Auto merge of #88781 - estebank:emoji-idents, r=oli-obk

Tokenize emoji as if they were valid identifiers

In the lexer, consider emojis to be valid identifiers and reject
them later to avoid knock down parse errors.

Partially address #86102.

Author: bors

Cargo.lock

@@ -4040,6 +4040,7 @@ name = "rustc_lexer"
 version = "0.1.0"
 dependencies = [
  "expect-test",
+ "unic-emoji-char",
  "unicode-xid",
 ]
 
@@ -5510,6 +5511,47 @@ version = "0.1.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "56dee185309b50d1f11bfedef0fe6d036842e3fb77413abef29f8f8d1c5d4c1c"
 
+[[package]]
+name = "unic-char-property"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a8c57a407d9b6fa02b4795eb81c5b6652060a15a7903ea981f3d723e6c0be221"
+dependencies = [
+ "unic-char-range",
+]
+
+[[package]]
+name = "unic-char-range"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0398022d5f700414f6b899e10b8348231abf9173fa93144cbc1a43b9793c1fbc"
+
+[[package]]
+name = "unic-common"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "80d7ff825a6a654ee85a63e80f92f054f904f21e7d12da4e22f9834a4aaa35bc"
+
+[[package]]
+name = "unic-emoji-char"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0b07221e68897210270a38bde4babb655869637af0f69407f96053a34f76494d"
+dependencies = [
+ "unic-char-property",
+ "unic-char-range",
+ "unic-ucd-version",
+]
+
+[[package]]
+name = "unic-ucd-version"
+version = "0.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "96bd2f2237fe450fcd0a1d2f5f4e91711124f7857ba2e964247776ebeeb7b0c4"
+dependencies = [
+ "unic-common",
+]
+
 [[package]]
 name = "unicase"
 version = "2.6.0"

compiler/rustc_errors/src/emitter.rs

@@ -730,7 +730,7 @@ impl EmitterWriter {
         }
 
         let source_string = match file.get_line(line.line_index - 1) {
-            Some(s) => replace_tabs(&*s),
+            Some(s) => normalize_whitespace(&*s),
             None => return Vec::new(),
         };
 
@@ -1286,7 +1286,7 @@ impl EmitterWriter {
             }
             for &(ref text, _) in msg.iter() {
                 // Account for newlines to align output to its label.
-                for (line, text) in replace_tabs(text).lines().enumerate() {
+                for (line, text) in normalize_whitespace(text).lines().enumerate() {
                     buffer.append(
                         0 + line,
                         &format!(
@@ -1550,7 +1550,7 @@ impl EmitterWriter {
 
                             self.draw_line(
                                 &mut buffer,
-                                &replace_tabs(&unannotated_line),
+                                &normalize_whitespace(&unannotated_line),
                                 annotated_file.lines[line_idx + 1].line_index - 1,
                                 last_buffer_line_num,
                                 width_offset,
@@ -1672,7 +1672,7 @@ impl EmitterWriter {
                     buffer.puts(
                         row_num - 1,
                         max_line_num_len + 3,
-                        &replace_tabs(
+                        &normalize_whitespace(
                             &*file_lines
                                 .file
                                 .get_line(file_lines.lines[line_pos].line_index)
@@ -1698,7 +1698,7 @@ impl EmitterWriter {
                 }
 
                 // print the suggestion
-                buffer.append(row_num, &replace_tabs(line), Style::NoStyle);
+                buffer.append(row_num, &normalize_whitespace(line), Style::NoStyle);
 
                 // Colorize addition/replacements with green.
                 for &SubstitutionHighlight { start, end } in highlight_parts {
@@ -2081,6 +2081,7 @@ fn num_decimal_digits(num: usize) -> usize {
 // We replace some characters so the CLI output is always consistent and underlines aligned.
 const OUTPUT_REPLACEMENTS: &[(char, &str)] = &[
     ('\t', "    "),   // We do our own tab replacement
+    ('\u{200D}', ""), // Replace ZWJ with nothing for consistent terminal output of grapheme clusters.
     ('\u{202A}', ""), // The following unicode text flow control characters are inconsistently
     ('\u{202B}', ""), // supported accross CLIs and can cause confusion due to the bytes on disk
     ('\u{202D}', ""), // not corresponding to the visible source code, so we replace them always.
@@ -2092,7 +2093,7 @@ const OUTPUT_REPLACEMENTS: &[(char, &str)] = &[
     ('\u{2069}', ""),
 ];
 
