_with_context!(_s @ @ @ @ @ @ a); +// quote_token_with_context!(_s @ @ @ @ @ a b); +// quote_token_with_context!(_s @ @ @ @ a b c); +// quote_token_with_context!(_s @ @ @ (a) b c d); +// quote_token_with_context!(_s @ @ a (b) c d e); +// quote_token_with_context!(_s @ a b (c) d e f); +// quote_token_with_context!(_s a b c (d) e f g); +// quote_token_with_context!(_s b c d (e) f g h); +// quote_token_with_context!(_s c d e (f) g h i); +// quote_token_with_context!(_s d e f (g) h i j); +// quote_token_with_context!(_s e f g (h) i j @); +// quote_token_with_context!(_s f g h (i) j @ @); +// quote_token_with_context!(_s g h i (j) @ @ @); +// quote_token_with_context!(_s h i j @ @ @ @); +// quote_token_with_context!(_s i j @ @ @ @ @); +// quote_token_with_context!(_s j @ @ @ @ @ @); +// +// Without having used muncher-style recursion, we get one invocation of +// quote_token_with_context for each original tt, with three tts of context on +// either side. This is enough for the longest possible interpolation form (a +// repetition with separator, as in `# (#var) , *`) to be fully represented with +// the first or last tt in the middle. +// +// The middle tt (surrounded by parentheses) is the tt being processed. +// +// - When it is a `#`, quote_token_with_context can do an interpolation. The +// interpolation kind will depend on the three subsequent tts. +// +// - When it is within a later part of an interpolation, it can be ignored +// because the interpolation has already been done. +// +// - When it is not part of an interpolation it can be pushed as a single +// token into the output. +// +// - When the middle token is an unparenthesized `@`, that call is one of the +// first 3 or last 3 calls of quote_token_with_context and does not +// correspond to one of the original input tokens, so turns into nothing. +#[macro_export] +#[doc(hidden)] +macro_rules! quote_each_token { + ($tokens:ident $($tts:tt)*) => { + $crate::quote_tokens_with_context!{$tokens + (@ @ @ @ @ @ $($tts)*) + (@ @ @ @ @ $($tts)* @) + (@ @ @ @ $($tts)* @ @) + (@ @ @ $(($tts))* @ @ @) + (@ @ $($tts)* @ @ @ @) + (@ $($tts)* @ @ @ @ @) + ($($tts)* @ @ @ @ @ @) + } + }; +} + +// See the explanation on quote_each_token. +#[macro_export] +#[doc(hidden)] +macro_rules! quote_each_token_spanned { + ($tokens:ident $span:ident $($tts:tt)*) => { + $crate::quote_tokens_with_context_spanned!{$tokens $span + (@ @ @ @ @ @ $($tts)*) + (@ @ @ @ @ $($tts)* @) + (@ @ @ @ $($tts)* @ @) + (@ @ @ $(($tts))* @ @ @) + (@ @ $($tts)* @ @ @ @) + (@ $($tts)* @ @ @ @ @) + ($($tts)* @ @ @ @ @ @) + } + }; +} + +// See the explanation on quote_each_token. +#[macro_export] +#[doc(hidden)] +macro_rules! quote_tokens_with_context { + ($tokens:ident + ($($b3:tt)*) ($($b2:tt)*) ($($b1:tt)*) + ($($curr:tt)*) + ($($a1:tt)*) ($($a2:tt)*) ($($a3:tt)*) + ) => { + $( + $crate::quote_token_with_context!{$tokens $b3 $b2 $b1 $curr $a1 $a2 $a3} + )* + }; +} + +// See the explanation on quote_each_token. +#[macro_export] +#[doc(hidden)] +macro_rules! quote_tokens_with_context_spanned { + ($tokens:ident $span:ident + ($($b3:tt)*) ($($b2:tt)*) ($($b1:tt)*) + ($($curr:tt)*) + ($($a1:tt)*) ($($a2:tt)*) ($($a3:tt)*) + ) => { + $( + $crate::quote_token_with_context_spanned!