quote_expr macro: embed Ident using special encoding that preserves h…

…ygiene.

This adds support to `quote_expr!` and friends for round-trip hygienic
preservation of Ident.

Here are the pieces of the puzzle:

* adding a method for encoding Ident for re-reading into token tree.

* Support for reading such encoded Idents in the lexer.  Note that one
  must peek ahead for MOD_SEP after scan_embedded_hygienic_ident.

* To ensure that encoded Idents are only read when we are in the midst
  of expanding a `quote_expr` or similar, added a
  `read_embedded_ident` flag on `StringReader`.

* pprust support for encoding Ident's as (uint,uint) pairs (for hygiene).

src/libsyntax/ast.rs

@@ -55,6 +55,12 @@ impl Ident {
     pub fn as_str<'a>(&'a self) -> &'a str {
         self.name.as_str()
     }
+
+    pub fn encode_with_hygiene(&self) -> String {
+        format!("\x00name_{:u},ctxt_{:u}\x00",
+                self.name.uint(),
+                self.ctxt)
+    }
 }
 
 impl Show for Ident {

src/libsyntax/ext/quote.rs

@@ -97,20 +97,39 @@ pub mod rt {
         fn to_source(&self) -> String;
     }
 
+    // FIXME (Issue #16472): This should go away after ToToken impls
+    // are revised to go directly to token-trees.
+    trait ToSourceWithHygiene : ToSource {
+        // Takes a thing and generates a string containing rust code
+        // for it, encoding Idents as special byte sequences to
+        // maintain hygiene across serialization and deserialization.
+        fn to_source_with_hygiene(&self) -> String;
+    }
+
     macro_rules! impl_to_source(
         (Gc<$t:ty>, $pp:ident) => (
             impl ToSource for Gc<$t> {
                 fn to_source(&self) -> String {
                     pprust::$pp(&**self)
                 }
             }
+            impl ToSourceWithHygiene for Gc<$t> {
+                fn to_source_with_hygiene(&self) -> String {
+                    pprust::with_hygiene::$pp(&**self)
+                }
+            }
         );
         ($t:ty, $pp:ident) => (
             impl ToSource for $t {
                 fn to_source(&self) -> String {
                     pprust::$pp(self)
                 }
             }
+            impl ToSourceWithHygiene for $t {
+                fn to_source_with_hygiene(&self) -> String {
+                    pprust::with_hygiene::$pp(self)
+                }
+            }
         );
     )
 
@@ -122,13 +141,28 @@ pub mod rt {
             .to_string()
     }
 
+    fn slice_to_source_with_hygiene<'a, T: ToSourceWithHygiene>(
+        sep: &'static str, xs: &'a [T]) -> String {
+        xs.iter()
+            .map(|i| i.to_source_with_hygiene())
+            .collect::<Vec<String>>()
+            .connect(sep)
+            .to_string()
+    }
+
     macro_rules! impl_to_source_slice(
         ($t:ty, $sep:expr) => (
             impl<'a> ToSource for &'a [$t] {
                 fn to_source(&self) -> String {
                     slice_to_source($sep, *self)
                 }
             }
+
+            impl<'a> ToSourceWithHygiene for &'a [$t] {
+                fn to_source_with_hygiene(&self) -> String {
+                    slice_to_source_with_hygiene($sep, *self)
+                }
+            }
         )
     )
 
@@ -138,6 +172,12 @@ pub mod rt {
         }
     }
 
+    impl ToSourceWithHygiene for ast::Ident {
+        fn to_source_with_hygiene(&self) -> String {
+            self.encode_with_hygiene()
+        }
+    }
+
     impl_to_source!(ast::Ty, ty_to_string)
     impl_to_source!(ast::Block, block_to_string)
     impl_to_source!(ast::Arg, arg_to_string)
@@ -156,6 +196,11 @@ pub mod rt {
             pprust::attribute_to_string(&dummy_spanned(*self))
         }
     }
+    impl ToSourceWithHygiene for ast::Attribute_ {
+        fn to_source_with_hygiene(&self) -> String {
+            self.to_source()
+        }
+    }
 
     impl<'a> ToSource for &'a str {
         fn to_source(&self) -> String {
@@ -164,26 +209,46 @@ pub mod rt {
             pprust::lit_to_string(&lit)
         }
     }
+    impl<'a> ToSourceWithHygiene for &'a str {
+        fn to_source_with_hygiene(&self) -> String {
+            self.to_source()
+        }
+    }
 
     impl ToSource for () {
         fn to_source(&self) -> String {
             "()".to_string()
         }
     }
+    impl ToSourceWithHygiene for () {
+        fn to_source_with_hygiene(&self) -> String {
+            self.to_source()
+        }
+    }
 
