Auto merge of #17024 - roife:fix-issue-16980, r=Veykril

fix: handle escaped chars in doc comments

fix #16980.

For `ast::LiteralKind::String`, store the original string value.
This commit is contained in:
bors 2024-04-19 13:19:22 +00:00
commit 05428c5640
5 changed files with 68 additions and 14 deletions

View file

@ -5,7 +5,7 @@ pub mod builtin;
#[cfg(test)]
mod tests;
use std::{hash::Hash, ops, slice::Iter as SliceIter};
use std::{borrow::Cow, hash::Hash, ops, slice::Iter as SliceIter};
use base_db::CrateId;
use cfg::{CfgExpr, CfgOptions};
@ -573,6 +573,10 @@ impl<'attr> AttrQuery<'attr> {
self.attrs().find_map(|attr| attr.string_value())
}
pub fn string_value_unescape(self) -> Option<Cow<'attr, str>> {
self.attrs().find_map(|attr| attr.string_value_unescape())
}
pub fn exists(self) -> bool {
self.attrs().next().is_some()
}

View file

@ -1917,7 +1917,7 @@ impl ModCollector<'_, '_> {
}
fn collect_module(&mut self, module_id: FileItemTreeId<Mod>, attrs: &Attrs) {
let path_attr = attrs.by_key("path").string_value();
let path_attr = attrs.by_key("path").string_value_unescape();
let is_macro_use = attrs.by_key("macro_use").exists();
let module = &self.item_tree[module_id];
match &module.kind {
@ -1931,7 +1931,8 @@ impl ModCollector<'_, '_> {
module_id,
);
let Some(mod_dir) = self.mod_dir.descend_into_definition(&module.name, path_attr)
let Some(mod_dir) =
self.mod_dir.descend_into_definition(&module.name, path_attr.as_deref())
else {
return;
};
@ -1952,8 +1953,12 @@ impl ModCollector<'_, '_> {
ModKind::Outline => {
let ast_id = AstId::new(self.file_id(), module.ast_id);
let db = self.def_collector.db;
match self.mod_dir.resolve_declaration(db, self.file_id(), &module.name, path_attr)
{
match self.mod_dir.resolve_declaration(
db,
self.file_id(),
&module.name,
path_attr.as_deref(),
) {
Ok((file_id, is_mod_rs, mod_dir)) => {
let item_tree = db.file_item_tree(file_id.into());
let krate = self.def_collector.def_map.krate;

View file

@ -1,5 +1,5 @@
//! A higher level attributes based on TokenTree, with also some shortcuts.
use std::{fmt, ops};
use std::{borrow::Cow, fmt, ops};
use base_db::CrateId;
use cfg::CfgExpr;
@ -8,6 +8,7 @@ use intern::Interned;
use mbe::{syntax_node_to_token_tree, DelimiterKind, Punct};
use smallvec::{smallvec, SmallVec};
use span::{Span, SyntaxContextId};
use syntax::unescape;
use syntax::{ast, format_smolstr, match_ast, AstNode, AstToken, SmolStr, SyntaxNode};
use triomphe::ThinArc;
@ -54,8 +55,7 @@ impl RawAttrs {
Attr {
id,
input: Some(Interned::new(AttrInput::Literal(tt::Literal {
// FIXME: Escape quotes from comment content
text: SmolStr::new(format_smolstr!("\"{doc}\"",)),
text: SmolStr::new(format_smolstr!("\"{}\"", Self::escape_chars(doc))),
span,
}))),
path: Interned::new(ModPath::from(crate::name!(doc))),
@ -74,6 +74,10 @@ impl RawAttrs {
RawAttrs { entries }
}
fn escape_chars(s: &str) -> String {
s.replace('\\', r#"\\"#).replace('"', r#"\""#)
}
pub fn from_attrs_owner(
db: &dyn ExpandDatabase,
owner: InFile<&dyn ast::HasAttrs>,
@ -297,6 +301,18 @@ impl Attr {
}
}
pub fn string_value_unescape(&self) -> Option<Cow<'_, str>> {
match self.input.as_deref()? {
AttrInput::Literal(it) => match it.text.strip_prefix('r') {
Some(it) => {
it.trim_matches('#').strip_prefix('"')?.strip_suffix('"').map(Cow::Borrowed)
}
None => it.text.strip_prefix('"')?.strip_suffix('"').and_then(unescape),
},
_ => None,
}
}
/// #[path(ident)]
pub fn single_ident_value(&self) -> Option<&tt::Ident> {
match self.input.as_deref()? {
@ -346,6 +362,33 @@ impl Attr {
}
}
fn unescape(s: &str) -> Option<Cow<'_, str>> {
let mut buf = String::new();
let mut prev_end = 0;
let mut has_error = false;
unescape::unescape_unicode(s, unescape::Mode::Str, &mut |char_range, unescaped_char| match (
unescaped_char,
buf.capacity() == 0,
) {
(Ok(c), false) => buf.push(c),
(Ok(_), true) if char_range.len() == 1 && char_range.start == prev_end => {
prev_end = char_range.end
}
(Ok(c), true) => {
buf.reserve_exact(s.len());
buf.push_str(&s[..prev_end]);
buf.push(c);
}
(Err(_), _) => has_error = true,
});
match (has_error, buf.capacity() == 0) {
(true, _) => None,
(false, false) => Some(Cow::Owned(buf)),
(false, true) => Some(Cow::Borrowed(s)),
}
}
pub fn collect_attrs(
owner: &dyn ast::HasAttrs,
) -> impl Iterator<Item = (AttrId, Either<ast::Attr, ast::Comment>)> {

View file

@ -91,8 +91,10 @@ pub fn docs_with_rangemap(
db: &dyn DefDatabase,
attrs: &AttrsWithOwner,
) -> Option<(Documentation, DocsRangeMap)> {
let docs =
attrs.by_key("doc").attrs().filter_map(|attr| attr.string_value().map(|s| (s, attr.id)));
let docs = attrs
.by_key("doc")
.attrs()
.filter_map(|attr| attr.string_value_unescape().map(|s| (s, attr.id)));
let indent = doc_indent(attrs);
let mut buf = String::new();
let mut mapping = Vec::new();
@ -132,7 +134,7 @@ pub fn docs_with_rangemap(
}
pub fn docs_from_attrs(attrs: &hir::Attrs) -> Option<String> {
let docs = attrs.by_key("doc").attrs().filter_map(|attr| attr.string_value());
let docs = attrs.by_key("doc").attrs().filter_map(|attr| attr.string_value_unescape());
let indent = doc_indent(attrs);
let mut buf = String::new();
for doc in docs {
@ -270,10 +272,9 @@ fn doc_indent(attrs: &hir::Attrs) -> usize {
attrs
.by_key("doc")
.attrs()
.filter_map(|attr| attr.string_value())
.filter_map(|attr| attr.string_value()) // no need to use unescape version here
.flat_map(|s| s.lines())
.filter(|line| !line.chars().all(|c| c.is_whitespace()))
.map(|line| line.chars().take_while(|c| c.is_whitespace()).count())
.filter_map(|line| line.chars().position(|c| !c.is_whitespace()))
.min()
.unwrap_or(0)
}

View file

@ -65,6 +65,7 @@ pub use rowan::{
api::Preorder, Direction, GreenNode, NodeOrToken, SyntaxText, TextRange, TextSize,
TokenAtOffset, WalkEvent,
};
pub use rustc_lexer::unescape;
pub use smol_str::{format_smolstr, SmolStr};
/// `Parse` is the result of the parsing: a syntax tree and a collection of