Refactor the lexer to use FilePos types

brson · brson · commit 4a0f4f5e311b · 2012-11-15T14:24:53.000-08:00
diff --git a/src/libsyntax/codemap.rs b/src/libsyntax/codemap.rs
@@ -200,11 +200,8 @@ pub impl FileMap {
                                      start_pos);
     }
 
-    fn next_line(@self, +chpos: CharPos, +byte_pos: BytePos) {
-        self.lines.push(FilePos {
-            ch: chpos,
-            byte: byte_pos + self.start_pos.byte
-        });
+    fn next_line(@self, +pos: FilePos) {
+        self.lines.push(pos);
     }
 
     pub fn get_line(@self, line: int) -> ~str unsafe {
@@ -231,6 +228,18 @@ pub impl CodeMap {
     }
 
     pub fn add_filemap(@self, filemap: @FileMap) {
+        let expected_byte_pos = if self.files.len() == 0 {
+            0
+        } else {
+            let last_start = self.files.last().start_pos.byte.to_uint();
+            let last_len = self.files.last().src.len();
+            last_start + last_len
+        };
+        let actual_byte_pos = filemap.start_pos.byte.to_uint();
+        debug!("codemap: adding filemap: %s", filemap.name);
+        debug!("codemap: expected offset: %u", expected_byte_pos);
+        debug!("codemap: actual offset: %u", actual_byte_pos);
+        assert expected_byte_pos == actual_byte_pos;
         self.files.push(filemap);
     }
 
diff --git a/src/libsyntax/ext/source_util.rs b/src/libsyntax/ext/source_util.rs
@@ -58,10 +58,13 @@ fn expand_include(cx: ext_ctxt, sp: span, arg: ast::mac_arg,
                   _body: ast::mac_body) -> @ast::expr {
     let args = get_mac_args(cx, sp, arg, 1u, option::Some(1u), ~"include");
     let file = expr_to_str(cx, args[0], ~"#include_str requires a string");
-    let p = parse::new_parser_from_file(cx.parse_sess(), cx.cfg(),
-                                        &res_rel_file(cx, sp, &Path(file)),
-                                        parse::parser::SOURCE_FILE);
-    return p.parse_expr();
+    let (p, rdr) = parse::new_parser_etc_from_file(
+        cx.parse_sess(), cx.cfg(),
+        &res_rel_file(cx, sp, &Path(file)),
+        parse::parser::SOURCE_FILE);
+    let e = p.parse_expr();
+    parse::update_parse_sess_position(&cx.parse_sess(), &rdr);
+    return e;
 }
 
 fn expand_include_str(cx: ext_ctxt, sp: codemap::span, arg: ast::mac_arg,
diff --git a/src/libsyntax/parse.rs b/src/libsyntax/parse.rs
@@ -11,6 +11,7 @@ export parse_crate_from_source_str;
 export parse_expr_from_source_str, parse_item_from_source_str;
 export parse_stmt_from_source_str;
 export parse_from_source_str;
+export update_parse_sess_position;
 
 use parser::Parser;
 use attr::parser_attr;
@@ -76,7 +77,7 @@ fn parse_crate_from_crate_file(input: &Path, cfg: ast::crate_cfg,
     let leading_attrs = p.parse_inner_attrs_and_next();
     let { inner: crate_attrs, next: first_cdir_attr } = leading_attrs;
     let cdirs = p.parse_crate_directives(token::EOF, first_cdir_attr);
-    eval::update_parse_sess_position(&sess, &rdr);
+    update_parse_sess_position(&sess, &rdr);
     let cx = @{sess: sess, cfg: /* FIXME (#2543) */ copy p.cfg};
     let companionmod = input.filestem().map(|s| Path(*s));
     let (m, attrs) = eval::eval_crate_directives_to_mod(
@@ -96,7 +97,7 @@ fn parse_crate_from_source_file(input: &Path, cfg: ast::crate_cfg,
     let (p, rdr) = new_parser_etc_from_file(sess, cfg, input,
                                             parser::SOURCE_FILE);
     let r = p.parse_crate_mod(cfg);
-    eval::update_parse_sess_position(&sess, &rdr);
+    update_parse_sess_position(&sess, &rdr);
     return r;
 }
 
@@ -106,7 +107,7 @@ fn parse_crate_from_source_str(name: ~str, source: @~str, cfg: ast::crate_cfg,
                                                   codemap::FssNone, source);
     let r = p.parse_crate_mod(cfg);
     p.abort_if_errors();
-    eval::update_parse_sess_position(&sess, &rdr);
+    update_parse_sess_position(&sess, &rdr);
     return r;
 }
 