-fn replace_tabs(str: &str) -> String {
+fn normalize_whitespace(str: &str) -> String {
     let mut s = str.to_string();
     for (c, replacement) in OUTPUT_REPLACEMENTS {
         s = s.replace(*c, replacement);

compiler/rustc_interface/src/passes.rs

@@ -35,7 +35,7 @@ use rustc_session::output::{filename_for_input, filename_for_metadata};
 use rustc_session::search_paths::PathKind;
 use rustc_session::{Limit, Session};
 use rustc_span::symbol::{sym, Ident, Symbol};
-use rustc_span::FileName;
+use rustc_span::{FileName, MultiSpan};
 use rustc_trait_selection::traits;
 use rustc_typeck as typeck;
 use tempfile::Builder as TempFileBuilder;
@@ -450,6 +450,19 @@ pub fn configure_and_expand(
         });
     }
 
+    // Gate identifiers containing invalid Unicode codepoints that were recovered during lexing.
+    sess.parse_sess.bad_unicode_identifiers.with_lock(|identifiers| {
+        let mut identifiers: Vec<_> = identifiers.drain().collect();
+        identifiers.sort_by_key(|&(key, _)| key);
+        for (ident, mut spans) in identifiers.into_iter() {
+            spans.sort();
+            sess.diagnostic().span_err(
+                MultiSpan::from(spans),
+                &format!("identifiers cannot contain emoji: `{}`", ident),
+            );
+        }
+    });
+
     Ok(krate)
 }
 

compiler/rustc_lexer/Cargo.toml

@@ -17,6 +17,7 @@ doctest = false
 # Note that this crate purposefully does not depend on other rustc crates
 [dependencies]
 unicode-xid = "0.2.0"
+unic-emoji-char = "0.9.0"
 
 [dev-dependencies]
 expect-test = "1.0"

compiler/rustc_lexer/src/lib.rs

@@ -64,6 +64,8 @@ pub enum TokenKind {
     /// "ident" or "continue"
     /// At this step keywords are also considered identifiers.
     Ident,
+    /// Like the above, but containing invalid unicode codepoints.
+    InvalidIdent,
     /// "r#ident"
     RawIdent,
     /// An unknown prefix like `foo#`, `foo'`, `foo"`. Note that only the
@@ -411,6 +413,10 @@ impl Cursor<'_> {
                 let kind = Str { terminated };
                 Literal { kind, suffix_start }
             }
+            // Identifier starting with an emoji. Only lexed for graceful error recovery.
+            c if !c.is_ascii() && unic_emoji_char::is_emoji(c) => {
+                self.fake_ident_or_unknown_prefix()
+            }
             _ => Unknown,
         };
         Token::new(token_kind, self.len_consumed())
@@ -492,10 +498,28 @@ impl Cursor<'_> {
         // we see a prefix here, it is definitely an unknown prefix.
         match self.first() {
             '#' | '"' | '\'' => UnknownPrefix,
+            c if !c.is_ascii() && unic_emoji_char::is_emoji(c) => {
+                self.fake_ident_or_unknown_prefix()
+            }
             _ => Ident,
         }
     }
 
+    fn fake_ident_or_unknown_prefix(&mut self) -> TokenKind {
+        // Start is already eaten, eat the rest of identifier.
+        self.eat_while(|c| {
+            unicode_xid::UnicodeXID::is_xid_continue(c)
+                || (!c.is_ascii() && unic_emoji_char::is_emoji(c))
+                || c == '\u{200d}'
+        });
+        // Known prefixes must have been handled earlier. So if
+        // we see a prefix here, it is definitely an unknown prefix.
+        match self.first() {
+            '#' | '"' | '\'' => UnknownPrefix,
+            _ => InvalidIdent,
+        }
+    }
+
     fn number(&mut self, first_digit: char) -> LiteralKind {
         debug_assert!('0' <= self.prev() && self.prev() <= '9');
         let mut base = Base::Decimal;

compiler/rustc_parse/src/lexer/mod.rs

@@ -1,3 +1,4 @@
+use crate::lexer::unicode_chars::UNICODE_ARRAY;
 use rustc_ast::ast::{self, AttrStyle};
 use rustc_ast::token::{self, CommentKind, Token, TokenKind};
 use rustc_ast::tokenstream::{Spacing, TokenStream};
@@ -222,6 +223,22 @@ impl<'a> StringReader<'a> {
                 }
                 token::Ident(sym, is_raw_ident)
             }
+            rustc_lexer::TokenKind::InvalidIdent
+                // Do not recover an identifier with emoji if the codepoint is a confusable
+                // with a recoverable substitution token, like `➖`.
+                if UNICODE_ARRAY
+                    .iter()
+                    .find(|&&(c, _, _)| {
+                        let sym = self.str_from(start);
+                        sym.chars().count() == 1 && c == sym.chars().next().unwrap()
+                    })
+                    .is_none() =>
+            {
+                let sym = nfc_normalize(self.str_from(start));
+                let span = self.mk_sp(start, self.pos);
+                self.sess.bad_unicode_identifiers.borrow_mut().entry(sym).or_default().push(span);
+                token::Ident(sym, false)
+            }
             rustc_lexer::TokenKind::Literal { kind, suffix_start } => {
                 let suffix_start = start + BytePos(suffix_start as u32);
                 let (kind, symbol) = self.cook_lexer_literal(start, suffix_start, kind);
@@ -293,7 +310,7 @@ impl<'a> StringReader<'a> {
             rustc_lexer::TokenKind::Caret => token::BinOp(token::Caret),
             rustc_lexer::TokenKind::Percent => token::BinOp(token::Percent),
 