     impl ToSource for bool {
         fn to_source(&self) -> String {
             let lit = dummy_spanned(ast::LitBool(*self));
             pprust::lit_to_string(&lit)
         }
     }
+    impl ToSourceWithHygiene for bool {
+        fn to_source_with_hygiene(&self) -> String {
+            self.to_source()
+        }
+    }
 
     impl ToSource for char {
         fn to_source(&self) -> String {
             let lit = dummy_spanned(ast::LitChar(*self));
             pprust::lit_to_string(&lit)
         }
     }
+    impl ToSourceWithHygiene for char {
+        fn to_source_with_hygiene(&self) -> String {
+            self.to_source()
+        }
+    }
 
     macro_rules! impl_to_source_int(
         (signed, $t:ty, $tag:ident) => (
@@ -194,6 +259,11 @@ pub mod rt {
                     pprust::lit_to_string(&dummy_spanned(lit))
                 }
             }
+            impl ToSourceWithHygiene for $t {
+                fn to_source_with_hygiene(&self) -> String {
+                    self.to_source()
+                }
+            }
         );
         (unsigned, $t:ty, $tag:ident) => (
             impl ToSource for $t {
@@ -202,6 +272,11 @@ pub mod rt {
                     pprust::lit_to_string(&dummy_spanned(lit))
                 }
             }
+            impl ToSourceWithHygiene for $t {
+                fn to_source_with_hygiene(&self) -> String {
+                    self.to_source()
+                }
+            }
         );
     )
 
@@ -223,7 +298,7 @@ pub mod rt {
         ($t:ty) => (
             impl ToTokens for $t {
                 fn to_tokens(&self, cx: &ExtCtxt) -> Vec<TokenTree> {
-                    cx.parse_tts(self.to_source())
+                    cx.parse_tts_with_hygiene(self.to_source_with_hygiene())
                 }
             }
         )
@@ -233,7 +308,7 @@ pub mod rt {
         ($t:ty) => (
             impl<'a> ToTokens for $t {
                 fn to_tokens(&self, cx: &ExtCtxt) -> Vec<TokenTree> {
-                    cx.parse_tts(self.to_source())
+                    cx.parse_tts_with_hygiene(self.to_source_with_hygiene())
                 }
             }
         )
@@ -272,7 +347,13 @@ pub mod rt {
         fn parse_item(&self, s: String) -> Gc<ast::Item>;
         fn parse_expr(&self, s: String) -> Gc<ast::Expr>;
         fn parse_stmt(&self, s: String) -> Gc<ast::Stmt>;
-        fn parse_tts(&self, s: String) -> Vec<ast::TokenTree> ;
+        fn parse_tts(&self, s: String) -> Vec<ast::TokenTree>;
+    }
+
+    trait ExtParseUtilsWithHygiene {
+        // FIXME (Issue #16472): This should go away after ToToken impls
+        // are revised to go directly to token-trees.
+        fn parse_tts_with_hygiene(&self, s: String) -> Vec<ast::TokenTree>;
     }
 
     impl<'a> ExtParseUtils for ExtCtxt<'a> {
@@ -315,6 +396,18 @@ pub mod rt {
         }
     }
 
+    impl<'a> ExtParseUtilsWithHygiene for ExtCtxt<'a> {
+
+        fn parse_tts_with_hygiene(&self, s: String) -> Vec<ast::TokenTree> {
+            use parse::with_hygiene::parse_tts_from_source_str;
+            parse_tts_from_source_str("<quote expansion>".to_string(),
+                                      s,
+                                      self.cfg(),
+                                      self.parse_sess())
+        }
+
+    }
+
 }
 
 pub fn expand_quote_tokens(cx: &mut ExtCtxt,

src/libsyntax/parse/lexer/mod.rs

@@ -17,7 +17,9 @@ use parse::token;
 use parse::token::{str_to_ident};
 
 use std::char;
+use std::fmt;
 use std::mem::replace;
+use std::num;
 use std::rc::Rc;
 use std::str;
 
@@ -55,6 +57,11 @@ pub struct StringReader<'a> {
     /* cached: */
     pub peek_tok: token::Token,
     pub peek_span: Span,
+
+    // FIXME (Issue #16472): This field should go away after ToToken impls
+    // are revised to go directly to token-trees.
+    /// Is \x00<name>,<ctxt>\x00 is interpreted as encoded ast::Ident?
+    read_embedded_ident: bool,
 }
 
 impl<'a> Reader for StringReader<'a> {
@@ -106,6 +113,17 @@ impl<'a> Reader for TtReader<'a> {
     }
 }
 
