Add AstElement trait, generate tokens, support tokens in enums

- Adds a new AstElement trait that is implemented by all generated
  node, token and enum structs

- Overhauls the code generators to code-generate all tokens, and
  also enhances enums to support including tokens, node, and nested
  enums
This commit is contained in:
Luca Barbieri 2020-04-03 21:12:08 +02:00 committed by Aleksey Kladov
parent 4762c6d9c6
commit 68196ccc10
3 changed files with 380 additions and 94 deletions

View file

@ -11,7 +11,10 @@ pub mod make;
use std::marker::PhantomData;
use crate::{
syntax_node::{SyntaxNode, SyntaxNodeChildren, SyntaxToken},
syntax_node::{
NodeOrToken, SyntaxElement, SyntaxElementChildren, SyntaxNode, SyntaxNodeChildren,
SyntaxToken,
},
SmolStr, SyntaxKind,
};
@ -30,16 +33,24 @@ pub use self::{
/// conversion itself has zero runtime cost: ast and syntax nodes have exactly
/// the same representation: a pointer to the tree root and a pointer to the
/// node itself.
pub trait AstNode: std::fmt::Display {
pub trait AstNode: AstElement {
fn can_cast(kind: SyntaxKind) -> bool
where
Self: Sized;
fn cast_or_return(syntax: SyntaxNode) -> Result<Self, SyntaxNode>
where
Self: Sized;
fn cast(syntax: SyntaxNode) -> Option<Self>
where
Self: Sized;
Self: Sized,
{
<Self as AstNode>::cast_or_return(syntax).ok()
}
fn syntax(&self) -> &SyntaxNode;
fn into_syntax(self) -> SyntaxNode;
}
#[test]
@ -48,16 +59,51 @@ fn assert_ast_is_object_safe() {
}
/// Like `AstNode`, but wraps tokens rather than interior nodes.
pub trait AstToken {
fn cast(token: SyntaxToken) -> Option<Self>
pub trait AstToken: AstElement {
fn can_cast(token: SyntaxKind) -> bool
where
Self: Sized;
fn cast_or_return(syntax: SyntaxToken) -> Result<Self, SyntaxToken>
where
Self: Sized;
fn cast(syntax: SyntaxToken) -> Option<Self>
where
Self: Sized,
{
<Self as AstToken>::cast_or_return(syntax).ok()
}
fn syntax(&self) -> &SyntaxToken;
fn into_syntax(self) -> SyntaxToken;
fn text(&self) -> &SmolStr {
self.syntax().text()
}
}
/// Like `AstNode`, but wraps either nodes or tokens rather than interior nodes.
pub trait AstElement: std::fmt::Display {
fn can_cast_element(kind: SyntaxKind) -> bool
where
Self: Sized;
fn cast_or_return_element(syntax: SyntaxElement) -> Result<Self, SyntaxElement>
where
Self: Sized;
fn cast_element(syntax: SyntaxElement) -> Option<Self>
where
Self: Sized,
{
<Self as AstElement>::cast_or_return_element(syntax).ok()
}
fn syntax_element(&self) -> NodeOrToken<&SyntaxNode, &SyntaxToken>;
fn into_syntax_element(self) -> SyntaxElement;
}
/// An iterator over `SyntaxNode` children of a particular AST type.
#[derive(Debug, Clone)]
pub struct AstChildren<N> {
@ -86,6 +132,64 @@ fn children<P: AstNode + ?Sized, C: AstNode>(parent: &P) -> AstChildren<C> {
AstChildren::new(parent.syntax())
}
/// An iterator over `SyntaxToken` children of a particular AST type.
#[derive(Debug, Clone)]
pub struct AstChildTokens<N> {
inner: SyntaxElementChildren,
ph: PhantomData<N>,
}
impl<N> AstChildTokens<N> {
fn new(parent: &SyntaxNode) -> Self {
AstChildTokens { inner: parent.children_with_tokens(), ph: PhantomData }
}
}
impl<N: AstToken> Iterator for AstChildTokens<N> {
type Item = N;
fn next(&mut self) -> Option<N> {
self.inner.by_ref().filter_map(|x| x.into_token()).find_map(N::cast)
}
}
fn child_token_opt<P: AstNode + ?Sized, C: AstToken>(parent: &P) -> Option<C> {
child_tokens(parent).next()
}
fn child_tokens<P: AstNode + ?Sized, C: AstToken>(parent: &P) -> AstChildTokens<C> {
AstChildTokens::new(parent.syntax())
}
/// An iterator over `SyntaxNode` children of a particular AST type.