@@ -116,7 +117,7 @@ fn parse_expr_from_source_str(name: ~str, source: @~str, cfg: ast::crate_cfg,
                                                   codemap::FssNone, source);
     let r = p.parse_expr();
     p.abort_if_errors();
-    eval::update_parse_sess_position(&sess, &rdr);
+    update_parse_sess_position(&sess, &rdr);
     return r;
 }
 
@@ -127,7 +128,7 @@ fn parse_item_from_source_str(name: ~str, source: @~str, cfg: ast::crate_cfg,
                                                   codemap::FssNone, source);
     let r = p.parse_item(attrs);
     p.abort_if_errors();
-    eval::update_parse_sess_position(&sess, &rdr);
+    update_parse_sess_position(&sess, &rdr);
     return r;
 }
 
@@ -138,7 +139,7 @@ fn parse_stmt_from_source_str(name: ~str, source: @~str, cfg: ast::crate_cfg,
                                                   codemap::FssNone, source);
     let r = p.parse_stmt(attrs);
     p.abort_if_errors();
-    eval::update_parse_sess_position(&sess, &rdr);
+    update_parse_sess_position(&sess, &rdr);
     return r;
 }
 
@@ -155,7 +156,7 @@ fn parse_from_source_str<T>(f: fn (p: Parser) -> T,
         p.reader.fatal(~"expected end-of-string");
     }
     p.abort_if_errors();
-    eval::update_parse_sess_position(&sess, &rdr);
+    update_parse_sess_position(&sess, &rdr);
     move r
 }
 
@@ -216,3 +217,10 @@ fn new_parser_from_tt(sess: parse_sess, cfg: ast::crate_cfg,
                                     None, tt);
     return Parser(sess, cfg, trdr as reader, parser::SOURCE_FILE)
 }
+
+fn update_parse_sess_position(sess: &parse_sess, r: &lexer::string_reader) {
+    sess.pos = FilePos {
+        ch: r.last_pos.ch,
+        byte: r.last_pos.byte
+    };
+}
diff --git a/src/libsyntax/parse/comments.rs b/src/libsyntax/parse/comments.rs
@@ -131,7 +131,7 @@ fn consume_non_eol_whitespace(rdr: string_reader) {
 fn push_blank_line_comment(rdr: string_reader, comments: &mut ~[cmnt]) {
     debug!(">>> blank-line comment");
     let v: ~[~str] = ~[];
-    comments.push({style: blank_line, lines: v, pos: rdr.chpos});
+    comments.push({style: blank_line, lines: v, pos: rdr.last_pos.ch});
 }
 
 fn consume_whitespace_counting_blank_lines(rdr: string_reader,
@@ -148,7 +148,7 @@ fn consume_whitespace_counting_blank_lines(rdr: string_reader,
 fn read_shebang_comment(rdr: string_reader, code_to_the_left: bool,
                                             comments: &mut ~[cmnt]) {
     debug!(">>> shebang comment");
-    let p = rdr.chpos;
+    let p = rdr.last_pos.ch;
     debug!("<<< shebang comment");
     comments.push({
         style: if code_to_the_left { trailing } else { isolated },
@@ -160,7 +160,7 @@ fn read_shebang_comment(rdr: string_reader, code_to_the_left: bool,
 fn read_line_comments(rdr: string_reader, code_to_the_left: bool,
                                           comments: &mut ~[cmnt]) {
     debug!(">>> line comments");
-    let p = rdr.chpos;
+    let p = rdr.last_pos.ch;
     let mut lines: ~[~str] = ~[];
     while rdr.curr == '/' && nextch(rdr) == '/' {
         let line = read_one_line_comment(rdr);
@@ -209,7 +209,7 @@ fn trim_whitespace_prefix_and_push_line(lines: &mut ~[~str],
 fn read_block_comment(rdr: string_reader, code_to_the_left: bool,
                                           comments: &mut ~[cmnt]) {
     debug!(">>> block comment");
-    let p = rdr.chpos;
+    let p = rdr.last_pos.ch;
     let mut lines: ~[~str] = ~[];
     let mut col: CharPos = rdr.col;
     bump(rdr);
@@ -319,7 +319,7 @@ fn gather_comments_and_literals(span_diagnostic: diagnostic::span_handler,
         }
 
