//! This module generates AST datatype used by rust-analyzer. //! //! Specifically, it generates the `SyntaxKind` enum and a number of newtype //! wrappers around `SyntaxNode` which implement `ra_syntax::AstNode`. use std::{collections::HashSet, fmt::Write}; use proc_macro2::{Punct, Spacing}; use quote::{format_ident, quote}; use crate::{ ast_src::{rust_ast, AstSrc, Field, FieldSrc, KindsSrc, KINDS_SRC}, codegen::{self, update, Mode}, project_root, Result, }; pub fn generate_syntax(mode: Mode) -> Result<()> { let ast = rust_ast(); let syntax_kinds_file = project_root().join(codegen::SYNTAX_KINDS); let syntax_kinds = generate_syntax_kinds(KINDS_SRC)?; update(syntax_kinds_file.as_path(), &syntax_kinds, mode)?; let ast_tokens_file = project_root().join(codegen::AST_TOKENS); let contents = generate_tokens(&ast)?; update(ast_tokens_file.as_path(), &contents, mode)?; let ast_nodes_file = project_root().join(codegen::AST_NODES); let contents = generate_nodes(KINDS_SRC, &ast)?; update(ast_nodes_file.as_path(), &contents, mode)?; Ok(()) } fn generate_tokens(grammar: &AstSrc) -> Result { let tokens = grammar.tokens.iter().map(|token| { let name = format_ident!("{}", token); let kind = format_ident!("{}", to_upper_snake_case(token)); quote! { #[derive(Debug, Clone, PartialEq, Eq, Hash)] pub struct #name { pub(crate) syntax: SyntaxToken, } impl std::fmt::Display for #name { fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { std::fmt::Display::fmt(&self.syntax, f) } } impl AstToken for #name { fn can_cast(kind: SyntaxKind) -> bool { kind == #kind } fn cast(syntax: SyntaxToken) -> Option { if Self::can_cast(syntax.kind()) { Some(Self { syntax }) } else { None } } fn syntax(&self) -> &SyntaxToken { &self.syntax } } } }); let pretty = crate::reformat(quote! { use crate::{SyntaxKind::{self, *}, SyntaxToken, ast::AstToken}; #(#tokens)* })? .replace("#[derive", "\n#[derive"); Ok(pretty) } fn generate_nodes(kinds: KindsSrc<'_>, grammar: &AstSrc) -> Result { let (node_defs, node_boilerplate_impls): (Vec<_>, Vec<_>) = grammar .nodes .iter() .map(|node| { let name = format_ident!("{}", node.name); let kind = format_ident!("{}", to_upper_snake_case(&node.name)); let traits = node.traits.iter().map(|trait_name| { let trait_name = format_ident!("{}", trait_name); quote!(impl ast::#trait_name for #name {}) }); let methods = node.fields.iter().map(|field| { let method_name = field.method_name(); let ty = field.ty(); if field.is_many() { quote! { pub fn #method_name(&self) -> AstChildren<#ty> { support::children(&self.syntax) } } } else { if let Some(token_kind) = field.token_kind() { quote! { pub fn #method_name(&self) -> Option<#ty> { support::token(&self.syntax, #token_kind) } } } else { quote! { pub fn #method_name(&self) -> Option<#ty> { support::child(&self.syntax) } } } } }); ( quote! { #[pretty_doc_comment_placeholder_workaround] #[derive(Debug, Clone, PartialEq, Eq, Hash)] pub struct #name { pub(crate) syntax: SyntaxNode, } #(#traits)* impl #name { #(#methods)* } }, quote! { impl AstNode for #name { fn can_cast(kind: SyntaxKind) -> bool { kind == #kind } fn cast(syntax: SyntaxNode) -> Option { if Self::can_cast(syntax.kind()) { Some(Self { syntax }) } else { None } } fn syntax(&self) -> &SyntaxNode { &self.syntax } } }, ) }) .unzip(); let (enum_defs, enum_boilerplate_impls): (Vec<_>, Vec<_>) = grammar .enums .iter() .map(|en| { let variants: Vec<_> = en.variants.iter().map(|var| format_ident!("{}", var)).collect(); let name = format_ident!("{}", en.name); let kinds: Vec<_> = variants .iter() .map(|name| format_ident!