{$tokens $span $b3 $b2 $b1 $curr $a1 $a2 $a3} + )* + }; +} + +// See the explanation on quote_each_token. +#[macro_export] +#[doc(hidden)] +macro_rules! quote_token_with_context { + // Unparenthesized `@` indicates this call does not correspond to one of the + // original input tokens. Ignore it. + ($tokens:ident $b3:tt $b2:tt $b1:tt @ $a1:tt $a2:tt $a3:tt) => {}; + + // A repetition with no separator. + ($tokens:ident $b3:tt $b2:tt $b1:tt (#) ( $($inner:tt)* ) * $a3:tt) => {{ + use $crate::__private::ext::*; + let has_iter = $crate::__private::ThereIsNoIteratorInRepetition; + $crate::pounded_var_names!{quote_bind_into_iter!(has_iter) () $($inner)*} + let _: $crate::__private::HasIterator = has_iter; + // This is `while true` instead of `loop` because if there are no + // iterators used inside of this repetition then the body would not + // contain any `break`, so the compiler would emit unreachable code + // warnings on anything below the loop. We use has_iter to detect and + // fail to compile when there are no iterators, so here we just work + // around the unneeded extra warning. + while true { + $crate::pounded_var_names!{quote_bind_next_or_break!() () $($inner)*} + $crate::quote_each_token!{$tokens $($inner)*} + } + }}; + // ... and one step later. + ($tokens:ident $b3:tt $b2:tt # (( $($inner:tt)* )) * $a2:tt $a3:tt) => {}; + // ... and one step later. + ($tokens:ident $b3:tt # ( $($inner:tt)* ) (*) $a1:tt $a2:tt $a3:tt) => {}; + + // A repetition with separator. + ($tokens:ident $b3:tt $b2:tt $b1:tt (#) ( $($inner:tt)* ) $sep:tt *) => {{ + use $crate::__private::ext::*; + let mut _i = 0usize; + let has_iter = $crate::__private::ThereIsNoIteratorInRepetition; + $crate::pounded_var_names!{quote_bind_into_iter!(has_iter) () $($inner)*} + let _: $crate::__private::HasIterator = has_iter; + while true { + $crate::pounded_var_names!{quote_bind_next_or_break!() () $($inner)*} + if _i > 0 { + $crate::quote_token!{$sep $tokens} + } + _i += 1; + $crate::quote_each_token!{$tokens $($inner)*} + } + }}; + // ... and one step later. + ($tokens:ident $b3:tt $b2:tt # (( $($inner:tt)* )) $sep:tt * $a3:tt) => {}; + // ... and one step later. + ($tokens:ident $b3:tt # ( $($inner:tt)* ) ($sep:tt) * $a2:tt $a3:tt) => {}; + // (A special case for `#(var)**`, where the first `*` is treated as the + // repetition symbol and the second `*` is treated as an ordinary token.) + ($tokens:ident # ( $($inner:tt)* ) * (*) $a1:tt $a2:tt $a3:tt) => { + // https://github.com/dtolnay/quote/issues/130 + $crate::quote_token!{* $tokens} + }; + // ... and one step later. + ($tokens:ident # ( $($inner:tt)* ) $sep:tt (*) $a1:tt $a2:tt $a3:tt) => {}; + + // A non-repetition interpolation. + ($tokens:ident $b3:tt $b2:tt $b1:tt (#) $var:ident $a2:tt $a3:tt) => { + $crate::ToTokens::to_tokens(&$var, &mut $tokens); + }; + // ... and one step later. + ($tokens:ident $b3:tt $b2:tt # ($var:ident) $a1:tt $a2:tt $a3:tt) => {}; + + // An ordinary token, not part of any interpolation. + ($tokens:ident $b3:tt $b2:tt $b1:tt ($curr:tt) $a1:tt $a2:tt $a3:tt) => { + $crate::quote_token!