1 files changed, 264 insertions, 26 deletions

diff --git a/xtask/src/codegen/gen_syntax.rs b/xtask/src/codegen/gen_syntax.rs
index 745a25862..45b788bdb 100644
--- a/xtask/src/codegen/gen_syntax.rs
+++ b/xtask/src/codegen/gen_syntax.rs
@@ -3,34 +3,43 @@
 //! 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 std::{
+    collections::{BTreeSet, HashSet},
+    fmt::Write,
+};
 
 use proc_macro2::{Punct, Spacing};
 use quote::{format_ident, quote};
+use ungrammar::{Grammar, Rule};
 
 use crate::{
-    ast_src::{AstSrc, Field, FieldSrc, KindsSrc, AST_SRC, KINDS_SRC},
+    ast_src::{AstEnumSrc, AstNodeSrc, AstSrc, Cardinality, Field, KindsSrc, KINDS_SRC},
     codegen::{self, update, Mode},
     project_root, Result,
 };
 
 pub fn generate_syntax(mode: Mode) -> Result<()> {
+    let grammar = include_str!("rust.ungram")
+        .parse::<Grammar>()
+        .unwrap_or_else(|err| panic!("\n    \x1b[91merror\x1b[0m: {}\n", err));
+    let ast = lower(&grammar);
+
     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_SRC)?;
+    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_SRC)?;
+    let contents = generate_nodes(KINDS_SRC, &ast)?;
     update(ast_nodes_file.as_path(), &contents, mode)?;
 
     Ok(())
 }
 
-fn generate_tokens(grammar: AstSrc<'_>) -> Result<String> {
+fn generate_tokens(grammar: &AstSrc) -> Result<String> {
     let tokens = grammar.tokens.iter().map(|token| {
         let name = format_ident!("{}", token);
         let kind = format_ident!("{}", to_upper_snake_case(token));
@@ -62,13 +71,13 @@ fn generate_tokens(grammar: AstSrc<'_>) -> Result<String> {
     Ok(pretty)
 }
 
-fn generate_nodes(kinds: KindsSrc<'_>, grammar: AstSrc<'_>) -> Result<String> {
+fn generate_nodes(kinds: KindsSrc<'_>, grammar: &AstSrc) -> Result<String> {
     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 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 {})
@@ -192,8 +201,8 @@ fn generate_nodes(kinds: KindsSrc<'_>, grammar: AstSrc<'_>) -> Result<String> {
         })
         .unzip();
 
-    let enum_names = grammar.enums.iter().map(|it| it.name);
-    let node_names = grammar.nodes.iter().map(|it| it.name);
+    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| {
@@ -212,9 +221,11 @@ fn generate_nodes(kinds: KindsSrc<'_>, grammar: AstSrc<'_>) -> Result<String> {
         .nodes
         .iter()
         .map(|kind| to_pascal_case(kind))
-        .filter(|name| !defined_nodes.contains(name.as_str()))
+        .filter(|name| !defined_nodes.iter().any(|&it| it == name))
     {
-        eprintln!("Warning: node {} not defined in ast source", node);
+        drop(node)
+        // TODO: restore this
+        // eprintln!("Warning: node {} not defined in ast source", node);
     }
 
     let ast = quote! {
@@ -236,12 +247,12 @@ fn generate_nodes(kinds: KindsSrc<'_>, grammar: AstSrc<'_>) -> Result<String> {
     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));
+        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);
+            write_doc_comment(&doc, &mut res);
         }
     }
 
@@ -249,7 +260,7 @@ fn generate_nodes(kinds: KindsSrc<'_>, grammar: AstSrc<'_>) -> Result<String> {
     Ok(pretty)
 }
 
-fn write_doc_comment(contents: &[&str], dest: &mut String) {
+fn write_doc_comment(contents: &[String], dest: &mut String) {
     for line in contents {
         writeln!(dest, "///{}", line).unwrap();
     }
@@ -296,7 +307,7 @@ fn generate_syntax_kinds(grammar: KindsSrc<'_>) -> Result<String> {
 
     let ast = quote! {
         #![allow(bad_style, missing_docs, unreachable_pub)]
-        /// The kind of syntax node, e.g. `IDENT`, `USE_KW`, or `STRUCT_DEF`.
+        /// The kind of syntax node, e.g. `IDENT`, `USE_KW`, or `STRUCT`.
         #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash, Debug)]
         #[repr(u16)]
         pub enum SyntaxKind {
@@ -363,6 +374,7 @@ fn generate_syntax_kinds(grammar: KindsSrc<'_>) -> Result<String> {
             #([#all_keywords_idents] => { $crate::SyntaxKind::#all_keywords };)*
             [lifetime] => { $crate::SyntaxKind::LIFETIME };
             [ident] => { $crate::SyntaxKind::IDENT };
+            [shebang] => { $crate::SyntaxKind::SHEBANG };
         }
     };
 