+// FIXME (Issue #16472): This function should go away after
+// ToToken impls are revised to go directly to token-trees.
+pub fn make_reader_with_embedded_idents<'b>(span_diagnostic: &'b SpanHandler,
+                                            filemap: Rc<codemap::FileMap>)
+                                            -> StringReader<'b> {
+    let mut sr = StringReader::new_raw(span_diagnostic, filemap);
+    sr.read_embedded_ident = true;
+    sr.advance_token();
+    sr
+}
+
 impl<'a> StringReader<'a> {
     /// For comments.rs, which hackily pokes into pos and curr
     pub fn new_raw<'b>(span_diagnostic: &'b SpanHandler,
@@ -120,6 +138,7 @@ impl<'a> StringReader<'a> {
             /* dummy values; not read */
             peek_tok: token::EOF,
             peek_span: codemap::DUMMY_SP,
+            read_embedded_ident: false,
         };
         sr.bump();
         sr
@@ -512,6 +531,81 @@ impl<'a> StringReader<'a> {
         })
     }
 
+    // FIXME (Issue #16472): The scan_embedded_hygienic_ident function
+    // should go away after we revise the syntax::ext::quote::ToToken
+    // impls to go directly to token-trees instead of thing -> string
+    // -> token-trees.  (The function is currently used to resolve
+    // Issues #15750 and #15962.)
+    //
+    // Since this function is only used for certain internal macros,
+    // and the functionality it provides is not exposed to end user
+    // programs, pnkfelix deliberately chose to write it in a way that
+    // favors rustc debugging effectiveness over runtime efficiency.
+
+    /// Scan through input of form \x00name_NNNNNN,ctxt_CCCCCCC\x00
+    /// where: `NNNNNN` is a string of characters forming an integer
+    /// (the name) and `CCCCCCC` is a string of characters forming an
+    /// integer (the ctxt), separate by a comma and delimited by a
+    /// `\x00` marker.
+    #[inline(never)]
+    fn scan_embedded_hygienic_ident(&mut self) -> ast::Ident {
+        fn bump_expecting_char<'a,D:fmt::Show>(r: &mut StringReader<'a>,
+                                               c: char,
+                                               described_c: D,
+                                               where: &str) {
+            match r.curr {
+                Some(r_c) if r_c == c => r.bump(),
+                Some(r_c) => fail!("expected {}, hit {}, {}", described_c, r_c, where),
+                None      => fail!("expected {}, hit EOF, {}", described_c, where),
+            }
+        }
+
+        let where = "while scanning embedded hygienic ident";
+
+        // skip over the leading `\x00`
+        bump_expecting_char(self, '\x00', "nul-byte", where);
+
+        // skip over the "name_"
+        for c in "name_".chars() {
+            bump_expecting_char(self, c, c, where);
+        }
+
+        let start_bpos = self.last_pos;
+        let base = 10;
+
+        // find the integer representing the name
+        self.scan_digits(base);
+        let encoded_name : u32 = self.with_str_from(start_bpos, |s| {
+            num::from_str_radix(s, 10).unwrap_or_else(|| {
+                fail!("expected digits representing a name, got `{}`, {}, range [{},{}]",
+                      s, where, start_bpos, self.last_pos);
+            })
+        });
+
+        // skip over the `,`
+        bump_expecting_char(self, ',', "comma", where);
+
+        // skip over the "ctxt_"
+        for c in "ctxt_".chars() {
+            bump_expecting_char(self, c, c, where);
+        }
+
+        // find the integer representing the ctxt
+        let start_bpos = self.last_pos;
+        self.scan_digits(base);
+        let encoded_ctxt : ast::SyntaxContext = self.with_str_from(start_bpos, |s| {
+            num::from_str_radix(s, 10).unwrap_or_else(|| {
+                fail!("expected digits representing a ctxt, got `{}`, {}", s, where);
+            })
+        });
+
+        // skip over the `\x00`
+        bump_expecting_char(self, '\x00', "nul-byte", where);
+
+        ast::Ident { name: ast::Name(encoded_name),
+                     ctxt: encoded_ctxt, }
+    }
+
     /// Scan through any digits (base `radix`) or underscores, and return how
     /// many digits there were.
     fn scan_digits(&mut self, radix: uint) -> uint {
@@ -839,6 +933,17 @@ impl<'a> StringReader<'a> {
             return self.scan_number(c.unwrap());
         }
 
+        if self.read_embedded_ident {
+            match (c.unwrap(), self.nextch(), self.nextnextch()) {
+                ('\x00', Some('n'), Some('a')) => {
+                    let ast_ident = self.scan_embedded_hygienic_ident();
+                    let is_mod_name = self.curr_is(':') && self.nextch_is(':');
+                    return token::IDENT(ast_ident, is_mod_name);
+                }
+                _ => {}
+            }
+        }
+
         match c.expect("next_token_inner called at EOF") {
           // One-byte tokens.
           ';' => { self.bump(); return token::SEMI; }