{$curr $tokens} + }; +} + +// See the explanation on quote_each_token, and on the individual rules of +// quote_token_with_context. +#[macro_export] +#[doc(hidden)] +macro_rules! quote_token_with_context_spanned { + ($tokens:ident $span:ident $b3:tt $b2:tt $b1:tt @ $a1:tt $a2:tt $a3:tt) => {}; + + ($tokens:ident $span:ident $b3:tt $b2:tt $b1:tt (#) ( $($inner:tt)* ) * $a3:tt) => {{ + use $crate::__private::ext::*; + let has_iter = $crate::__private::ThereIsNoIteratorInRepetition; + $crate::pounded_var_names!{quote_bind_into_iter!(has_iter) () $($inner)*} + let _: $crate::__private::HasIterator = has_iter; + while true { + $crate::pounded_var_names!{quote_bind_next_or_break!() () $($inner)*} + $crate::quote_each_token_spanned!{$tokens $span $($inner)*} + } + }}; + ($tokens:ident $span:ident $b3:tt $b2:tt # (( $($inner:tt)* )) * $a2:tt $a3:tt) => {}; + ($tokens:ident $span:ident $b3:tt # ( $($inner:tt)* ) (*) $a1:tt $a2:tt $a3:tt) => {}; + + ($tokens:ident $span:ident $b3:tt $b2:tt $b1:tt (#) ( $($inner:tt)* ) $sep:tt *) => {{ + use $crate::__private::ext::*; + let mut _i = 0usize; + let has_iter = $crate::__private::ThereIsNoIteratorInRepetition; + $crate::pounded_var_names!{quote_bind_into_iter!(has_iter) () $($inner)*} + let _: $crate::__private::HasIterator = has_iter; + while true { + $crate::pounded_var_names!{quote_bind_next_or_break!() () $($inner)*} + if _i > 0 { + $crate::quote_token_spanned!{$sep $tokens $span} + } + _i += 1; + $crate::quote_each_token_spanned!{$tokens $span $($inner)*} + } + }}; + ($tokens:ident $span:ident $b3:tt $b2:tt # (( $($inner:tt)* )) $sep:tt * $a3:tt) => {}; + ($tokens:ident $span:ident $b3:tt # ( $($inner:tt)* ) ($sep:tt) * $a2:tt $a3:tt) => {}; + ($tokens:ident $span:ident # ( $($inner:tt)* ) * (*) $a1:tt $a2:tt $a3:tt) => { + // https://github.com/dtolnay/quote/issues/130 + $crate::quote_token_spanned!{* $tokens $span} + }; + ($tokens:ident $span:ident # ( $($inner:tt)* ) $sep:tt (*) $a1:tt $a2:tt $a3:tt) => {}; + + ($tokens:ident $span:ident $b3:tt $b2:tt $b1:tt (#) $var:ident $a2:tt $a3:tt) => { + $crate::ToTokens::to_tokens(&$var, &mut $tokens); + }; + ($tokens:ident $span:ident $b3:tt $b2:tt # ($var:ident) $a1:tt $a2:tt $a3:tt) => {}; + + ($tokens:ident $span:ident $b3:tt $b2:tt $b1:tt ($curr:tt) $a1:tt $a2:tt $a3:tt) => { + $crate::quote_token_spanned!{$curr $tokens $span} + }; +} + +// These rules are ordered by approximate token frequency, at least for the +// first 10 or so, to improve compile times. Having `ident` first is by far the +// most important because it's typically 2-3x more common than the next most +// common token. +// +// Separately, we put the token being matched in the very front so that failing +// rules may fail to match as quickly as possible. +#[macro_export] +#[doc(hidden)] +macro_rules! quote_token { + ($ident:ident $tokens:ident) => { + $crate::__private::push_ident(&mut $tokens, stringify!($ident)); + }; + + (:: $tokens:ident) => { + $crate::__private::push_colon2(&mut $tokens); + }; + + (( $($inner:tt)* ) $tokens:ident) => { + $crate::__private::push_group( + &mut $tokens, + $crate::__private::Delimiter::Parenthesis, + $crate::quote!($($inner)*), + ); + }; + + ([ $($inner:tt)* ] $tokens:ident) => { + $crate::__private::push_group( + &mut $tokens, + $crate::__private::Delimiter::Bracket, + $crate::quote!($($inner)*), + ); + }; + + ({ $($inner:tt)* } $tokens:ident) => { + $crate::__private::push_group( + &mut $tokens, + $crate::__private::Delimiter::Brace, + $crate::quote!