 
-        let bstart = rdr.pos;
+        let bstart = rdr.pos.byte;
         rdr.next_token();
         //discard, and look ahead; we're working with internal state
         let {tok: tok, sp: sp} = rdr.peek();
diff --git a/src/libsyntax/parse/eval.rs b/src/libsyntax/parse/eval.rs
@@ -1,8 +1,8 @@
+use parse::update_parse_sess_position;
 use parser::{Parser, SOURCE_FILE};
 use attr::parser_attr;
 
 export eval_crate_directives_to_mod;
-export update_parse_sess_position;
 
 type ctx =
     @{sess: parse::parse_sess,
@@ -74,13 +74,6 @@ fn parse_companion_mod(cx: ctx, prefix: &Path, suffix: &Option<Path>)
     }
 }
 
-fn update_parse_sess_position(sess: &parse_sess, r: &lexer::string_reader) {
-    sess.pos = FilePos {
-        ch: r.chpos,
-        byte: sess.pos.byte + r.pos
-    };
-}
-
 fn cdir_path_opt(default: ~str, attrs: ~[ast::attribute]) -> ~str {
     match ::attr::first_attr_value_str_by_name(attrs, ~"path") {
       Some(d) => d,
diff --git a/src/libsyntax/parse/lexer.rs b/src/libsyntax/parse/lexer.rs
@@ -1,5 +1,5 @@
 use diagnostic::span_handler;
-use codemap::{span, CodeMap, CharPos, BytePos};
+use codemap::{span, CodeMap, CharPos, BytePos, FilePos};
 use ext::tt::transcribe::{tt_reader,  new_tt_reader, dup_tt_reader,
                              tt_next_token};
 
@@ -21,10 +21,10 @@ trait reader {
 type string_reader = @{
     span_diagnostic: span_handler,
     src: @~str,
+    mut pos: FilePos,
+    mut last_pos: FilePos,
     mut col: CharPos,
-    mut pos: BytePos,
     mut curr: char,
-    mut chpos: CharPos,
     filemap: @codemap::FileMap,
     interner: @token::ident_interner,
     /* cached: */
@@ -48,9 +48,10 @@ fn new_low_level_string_reader(span_diagnostic: span_handler,
     // Force the initial reader bump to start on a fresh line
     let initial_char = '\n';
     let r = @{span_diagnostic: span_diagnostic, src: filemap.src,
-              mut col: CharPos(0), mut pos: BytePos(0),
+              mut pos: filemap.start_pos,
+              mut last_pos: filemap.start_pos,
+              mut col: CharPos(0),
               mut curr: initial_char,
-              mut chpos: filemap.start_pos.ch,
               filemap: filemap, interner: itr,
               /* dummy values; not read */
               mut peek_tok: token::EOF,
@@ -61,7 +62,9 @@ fn new_low_level_string_reader(span_diagnostic: span_handler,
 
 fn dup_string_reader(&&r: string_reader) -> string_reader {
     @{span_diagnostic: r.span_diagnostic, src: r.src,
-      mut col: r.col, mut pos: r.pos, mut curr: r.curr, mut chpos: r.chpos,
+      mut pos: r.pos,
+      mut last_pos: r.last_pos,
+      mut col: r.col, mut curr: r.curr,
       filemap: r.filemap, interner: r.interner,
       mut peek_tok: r.peek_tok, mut peek_span: r.peek_span}
 }
@@ -116,34 +119,48 @@ fn string_advance_token(&&r: string_reader) {
     if is_eof(r) {
         r.peek_tok = token::EOF;
     } else {
-        let start_chpos = r.chpos;
+        let start_chpos = r.last_pos.ch;
         r.peek_tok = next_token_inner(r);
-        r.peek_span = ast_util::mk_sp(start_chpos, r.chpos);
+        r.peek_span = ast_util::mk_sp(start_chpos, r.last_pos.ch);
     };
 
 }
 
+fn byte_offset(rdr: string_reader) -> BytePos {
+    (rdr.pos.byte - rdr.filemap.start_pos.byte)
+}
+
 fn get_str_from(rdr: string_reader, start: BytePos) -> ~str unsafe {
     // I'm pretty skeptical about this subtraction. What if there's a
     // multi-byte character before the mark?
-    return str::slice(*rdr.src, start.to_uint() - 1u, rdr.pos.to_uint() - 1u);
+    return str::slice(*rdr.src, start.to_uint() - 1u,
+                      byte_offset(rdr).to_uint() - 1u);
 }
 