#[derive(Debug, Clone)]
pub struct AstChildElements<N> {
inner: SyntaxElementChildren,
ph: PhantomData<N>,
}
impl<N> AstChildElements<N> {
fn new(parent: &SyntaxNode) -> Self {
AstChildElements { inner: parent.children_with_tokens(), ph: PhantomData }
}
}
impl<N: AstElement> Iterator for AstChildElements<N> {
type Item = N;
fn next(&mut self) -> Option<N> {
self.inner.by_ref().find_map(N::cast_element)
}
}
#[allow(dead_code)]
fn child_element_opt<P: AstNode + ?Sized, C: AstElement>(parent: &P) -> Option<C> {
child_elements(parent).next()
}
#[allow(dead_code)]
fn child_elements<P: AstNode + ?Sized, C: AstElement>(parent: &P) -> AstChildElements<C> {
AstChildElements::new(parent.syntax())
}
#[test]
fn test_doc_comment_none() {
let file = SourceFile::parse(

View file

@ -1,26 +1,10 @@
//! There are many AstNodes, but only a few tokens, so we hand-write them here.
use crate::{
ast::AstToken,
SyntaxKind::{COMMENT, RAW_STRING, STRING, WHITESPACE},
SyntaxToken, TextRange, TextUnit,
ast::{AstToken, Comment, RawString, String, Whitespace},
TextRange, TextUnit,
};
#[derive(Debug, Clone, PartialEq, Eq, Hash)]
pub struct Comment(SyntaxToken);
impl AstToken for Comment {
fn cast(token: SyntaxToken) -> Option<Self> {
match token.kind() {
COMMENT => Some(Comment(token)),
_ => None,
}
}
fn syntax(&self) -> &SyntaxToken {
&self.0
}
}
impl Comment {
pub fn kind(&self) -> CommentKind {
kind_by_prefix(self.text())
@ -89,20 +73,6 @@ fn prefix_by_kind(kind: CommentKind) -> &'static str {
unreachable!()
}
pub struct Whitespace(SyntaxToken);
impl AstToken for Whitespace {
fn cast(token: SyntaxToken) -> Option<Self> {
match token.kind() {
WHITESPACE => Some(Whitespace(token)),
_ => None,
}
}
fn syntax(&self) -> &SyntaxToken {
&self.0
}
}
impl Whitespace {
pub fn spans_multiple_lines(&self) -> bool {
let text = self.text();
@ -168,20 +138,6 @@ pub trait HasStringValue: HasQuotes {
fn value(&self) -> Option<std::string::String>;
}
pub struct String(SyntaxToken);
impl AstToken for String {
fn cast(token: SyntaxToken) -> Option<Self> {
match token.kind() {
STRING => Some(String(token)),
_ => None,
}
}
fn syntax(&self) -> &SyntaxToken {
&self.0
}
}
impl HasStringValue for String {
fn value(&self) -> Option<std::string::String> {
let text = self.text().as_str();
@ -201,20 +157,6 @@ impl HasStringValue for String {
}
}
pub struct RawString(SyntaxToken);
impl AstToken for RawString {
fn cast(token: SyntaxToken) -> Option<Self> {
match token.kind() {
RAW_STRING => Some(RawString(token)),
_ => None,
}
}
fn syntax(&self) -> &SyntaxToken {
&self.0
}
}
impl HasStringValue for RawString {
fn value(&self) -> Option<std::string::String> {
let text = self.text().as_str();

View file

@ -5,6 +5,8 @@
use proc_macro2::{Punct, Spacing};
use quote::{format_ident, quote};
use std::borrow::Cow;
use std::collections::{BTreeSet, HashMap, HashSet};
use crate::{
ast_src::{AstSrc, FieldSrc, KindsSrc, AST_SRC, KINDS_SRC},
@ -18,13 +20,125 @@ pub fn generate_syntax(mode: Mode) -> Result<()> {
update(syntax_kinds_file.as_path(), &syntax_kinds, mode)?;
let ast_file = project_root().join(codegen::AST);