@@ -413,9 +425,13 @@ fn to_pascal_case(s: &str) -> String {
     buf
 }
 
-impl Field<'_> {
+fn pluralize(s: &str) -> String {
+    format!("{}s", s)
+}
+
+impl Field {
     fn is_many(&self) -> bool {
-        matches!(self, Field::Node { src: FieldSrc::Many(_), .. })
+        matches!(self, Field::Node { cardinality: Cardinality::Many, .. })
     }
     fn token_kind(&self) -> Option<proc_macro2::TokenStream> {
         match self {
@@ -429,7 +445,7 @@ impl Field<'_> {
     fn method_name(&self) -> proc_macro2::Ident {
         match self {
             Field::Token(name) => {
-                let name = match *name {
+                let name = match name.as_str() {
                     ";" => "semicolon",
                     "->" => "thin_arrow",
                     "'{'" => "l_curly",
@@ -448,29 +464,251 @@ impl Field<'_> {
                     "." => "dot",
                     ".." => "dotdot",
                     "..." => "dotdotdot",
+                    "..=" => "dotdoteq",
                     "=>" => "fat_arrow",
                     "@" => "at",
                     ":" => "colon",
                     "::" => "coloncolon",
                     "#" => "pound",
                     "?" => "question_mark",
+                    "," => "comma",
                     _ => name,
                 };
                 format_ident!("{}_token", name)
             }
-            Field::Node { name, src } => match src {
-                FieldSrc::Shorthand => format_ident!("{}", to_lower_snake_case(name)),
-                _ => format_ident!("{}", name),
-            },
+            Field::Node { 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),
-            },
+            Field::Node { ty, .. } => format_ident!("{}", ty),
         }
     }
 }
+
+fn lower(grammar: &Grammar) -> AstSrc {
+    let mut res = AstSrc::default();
+    res.tokens = vec!["Whitespace".into(), "Comment".into(), "String".into(), "RawString".into()];
+
+    let nodes = grammar
+        .iter()
+        .filter(|&node| match grammar[node].rule {
+            Rule::Node(it) if it == node => false,
+            _ => true,
+        })
+        .collect::<Vec<_>>();
+
+    for &node in &nodes {
+        let name = grammar[node].name.clone();
+        let rule = &grammar[node].rule;
+        match lower_enum(grammar, rule) {
+            Some(variants) => {
+                let enum_src = AstEnumSrc { doc: Vec::new(), name, traits: Vec::new(), variants };
+                res.enums.push(enum_src);
+            }
+            None => {
+                let mut fields = Vec::new();
+                lower_rule(&mut fields, grammar, None, rule);
+                res.nodes.push(AstNodeSrc { doc: Vec::new(), name, traits: Vec::new(), fields });
+            }
+        }
+    }
+
+    deduplicate_fields(&mut res);
+    extract_enums(&mut res);
+    extract_struct_traits(&mut res);
+    extract_enum_traits(&mut res);
+    res
+}
+
+fn lower_enum(grammar: &Grammar, rule: &Rule) -> Option<Vec<String>> {
+    let alternatives = match rule {
+        Rule::Alt(it) => it,
+        _ => return None,
+    };
+    let mut variants = Vec::new();
+    for alternative in alternatives {
+        match alternative {
+            Rule::Node(it) => variants.push(grammar[*it].name.clone()),
+            _ => return None,
+        }
+    }
+    Some(variants)
+}
+
+fn lower_rule(acc: &mut Vec<Field>, grammar: &Grammar, label: Option<&String>, rule: &Rule) {
+    if lower_comma_list(acc, grammar, label, rule) {
+        return;
+    }
+
+    match rule {
+        Rule::Node(node) => {
+            let ty = grammar[*node].name.clone();
+            let name = label.cloned().unwrap_or_else(|| to_lower_snake_case(&ty));
+            let field = Field::Node { name, ty, cardinality: Cardinality::Optional };
+            acc.push(field);
+        }
+        Rule::Token(token) => {
+            assert!(label.is_none());
+            let mut name = grammar[*token].name.clone();
+            if name != "int_number" && name != "string" {
+                if "[]{}()".contains(&name) {
+                    name = format!("'{}'", name);
+                }
+                let field = Field::Token(name);
+                acc.push(field);
+            }
+        }
+        Rule::Rep(inner) => {
+            if let Rule::Node(node) = &**inner {
+                let ty = grammar[*node].name.clone();
+                let name = label.cloned().unwrap_or_else(|| pluralize(&to_lower_snake_case(&ty)));
+                let field = Field::Node { name, ty, cardinality: Cardinality::Many };
+                acc.push(field);