("{}", to_upper_snake_case(&name.to_string()))) .collect(); let traits = en.traits.iter().map(|trait_name| { let trait_name = format_ident!("{}", trait_name); quote!(impl ast::#trait_name for #name {}) }); ( quote! { #[pretty_doc_comment_placeholder_workaround] #[derive(Debug, Clone, PartialEq, Eq, Hash)] pub enum #name { #(#variants(#variants),)* } #(#traits)* }, quote! { #( impl From<#variants> for #name { fn from(node: #variants) -> #name { #name::#variants(node) } } )* impl AstNode for #name { fn can_cast(kind: SyntaxKind) -> bool { match kind { #(#kinds)|* => true, _ => false, } } fn cast(syntax: SyntaxNode) -> Option { let res = match syntax.kind() { #( #kinds => #name::#variants(#variants { syntax }), )* _ => return None, }; Some(res) } fn syntax(&self) -> &SyntaxNode { match self { #( #name::#variants(it) => &it.syntax, )* } } } }, ) }) .unzip(); let enum_names = grammar.enums.iter().map(|it| &it.name); let node_names = grammar.nodes.iter().map(|it| &it.name); let display_impls = enum_names.chain(node_names.clone()).map(|it| format_ident!("{}", it)).map(|name| { quote! { impl std::fmt::Display for #name { fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { std::fmt::Display::fmt(self.syntax(), f) } } } }); let defined_nodes: HashSet<_> = node_names.collect(); for node in kinds .nodes .iter() .map(|kind| to_pascal_case(kind)) .filter(|name| !defined_nodes.iter().any(|&it| it == name)) { eprintln!("Warning: node {} not defined in ast source", node); } let ast = quote! { use crate::{ SyntaxNode, SyntaxToken, SyntaxKind::{self, *}, ast::{self, AstNode, AstChildren, support}, T, }; #(#node_defs)* #(#enum_defs)* #(#node_boilerplate_impls)* #(#enum_boilerplate_impls)* #(#display_impls)* }; let ast = ast.to_string().replace("T ! [ ", "T![").replace(" ] )", "])"); let mut res = String::with_capacity(ast.len() * 2); let mut docs = grammar.nodes.iter().map(|it| &it.doc).chain(grammar.enums.iter().map(|it| &it.doc)); for chunk in ast.split("# [ pretty_doc_comment_placeholder_workaround ]") { res.push_str(chunk); if let Some(doc) = docs.next() { write_doc_comment(&doc, &mut res); } } let pretty = crate::reformat(res)?; Ok(pretty) } fn write_doc_comment(contents: &[String], dest: &mut String) { for line in contents { writeln!(dest, "///{}", line).unwrap(); } } fn generate_syntax_kinds(grammar: KindsSrc<'_>) -> Result { let (single_byte_tokens_values, single_byte_tokens): (Vec<_>, Vec<_>) = grammar .punct .iter() .filter(|(token, _name)| token.len() == 1) .map(|(token, name)| (token.chars().next().unwrap(), format_ident!("{}", name))) .unzip(); let punctuation_values = grammar.punct.iter().map(|(token, _name)| { if "{}[]()".contains(token) { let c = token.chars().next().unwrap(); quote! { #c } } else { let cs = token.chars().map(|c| Punct::new(c, Spacing::Joint)); quote! { #(#cs)* } } }); let punctuation = grammar.punct.iter().map(|(_token, name)| format_ident!("{}", name)).collect::>(); let full_keywords_values = &grammar.keywords; let full_keywords = full_keywords_values.iter().map(|kw| format_ident!("{}_KW", to_upper_snake_case(&kw))); let all_keywords_values = grammar.keywords.iter().chain(grammar.contextual_keywords.iter()).collect::>(); let all_keywords_idents = all_keywords_values.iter().map(|kw| format_ident!("{}", kw)); let all_keywords = all_keywords_values .iter() .map(|name| format_ident!("{}_KW", to_upper_snake_case(&name))) .collect::>(); let literals = grammar.literals.iter().map(|name| format_ident!("{}", name)).collect::>(); let tokens = grammar.tokens.iter().map(|name| format_ident!("{}", name)).collect::>(); let nodes = grammar.nodes.iter().map(|name| format_ident!