($($inner)*), + ); + }; + + (# $tokens:ident) => { + $crate::__private::push_pound(&mut $tokens); + }; + + (, $tokens:ident) => { + $crate::__private::push_comma(&mut $tokens); + }; + + (. $tokens:ident) => { + $crate::__private::push_dot(&mut $tokens); + }; + + (; $tokens:ident) => { + $crate::__private::push_semi(&mut $tokens); + }; + + (: $tokens:ident) => { + $crate::__private::push_colon(&mut $tokens); + }; + + (+ $tokens:ident) => { + $crate::__private::push_add(&mut $tokens); + }; + + (+= $tokens:ident) => { + $crate::__private::push_add_eq(&mut $tokens); + }; + + (& $tokens:ident) => { + $crate::__private::push_and(&mut $tokens); + }; + + (&& $tokens:ident) => { + $crate::__private::push_and_and(&mut $tokens); + }; + + (&= $tokens:ident) => { + $crate::__private::push_and_eq(&mut $tokens); + }; + + (@ $tokens:ident) => { + $crate::__private::push_at(&mut $tokens); + }; + + (! $tokens:ident) => { + $crate::__private::push_bang(&mut $tokens); + }; + + (^ $tokens:ident) => { + $crate::__private::push_caret(&mut $tokens); + }; + + (^= $tokens:ident) => { + $crate::__private::push_caret_eq(&mut $tokens); + }; + + (/ $tokens:ident) => { + $crate::__private::push_div(&mut $tokens); + }; + + (/= $tokens:ident) => { + $crate::__private::push_div_eq(&mut $tokens); + }; + + (.. $tokens:ident) => { + $crate::__private::push_dot2(&mut $tokens); + }; + + (... $tokens:ident) => { + $crate::__private::push_dot3(&mut $tokens); + }; + + (..= $tokens:ident) => { + $crate::__private::push_dot_dot_eq(&mut $tokens); + }; + + (= $tokens:ident) => { + $crate::__private::push_eq(&mut $tokens); + }; + + (== $tokens:ident) => { + $crate::__private::push_eq_eq(&mut $tokens); + }; + + (>= $tokens:ident) => { + $crate::__private::push_ge(&mut $tokens); + }; + + (> $tokens:ident) => { + $crate::__private::push_gt(&mut $tokens); + }; + + (<= $tokens:ident) => { + $crate::__private::push_le(&mut $tokens); + }; + + (< $tokens:ident) => { + $crate::__private::push_lt(&mut $tokens); + }; + + (*= $tokens:ident) => { + $crate::__private::push_mul_eq(&mut $tokens); + }; + + (!= $tokens:ident) => { + $crate::__private::push_ne(&mut $tokens); + }; + + (| $tokens:ident) => { + $crate::__private::push_or(&mut $tokens); + }; + + (|= $tokens:ident) => { + $crate::__private::push_or_eq(&mut $tokens); + }; + + (|| $tokens:ident) => { + $crate::__private::push_or_or(&mut $tokens); + }; + + (? $tokens:ident) => { + $crate::__private::push_question(&mut $tokens); + }; + + (-> $tokens:ident) => { + $crate::__private::push_rarrow(&mut $tokens); + }; + + (<- $tokens:ident) => { + $crate::__private::push_larrow(&mut $tokens); + }; + + (% $tokens:ident) => { + $crate::__private::push_rem(&mut $tokens); + }; + + (%= $tokens:ident) => { + $crate::__private::push_rem_eq(&mut $tokens); + }; + + (=> $tokens:ident) => { + $crate::__private::push_fat_arrow(&mut $tokens); + }; + + (<< $tokens:ident) => { + $crate::__private::push_shl(&mut $tokens); + }; + + (<<= $tokens:ident) => { + $crate::__private::push_shl_eq(&mut $tokens); + }; + + (>> $tokens:ident) => { + $crate::__private::push_shr(&mut $tokens); + }; + + (>>= $tokens:ident) => { + $crate::__private::push_shr_eq(&mut $tokens); + }; + + (* $tokens:ident) => { + $crate::__private::push_star(&mut $tokens); + }; + + (- $tokens:ident) => { + $crate::__private::push_sub(&mut $tokens); + }; + + (-= $tokens:ident) => { + $crate::__private::push_sub_eq(&mut $tokens); + }; + + ($lifetime:lifetime $tokens:ident) => { + $crate::__private::push_lifetime(&mut $tokens, stringify!