 fn bump(rdr: string_reader) {
-    if rdr.pos.to_uint() < (*rdr.src).len() {
+    rdr.last_pos = rdr.pos;
+    let current_byte_offset = byte_offset(rdr).to_uint();;
+    if current_byte_offset < (*rdr.src).len() {
+        let last_char = rdr.curr;
+        let next = str::char_range_at(*rdr.src, current_byte_offset);
+        let byte_offset_diff = next.next - current_byte_offset;
+        rdr.pos = FilePos {
+            ch: rdr.pos.ch + CharPos(1u),
+            byte: rdr.pos.byte + BytePos(byte_offset_diff)
+        };
+        rdr.curr = next.ch;
         rdr.col += CharPos(1u);
-        rdr.chpos += CharPos(1u);
-        if rdr.curr == '\n' {
-            rdr.filemap.next_line(rdr.chpos, rdr.pos);
+        if last_char == '\n' {
+            rdr.filemap.next_line(rdr.last_pos);
             rdr.col = CharPos(0u);
         }
-        let next = str::char_range_at(*rdr.src, rdr.pos.to_uint());
-        rdr.pos = BytePos(next.next);
-        rdr.curr = next.ch;
     } else {
         // XXX: What does this accomplish?
         if (rdr.curr != -1 as char) {
-            rdr.chpos += CharPos(1u);
+            rdr.pos = FilePos {
+                ch: rdr.pos.ch + CharPos(1u),
+                byte: rdr.pos.byte + BytePos(1u)
+            };
             rdr.col += CharPos(1u);
             rdr.curr = -1 as char;
         }
@@ -153,8 +170,9 @@ fn is_eof(rdr: string_reader) -> bool {
     rdr.curr == -1 as char
 }
 fn nextch(rdr: string_reader) -> char {
-    if rdr.pos.to_uint() < (*rdr.src).len() {
-        return str::char_at(*rdr.src, rdr.pos.to_uint());
+    let offset = byte_offset(rdr).to_uint();
+    if offset < (*rdr.src).len() {
+        return str::char_at(*rdr.src, offset);
     } else { return -1 as char; }
 }
 
@@ -211,15 +229,15 @@ fn consume_any_line_comment(rdr: string_reader)
             bump(rdr);
             // line comments starting with "///" or "//!" are doc-comments
             if rdr.curr == '/' || rdr.curr == '!' {
-                let start_chpos = rdr.chpos - CharPos(2u);
+                let start_chpos = rdr.pos.ch - CharPos(2u);
                 let mut acc = ~"//";
                 while rdr.curr != '\n' && !is_eof(rdr) {
                     str::push_char(&mut acc, rdr.curr);
                     bump(rdr);
                 }
                 return Some({
                     tok: token::DOC_COMMENT(rdr.interner.intern(@acc)),
-                    sp: ast_util::mk_sp(start_chpos, rdr.chpos)
+                    sp: ast_util::mk_sp(start_chpos, rdr.pos.ch)
                 });
             } else {
                 while rdr.curr != '\n' && !is_eof(rdr) { bump(rdr); }
@@ -234,7 +252,7 @@ fn consume_any_line_comment(rdr: string_reader)
         if nextch(rdr) == '!' {
             let cmap = @CodeMap::new();
             (*cmap).files.push(rdr.filemap);
-            let loc = cmap.lookup_char_pos_adj(rdr.chpos);
+            let loc = cmap.lookup_char_pos_adj(rdr.last_pos.ch);
             if loc.line == 1u && loc.col == CharPos(0u) {
                 while rdr.curr != '\n' && !is_eof(rdr) { bump(rdr); }
                 return consume_whitespace_and_comments(rdr);
@@ -250,7 +268,7 @@ fn consume_block_comment(rdr: string_reader)
 
     // block comments starting with "/**" or "/*!" are doc-comments
     if rdr.curr == '*' || rdr.curr == '!' {
-        let start_chpos = rdr.chpos - CharPos(2u);
+        let start_chpos = rdr.pos.ch - CharPos(2u);
         let mut acc = ~"/*";
         while !(rdr.curr == '*' && nextch(rdr) == '/') && !is_eof(rdr) {
             str::push_char(&mut acc, rdr.curr);
@@ -264,7 +282,7 @@ fn consume_block_comment(rdr: string_reader)
             bump(rdr);
             return Some({
                 tok: token::DOC_COMMENT(rdr.interner.intern(@acc)),
-                sp: ast_util::mk_sp(start_chpos, rdr.chpos)
+                sp: ast_util::mk_sp(start_chpos, rdr.pos.ch)
             });
         }
     } else {
@@ -584,7 +602,7 @@ fn next_token_inner(rdr: string_reader) -> token::Token {
         return token::LIT_INT(c2 as i64, ast::ty_char);
       }
       '"' => {
-        let n = rdr.pos;
+        let n = byte_offset(rdr);
         bump(rdr);
         while rdr.curr != '"' {
             if is_eof(rdr) {