let ast = generate_ast(AST_SRC)?;
let ast = generate_ast(KINDS_SRC, AST_SRC)?;
update(ast_file.as_path(), &ast, mode)?;
Ok(())
}
fn generate_ast(grammar: AstSrc<'_>) -> Result<String> {
#[derive(Debug, Default, Clone)]
struct ElementKinds {
kinds: BTreeSet<proc_macro2::Ident>,
has_nodes: bool,
has_tokens: bool,
}
fn generate_ast(kinds: KindsSrc<'_>, grammar: AstSrc<'_>) -> Result<String> {
let all_token_kinds: Vec<_> = kinds
.punct
.into_iter()
.map(|(_, kind)| kind)
.copied()
.map(|x| x.into())
.chain(
kinds
.keywords
.into_iter()
.chain(kinds.contextual_keywords.into_iter())
.map(|name| Cow::Owned(format!("{}_KW", to_upper_snake_case(&name)))),
)
.chain(kinds.literals.into_iter().copied().map(|x| x.into()))
.chain(kinds.tokens.into_iter().copied().map(|x| x.into()))
.collect();
let mut element_kinds_map = HashMap::new();
for kind in &all_token_kinds {
let kind = &**kind;
let name = to_pascal_case(kind);
element_kinds_map.insert(
name,
ElementKinds {
kinds: Some(format_ident!("{}", kind)).into_iter().collect(),
has_nodes: false,
has_tokens: true,
},
);
}
for kind in kinds.nodes {
let name = to_pascal_case(kind);
element_kinds_map.insert(
name,
ElementKinds {
kinds: Some(format_ident!("{}", *kind)).into_iter().collect(),
has_nodes: true,
has_tokens: false,
},
);
}
for en in grammar.enums {
let mut element_kinds: ElementKinds = Default::default();
for variant in en.variants {
if let Some(variant_element_kinds) = element_kinds_map.get(*variant) {
element_kinds.kinds.extend(variant_element_kinds.kinds.iter().cloned());
element_kinds.has_tokens |= variant_element_kinds.has_tokens;
element_kinds.has_nodes |= variant_element_kinds.has_nodes;
} else {
panic!("Enum variant has type that does not exist or was not declared before the enum: {}", *variant);
}
}
element_kinds_map.insert(en.name.to_string(), element_kinds);
}
let tokens = all_token_kinds.iter().map(|kind_str| {
let kind_str = &**kind_str;
let kind = format_ident!("{}", kind_str);
let name = format_ident!("{}", to_pascal_case(kind_str));
quote! {
#[derive(Debug, Clone, PartialEq, Eq, Hash)]
pub struct #name(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 {
match kind {
#kind => true,
_ => false,
}
}
fn cast_or_return(syntax: SyntaxToken) -> Result<Self, SyntaxToken> {
if Self::can_cast(syntax.kind()) { Ok(Self(syntax)) } else { Err(syntax) }
}
fn syntax(&self) -> &SyntaxToken { &self.0 }
fn into_syntax(self) -> SyntaxToken { self.0 }
}
impl AstElement for #name {
fn can_cast_element(kind: SyntaxKind) -> bool {
match kind {
#kind => true,
_ => false,
}
}
fn cast_or_return_element(syntax: SyntaxElement) -> Result<Self, SyntaxElement> {
if Self::can_cast_element(syntax.kind()) { Ok(Self(syntax.into_token().unwrap())) } else { Err(syntax) }
}
fn syntax_element(&self) -> NodeOrToken<&SyntaxNode, &SyntaxToken> {
NodeOrToken::Token(&self.0)
}
fn into_syntax_element(self) -> SyntaxElement {
NodeOrToken::Token(self.0)
}
}
}
});
let nodes = grammar.nodes.iter().map(|node| {
let name = format_ident!("{}", node.name);
let kind = format_ident!("{}", to_upper_snake_case(&name.to_string()));