-            rustc_lexer::TokenKind::Unknown => {
+            rustc_lexer::TokenKind::Unknown | rustc_lexer::TokenKind::InvalidIdent => {
                 let c = self.str_from(start).chars().next().unwrap();
                 let mut err =
                     self.struct_fatal_span_char(start, self.pos, "unknown start of token", c);

compiler/rustc_parse/src/lexer/unicode_chars.rs

@@ -7,7 +7,7 @@ use rustc_errors::{Applicability, DiagnosticBuilder};
 use rustc_span::{symbol::kw, BytePos, Pos, Span};
 
 #[rustfmt::skip] // for line breaks
-const UNICODE_ARRAY: &[(char, &str, char)] = &[
+pub(crate) const UNICODE_ARRAY: &[(char, &str, char)] = &[
     ('
', "Line Separator", ' '),
     ('
', "Paragraph Separator", ' '),
     (' ', "Ogham Space mark", ' '),

compiler/rustc_session/src/parse.rs

@@ -119,8 +119,13 @@ pub struct ParseSess {
     pub config: CrateConfig,
     pub edition: Edition,
     pub missing_fragment_specifiers: Lock<FxHashMap<Span, NodeId>>,
-    /// Places where raw identifiers were used. This is used for feature-gating raw identifiers.
+    /// Places where raw identifiers were used. This is used to avoid complaining about idents
+    /// clashing with keywords in new editions.
     pub raw_identifier_spans: Lock<Vec<Span>>,
+    /// Places where identifiers that contain invalid Unicode codepoints but that look like they
+    /// should be. Useful to avoid bad tokenization when encountering emoji. We group them to
+    /// provide a single error per unique incorrect identifier.
+    pub bad_unicode_identifiers: Lock<FxHashMap<Symbol, Vec<Span>>>,
     source_map: Lrc<SourceMap>,
     pub buffered_lints: Lock<Vec<BufferedEarlyLint>>,
     /// Contains the spans of block expressions that could have been incomplete based on the
@@ -160,6 +165,7 @@ impl ParseSess {
             edition: ExpnId::root().expn_data().edition,
             missing_fragment_specifiers: Default::default(),
             raw_identifier_spans: Lock::new(Vec::new()),
+            bad_unicode_identifiers: Lock::new(Default::default()),
             source_map,
             buffered_lints: Lock::new(vec![]),
             ambiguous_block_expr_parse: Lock::new(FxHashMap::default()),

src/librustdoc/html/highlight.rs

@@ -632,7 +632,7 @@ impl<'a> Classifier<'a> {
                 },
                 Some(c) => c,
             },
-            TokenKind::RawIdent | TokenKind::UnknownPrefix => {
+            TokenKind::RawIdent | TokenKind::UnknownPrefix | TokenKind::InvalidIdent => {
                 Class::Ident(self.new_span(before, text))
             }
             TokenKind::Lifetime { .. } => Class::Lifetime,

src/test/ui/parser/emoji-identifiers.rs

@@ -0,0 +1,16 @@
+struct ABig👩‍👩‍👧‍👧Family; //~ ERROR identifiers cannot contain emoji
+struct 👀; //~ ERROR identifiers cannot contain emoji
+impl 👀 {
+    fn full_of_✨() -> 👀 { //~ ERROR identifiers cannot contain emoji
+        👀
+    }
+}
+fn i_like_to_😅_a_lot() -> 👀 { //~ ERROR identifiers cannot contain emoji
+    👀::full_of✨() //~ ERROR no function or associated item named `full_of✨` found for struct `👀`
+    //~^ ERROR identifiers cannot contain emoji
+}
+fn main() {
+    let _ = i_like_to_😄_a_lot() ➖ 4; //~ ERROR cannot find function `i_like_to_😄_a_lot` in this scope
+    //~^ ERROR identifiers cannot contain emoji
+    //~| ERROR unknown start of token: \u{2796}
+}