($lifetime)); + }; + + (_ $tokens:ident) => { + $crate::__private::push_underscore(&mut $tokens); + }; + + ($other:tt $tokens:ident) => { + $crate::__private::parse(&mut $tokens, stringify!($other)); + }; +} + +// See the comment above `quote_token!` about the rule ordering. +#[macro_export] +#[doc(hidden)] +macro_rules! quote_token_spanned { + ($ident:ident $tokens:ident $span:ident) => { + $crate::__private::push_ident_spanned(&mut $tokens, $span, stringify!($ident)); + }; + + (:: $tokens:ident $span:ident) => { + $crate::__private::push_colon2_spanned(&mut $tokens, $span); + }; + + (( $($inner:tt)* ) $tokens:ident $span:ident) => { + $crate::__private::push_group_spanned( + &mut $tokens, + $span, + $crate::__private::Delimiter::Parenthesis, + $crate::quote_spanned!($span=> $($inner)*), + ); + }; + + ([ $($inner:tt)* ] $tokens:ident $span:ident) => { + $crate::__private::push_group_spanned( + &mut $tokens, + $span, + $crate::__private::Delimiter::Bracket, + $crate::quote_spanned!($span=> $($inner)*), + ); + }; + + ({ $($inner:tt)* } $tokens:ident $span:ident) => { + $crate::__private::push_group_spanned( + &mut $tokens, + $span, + $crate::__private::Delimiter::Brace, + $crate::quote_spanned!($span=> $($inner)*), + ); + }; + + (# $tokens:ident $span:ident) => { + $crate::__private::push_pound_spanned(&mut $tokens, $span); + }; + + (, $tokens:ident $span:ident) => { + $crate::__private::push_comma_spanned(&mut $tokens, $span); + }; + + (. $tokens:ident $span:ident) => { + $crate::__private::push_dot_spanned(&mut $tokens, $span); + }; + + (; $tokens:ident $span:ident) => { + $crate::__private::push_semi_spanned(&mut $tokens, $span); + }; + + (: $tokens:ident $span:ident) => { + $crate::__private::push_colon_spanned(&mut $tokens, $span); + }; + + (+ $tokens:ident $span:ident) => { + $crate::__private::push_add_spanned(&mut $tokens, $span); + }; + + (+= $tokens:ident $span:ident) => { + $crate::__private::push_add_eq_spanned(&mut $tokens, $span); + }; + + (& $tokens:ident $span:ident) => { + $crate::__private::push_and_spanned(&mut $tokens, $span); + }; + + (&& $tokens:ident $span:ident) => { + $crate::__private::push_and_and_spanned(&mut $tokens, $span); + }; + + (&= $tokens:ident $span:ident) => { + $crate::__private::push_and_eq_spanned(&mut $tokens, $span); + }; + + (@ $tokens:ident $span:ident) => { + $crate::__private::push_at_spanned(&mut $tokens, $span); + }; + + (! $tokens:ident $span:ident) => { + $crate::__private::push_bang_spanned(&mut $tokens, $span); + }; + + (^ $tokens:ident $span:ident) => { + $crate::__private::push_caret_spanned(&mut $tokens, $span); + }; + + (^= $tokens:ident $span:ident) => { + $crate::__private::push_caret_eq_spanned(&mut $tokens, $span); + }; + + (/ $tokens:ident $span:ident) => { + $crate::__private::push_div_spanned(&mut $tokens, $span); + }; + + (/= $tokens:ident $span:ident) => { + $crate::__private::push_div_eq_spanned(&mut $tokens, $span); + }; + + (.. $tokens:ident $span:ident) => { + $crate::__private::push_dot2_spanned(&mut $tokens, $span); + }; + + (... $tokens:ident $span:ident) => { + $crate::__private::push_dot3_spanned(&mut $tokens, $span); + }; + + (..