@ -42,20 +156,28 @@ fn generate_ast(grammar: AstSrc<'_>) -> Result<String> {
FieldSrc::Optional(ty) | FieldSrc::Many(ty) => ty,
FieldSrc::Shorthand => name,
};
let element_kinds = &element_kinds_map.get(*ty).unwrap_or_else(|| panic!("type not found: {}", *ty));
let iter = if !element_kinds.has_tokens {
format_ident!("AstChildren")
} else if !element_kinds.has_nodes {
format_ident!("AstChildTokens")
} else {
format_ident!("AstChildElements")
};
let ty = format_ident!("{}", ty);
match field {
FieldSrc::Many(_) => {
quote! {
pub fn #method_name(&self) -> AstChildren<#ty> {
AstChildren::new(&self.syntax)
pub fn #method_name(&self) -> #iter<#ty> {
#iter::new(&self.syntax)
}
}
}
FieldSrc::Optional(_) | FieldSrc::Shorthand => {
quote! {
pub fn #method_name(&self) -> Option<#ty> {
AstChildren::new(&self.syntax).next()
#iter::new(&self.syntax).next()
}
}
}
@ -81,11 +203,31 @@ fn generate_ast(grammar: AstSrc<'_>) -> Result<String> {
_ => false,
}
}
fn cast(syntax: SyntaxNode) -> Option<Self> {
if Self::can_cast(syntax.kind()) { Some(Self { syntax }) } else { None }
fn cast_or_return(syntax: SyntaxNode) -> Result<Self, SyntaxNode> {
if Self::can_cast(syntax.kind()) { Ok(Self { syntax }) } else { Err(syntax) }
}
fn syntax(&self) -> &SyntaxNode { &self.syntax }
fn into_syntax(self) -> SyntaxNode { self.syntax }
}
impl AstElement for #name {
fn can_cast_element(kind: SyntaxKind) -> bool {
match kind {
#kind => true,
_ => false,
}
}
fn cast_or_return_element(syntax: SyntaxElement) -> Result<Self, SyntaxElement> {
if Self::can_cast_element(syntax.kind()) { Ok(Self { syntax: syntax.into_node().unwrap() }) } else { Err(syntax) }
}
fn syntax_element(&self) -> NodeOrToken<&SyntaxNode, &SyntaxToken> {
NodeOrToken::Node(&self.syntax)
}
fn into_syntax_element(self) -> SyntaxElement {
NodeOrToken::Node(self.syntax)
}
}
#(#traits)*
impl #name {
@ -96,16 +238,71 @@ fn generate_ast(grammar: AstSrc<'_>) -> Result<String> {
let enums = grammar.enums.iter().map(|en| {
let variants = en.variants.iter().map(|var| format_ident!("{}", var)).collect::<Vec<_>>();
let element_kinds = &element_kinds_map[&en.name.to_string()];
let name = format_ident!("{}", en.name);
let kinds = variants
let kinds = en.variants
.iter()
.map(|name| format_ident!("{}", to_upper_snake_case(&name.to_string())))
.map(|name| {
element_kinds_map[*name].kinds.iter().collect::<Vec<_>>()
})
.collect::<Vec<_>>();
let traits = en.traits.iter().map(|trait_name| {
let trait_name = format_ident!("{}", trait_name);
quote!(impl ast::#trait_name for #name {})
});
let all_kinds = &element_kinds.kinds;
let specific_ast_trait = if element_kinds.has_nodes != element_kinds.has_tokens {
let (ast_trait, syntax_type) = if element_kinds.has_tokens {
(
quote!(AstToken),
quote!(SyntaxToken),
)
} else {
(
quote!(AstNode),
quote!(SyntaxNode),
)
};
quote! {
impl #ast_trait for #name {
fn can_cast(kind: SyntaxKind) -> bool {
match kind {
#(#all_kinds)|* => true,
_ => false,
}
}
#[allow(unreachable_patterns)]
fn cast_or_return(syntax: #syntax_type) -> Result<Self, #syntax_type> {
match syntax.kind() {
#(
#(#kinds)|* => #variants::cast_or_return(syntax).map(|x| #name::#variants(x)),
)*