("{}", name)).collect::>(); let ast = quote! { #![allow(bad_style, missing_docs, unreachable_pub)] /// The kind of syntax node, e.g. `IDENT`, `USE_KW`, or `STRUCT_DEF`. #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash, Debug)] #[repr(u16)] pub enum SyntaxKind { // Technical SyntaxKinds: they appear temporally during parsing, // but never end up in the final tree #[doc(hidden)] TOMBSTONE, #[doc(hidden)] EOF, #(#punctuation,)* #(#all_keywords,)* #(#literals,)* #(#tokens,)* #(#nodes,)* // Technical kind so that we can cast from u16 safely #[doc(hidden)] __LAST, } use self::SyntaxKind::*; impl SyntaxKind { pub fn is_keyword(self) -> bool { match self { #(#all_keywords)|* => true, _ => false, } } pub fn is_punct(self) -> bool { match self { #(#punctuation)|* => true, _ => false, } } pub fn is_literal(self) -> bool { match self { #(#literals)|* => true, _ => false, } } pub fn from_keyword(ident: &str) -> Option { let kw = match ident { #(#full_keywords_values => #full_keywords,)* _ => return None, }; Some(kw) } pub fn from_char(c: char) -> Option { let tok = match c { #(#single_byte_tokens_values => #single_byte_tokens,)* _ => return None, }; Some(tok) } } #[macro_export] macro_rules! T { #([#punctuation_values] => { $crate::SyntaxKind::#punctuation };)* #([#all_keywords_idents] => { $crate::SyntaxKind::#all_keywords };)* [lifetime] => { $crate::SyntaxKind::LIFETIME }; [ident] => { $crate::SyntaxKind::IDENT }; } }; crate::reformat(ast) } fn to_upper_snake_case(s: &str) -> String { let mut buf = String::with_capacity(s.len()); let mut prev = false; for c in s.chars() { if c.is_ascii_uppercase() && prev { buf.push('_') } prev = true; buf.push(c.to_ascii_uppercase()); } buf } fn to_lower_snake_case(s: &str) -> String { let mut buf = String::with_capacity(s.len()); let mut prev = false; for c in s.chars() { if c.is_ascii_uppercase() && prev { buf.push('_') } prev = true; buf.push(c.to_ascii_lowercase()); } buf } fn to_pascal_case(s: &str) -> String { let mut buf = String::with_capacity(s.len()); let mut prev_is_underscore = true; for c in s.chars() { if c == '_' { prev_is_underscore = true; } else if prev_is_underscore { buf.push(c.to_ascii_uppercase()); prev_is_underscore = false; } else { buf.push(c.to_ascii_lowercase()); } } buf } impl Field { fn is_many(&self) -> bool { matches!(self, Field::Node { src: FieldSrc::Many(_), .. }) } fn token_kind(&self) -> Option { match self { Field::Token(token) => { let token: proc_macro2::TokenStream = token.parse().unwrap(); Some(quote! { T![#token] }) } _ => None, } } fn method_name(&self) -> proc_macro2::Ident { match self { Field::Token(name) => { let name = match name.as_str() { ";" => "semicolon", "->" => "thin_arrow", "'{'" => "l_curly", "'}'" => "r_curly", "'('" => "l_paren", "')'" => "r_paren", "'['" => "l_brack", "']'" => "r_brack", "<" => "l_angle", ">" => "r_angle", "=" => "eq", "!" => "excl", "*" => "star", "&" => "amp", "_" => "underscore", "." => "dot", ".." => "dotdot", "..." => "dotdotdot", "=>" => "fat_arrow", "@" => "at", ":" => "colon", "::" => "coloncolon", "#" => "pound", "?" => "question_mark", _ => name, }; format_ident!("{}_token", name) } Field::Node { name, src } => match src { FieldSrc::Shorthand => format_ident!("{}", to_lower_snake_case(name)), _ => format_ident!("{}", name), }, } } fn ty(&self) -> proc_macro2::Ident { match self { Field::Token(_) => format_ident!("SyntaxToken"), Field::Node { name, src } => match src { FieldSrc::Optional(ty) | FieldSrc::Many(ty) => format_ident!("{}", ty), FieldSrc::Shorthand => format_ident!("{}", name), }, } } }