+                return;
+            }
+            todo!("{:?}", rule)
+        }
+        Rule::Labeled { label: l, rule } => {
+            assert!(label.is_none());
+            lower_rule(acc, grammar, Some(l), rule);
+        }
+        Rule::Seq(rules) | Rule::Alt(rules) => {
+            for rule in rules {
+                lower_rule(acc, grammar, label, rule)
+            }
+        }
+        Rule::Opt(rule) => lower_rule(acc, grammar, label, rule),
+    }
+}
+
+// (T (',' T)* ','?)
+fn lower_comma_list(
+    acc: &mut Vec<Field>,
+    grammar: &Grammar,
+    label: Option<&String>,
+    rule: &Rule,
+) -> bool {
+    let rule = match rule {
+        Rule::Seq(it) => it,
+        _ => return false,
+    };
+    let (node, repeat, trailing_comma) = match rule.as_slice() {
+        [Rule::Node(node), Rule::Rep(repeat), Rule::Opt(trailing_comma)] => {
+            (node, repeat, trailing_comma)
+        }
+        _ => return false,
+    };
+    let repeat = match &**repeat {
+        Rule::Seq(it) => it,
+        _ => return false,
+    };
+    match repeat.as_slice() {
+        [comma, Rule::Node(n)] if comma == &**trailing_comma && n == node => (),
+        _ => return false,
+    }
+    let ty = grammar[*node].name.clone();
+    let name = label.cloned().unwrap_or_else(|| pluralize(&to_lower_snake_case(&ty)));
+    let field = Field::Node { name, ty, cardinality: Cardinality::Many };
+    acc.push(field);
+    true
+}
+
+fn deduplicate_fields(ast: &mut AstSrc) {
+    for node in &mut ast.nodes {
+        let mut i = 0;
+        'outer: while i < node.fields.len() {
+            for j in 0..i {
+                let f1 = &node.fields[i];
+                let f2 = &node.fields[j];
+                if f1 == f2 {
+                    node.fields.remove(i);
+                    continue 'outer;
+                }
+            }
+            i += 1;
+        }
+    }
+}
+
+fn extract_enums(ast: &mut AstSrc) {
+    for node in &mut ast.nodes {
+        for enm in &ast.enums {
+            let mut to_remove = Vec::new();
+            for (i, field) in node.fields.iter().enumerate() {
+                let ty = field.ty().to_string();
+                if enm.variants.iter().any(|it| it == &ty) {
+                    to_remove.push(i);
+                }
+            }
+            if to_remove.len() == enm.variants.len() {
+                node.remove_field(to_remove);
+                let ty = enm.name.clone();
+                let name = to_lower_snake_case(&ty);
+                node.fields.push(Field::Node { name, ty, cardinality: Cardinality::Optional });
+            }
+        }
+    }
+}
+
+fn extract_struct_traits(ast: &mut AstSrc) {
+    let traits: &[(&str, &[&str])] = &[
+        ("AttrsOwner", &["attrs"]),
+        ("NameOwner", &["name"]),
+        ("VisibilityOwner", &["visibility"]),
+        ("GenericParamsOwner", &["generic_param_list", "where_clause"]),
+        ("TypeBoundsOwner", &["type_bound_list", "colon_token"]),
+        ("ModuleItemOwner", &["items"]),
+        ("LoopBodyOwner", &["label", "loop_body"]),
+        ("ArgListOwner", &["arg_list"]),
+    ];
+
+    for node in &mut ast.nodes {
+        for (name, methods) in traits {
+            extract_struct_trait(node, name, methods);
+        }
+    }
+}
+
+fn extract_struct_trait(node: &mut AstNodeSrc, trait_name: &str, methods: &[&str]) {
+    let mut to_remove = Vec::new();
+    for (i, field) in node.fields.iter().enumerate() {
+        let method_name = field.method_name().to_string();
+        if methods.iter().any(|&it| it == &method_name) {
+            to_remove.push(i);
+        }
+    }
+    if to_remove.len() == methods.len() {
+        node.traits.push(trait_name.to_string());
+        node.remove_field(to_remove);
+    }
+}
+
+fn extract_enum_traits(ast: &mut AstSrc) {
+    for enm in &mut ast.enums {
+        let nodes = &ast.nodes;
+        let mut variant_traits = enm
+            .variants
+            .iter()
+            .map(|var| nodes.iter().find(|it| &it.name == var).unwrap())
+            .map(|node| node.traits.iter().cloned().collect::<BTreeSet<_>>());
+
+        let mut enum_traits = match variant_traits.next() {
+            Some(it) => it,
+            None => continue,
+        };
+        for traits in variant_traits {
+            enum_traits = enum_traits.intersection(&traits).cloned().collect();
+        }
+        enm.traits = enum_traits.into_iter().collect();
+    }
+}
+
+impl AstNodeSrc {
+    fn remove_field(&mut self, to_remove: Vec<usize>) {
+        to_remove.into_iter().rev().for_each(|idx| {
+            self.fields.remove(idx);
+        });
+    }
+}