_ => Err(syntax),
}
}
fn syntax(&self) -> &#syntax_type {
match self {
#(
#name::#variants(it) => it.syntax(),
)*
}
}
fn into_syntax(self) -> #syntax_type {
match self {
#(
#name::#variants(it) => it.into_syntax(),
)*
}
}
}
}
} else {
Default::default()
};
quote! {
#[derive(Debug, Clone, PartialEq, Eq, Hash)]
pub enum #name {
@ -122,44 +319,71 @@ fn generate_ast(grammar: AstSrc<'_>) -> Result<String> {
impl std::fmt::Display for #name {
fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
std::fmt::Display::fmt(self.syntax(), f)
match self {
#(
#name::#variants(it) => std::fmt::Display::fmt(it, f),
)*
}
}
}
impl AstNode for #name {
fn can_cast(kind: SyntaxKind) -> bool {
#specific_ast_trait
impl AstElement for #name {
fn can_cast_element(kind: SyntaxKind) -> bool {
match kind {
#(#kinds)|* => true,
#(#all_kinds)|* => true,
_ => false,
}
}
fn cast(syntax: SyntaxNode) -> Option<Self> {
let res = match syntax.kind() {
#[allow(unreachable_patterns)]
fn cast_or_return_element(syntax: SyntaxElement) -> Result<Self, SyntaxElement> {
match syntax.kind() {
#(
#kinds => #name::#variants(#variants { syntax }),
#(#kinds)|* => #variants::cast_or_return_element(syntax).map(|x| #name::#variants(x)),
)*
_ => return None,
};
Some(res)
_ => Err(syntax),
}
}
fn syntax(&self) -> &SyntaxNode {
fn syntax_element(&self) -> NodeOrToken<&SyntaxNode, &SyntaxToken> {
match self {
#(
#name::#variants(it) => &it.syntax,
#name::#variants(it) => it.syntax_element(),
)*
}
}
fn into_syntax_element(self) -> SyntaxElement {
match self {
#(
#name::#variants(it) => it.into_syntax_element(),
)*
}
}
}
#(#traits)*
}
});
let defined_nodes: HashSet<_> = grammar.nodes.iter().map(|node| node.name).collect();
for node in kinds
.nodes
.iter()
.map(|kind| to_pascal_case(*kind))
.filter(|name| !defined_nodes.contains(&**name))
{
eprintln!("Warning: node {} not defined in ast source", node);
}
let ast = quote! {
#[allow(unused_imports)]
use crate::{
SyntaxNode, SyntaxKind::{self, *},
ast::{self, AstNode, AstChildren},
SyntaxNode, SyntaxToken, SyntaxElement, NodeOrToken, SyntaxKind::{self, *},
ast::{self, AstNode, AstToken, AstElement, AstChildren, AstChildTokens, AstChildElements},
};
#(#tokens)*
#(#nodes)*
#(#enums)*
};
@ -282,12 +506,12 @@ fn generate_syntax_kinds(grammar: KindsSrc<'_>) -> Result<String> {
fn to_upper_snake_case(s: &str) -> String {
let mut buf = String::with_capacity(s.len());
let mut prev_is_upper = None;
let mut prev = false;
for c in s.chars() {
if c.is_ascii_uppercase() && prev_is_upper == Some(false) {
if c.is_ascii_uppercase() && prev {
buf.push('_')
}
prev_is_upper = Some(c.is_ascii_uppercase());
prev = true;
buf.push(c.to_ascii_uppercase());
}
@ -296,14 +520,30 @@ fn to_upper_snake_case(s: &str) -> String {
fn to_lower_snake_case(s: &str) -> String {
let mut buf = String::with_capacity(s.len());
let mut prev_is_upper = None;
let mut prev = false;
for c in s.chars() {
if c.is_ascii_uppercase() && prev_is_upper == Some(false) {
if c.is_ascii_uppercase() && prev {
buf.push('_')
}
prev_is_upper = Some(c.is_ascii_uppercase());
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
}