= $tokens:ident $span:ident) => { + $crate::__private::push_dot_dot_eq_spanned(&mut $tokens, $span); + }; + + (= $tokens:ident $span:ident) => { + $crate::__private::push_eq_spanned(&mut $tokens, $span); + }; + + (== $tokens:ident $span:ident) => { + $crate::__private::push_eq_eq_spanned(&mut $tokens, $span); + }; + + (>= $tokens:ident $span:ident) => { + $crate::__private::push_ge_spanned(&mut $tokens, $span); + }; + + (> $tokens:ident $span:ident) => { + $crate::__private::push_gt_spanned(&mut $tokens, $span); + }; + + (<= $tokens:ident $span:ident) => { + $crate::__private::push_le_spanned(&mut $tokens, $span); + }; + + (< $tokens:ident $span:ident) => { + $crate::__private::push_lt_spanned(&mut $tokens, $span); + }; + + (*= $tokens:ident $span:ident) => { + $crate::__private::push_mul_eq_spanned(&mut $tokens, $span); + }; + + (!= $tokens:ident $span:ident) => { + $crate::__private::push_ne_spanned(&mut $tokens, $span); + }; + + (| $tokens:ident $span:ident) => { + $crate::__private::push_or_spanned(&mut $tokens, $span); + }; + + (|= $tokens:ident $span:ident) => { + $crate::__private::push_or_eq_spanned(&mut $tokens, $span); + }; + + (|| $tokens:ident $span:ident) => { + $crate::__private::push_or_or_spanned(&mut $tokens, $span); + }; + + (? $tokens:ident $span:ident) => { + $crate::__private::push_question_spanned(&mut $tokens, $span); + }; + + (-> $tokens:ident $span:ident) => { + $crate::__private::push_rarrow_spanned(&mut $tokens, $span); + }; + + (<- $tokens:ident $span:ident) => { + $crate::__private::push_larrow_spanned(&mut $tokens, $span); + }; + + (% $tokens:ident $span:ident) => { + $crate::__private::push_rem_spanned(&mut $tokens, $span); + }; + + (%= $tokens:ident $span:ident) => { + $crate::__private::push_rem_eq_spanned(&mut $tokens, $span); + }; + + (=> $tokens:ident $span:ident) => { + $crate::__private::push_fat_arrow_spanned(&mut $tokens, $span); + }; + + (<< $tokens:ident $span:ident) => { + $crate::__private::push_shl_spanned(&mut $tokens, $span); + }; + + (<<= $tokens:ident $span:ident) => { + $crate::__private::push_shl_eq_spanned(&mut $tokens, $span); + }; + + (>> $tokens:ident $span:ident) => { + $crate::__private::push_shr_spanned(&mut $tokens, $span); + }; + + (>>= $tokens:ident $span:ident) => { + $crate::__private::push_shr_eq_spanned(&mut $tokens, $span); + }; + + (* $tokens:ident $span:ident) => { + $crate::__private::push_star_spanned(&mut $tokens, $span); + }; + + (- $tokens:ident $span:ident) => { + $crate::__private::push_sub_spanned(&mut $tokens, $span); + }; + + (-= $tokens:ident $span:ident) => { + $crate::__private::push_sub_eq_spanned(&mut $tokens, $span); + }; + + ($lifetime:lifetime $tokens:ident $span:ident) => { + $crate::__private::push_lifetime_spanned(&mut $tokens, $span, stringify!($lifetime)); + }; + + (_ $tokens:ident $span:ident) => { + $crate::__private::push_underscore_spanned(&mut $tokens, $span); + }; + + ($other:tt $tokens:ident $span:ident) => { + $crate::__private::parse_spanned(&mut $tokens, $span, stringify!($other)); + }; +} |