syntax: Divide keywords into contextual/restricted. No bad words

Author: brson

src/librustsyntax/parse.rs

@@ -132,7 +132,7 @@ fn new_parser(sess: parse_sess, cfg: ast::crate_cfg, rdr: lexer::reader,
       reader: rdr,
       binop_precs: prec::binop_prec_table(),
       keywords: token::keyword_table(),
-      bad_expr_words: token::bad_expr_word_table()}
+      restricted_keywords: token::restricted_keyword_table()}
 }
 
 fn new_parser_from_source_str(sess: parse_sess, cfg: ast::crate_cfg,

src/librustsyntax/parse/common.rs

@@ -43,7 +43,7 @@ fn parse_path_list_ident(p: parser) -> ast::path_list_ident {
 }
 
 fn parse_value_ident(p: parser) -> ast::ident {
-    check_bad_expr_word(p);
+    check_restricted_keywords(p);
     ret parse_ident(p);
 }
 
@@ -87,15 +87,15 @@ fn expect_keyword(p: parser, word: str) {
     }
 }
 
-fn is_bad_expr_word(p: parser, word: str) -> bool {
-    p.bad_expr_words.contains_key(word)
+fn is_restricted_keyword(p: parser, word: str) -> bool {
+    p.restricted_keywords.contains_key(word)
 }
 
-fn check_bad_expr_word(p: parser) {
+fn check_restricted_keywords(p: parser) {
     alt p.token {
       token::IDENT(_, false) {
         let w = token_to_str(p.reader, p.token);
-        if is_bad_expr_word(p, w) {
+        if is_restricted_keyword(p, w) {
             p.fatal("found `" + w + "` in expression position");
         }
       }

src/librustsyntax/parse/parser.rs

@@ -59,7 +59,7 @@ type parser = @{
     reader: reader,
     binop_precs: @[op_spec],
     keywords: hashmap<str, ()>,
-    bad_expr_words: hashmap<str, ()>
+    restricted_keywords: hashmap<str, ()>
 };
 
 impl parser for parser {
@@ -543,7 +543,7 @@ fn parse_path(p: parser) -> @ast::path {
 fn parse_value_path(p: parser) -> @ast::path {
     let pt = parse_path(p);
     let last_word = vec::last(pt.idents);
-    if is_bad_expr_word(p, last_word) {
+    if is_restricted_keyword(p, last_word) {
         p.fatal("found " + last_word + " in expression position");
     }
     pt
@@ -802,7 +802,7 @@ fn parse_bottom_expr(p: parser) -> pexpr {
     } else if p.token == token::MOD_SEP ||
                   is_ident(p.token) && !is_keyword(p, "true") &&
                       !is_keyword(p, "false") {
-        check_bad_expr_word(p);
+        check_restricted_keywords(p);
         let pth = parse_path_and_ty_param_substs(p, true);
         hi = pth.span.hi;
         ex = ast::expr_path(pth);
@@ -1370,7 +1370,7 @@ fn parse_pat(p: parser) -> @ast::pat {
                 p.bump();
                 subpat = parse_pat(p);
             } else {
-                if is_bad_expr_word(p, fieldname) {
+                if is_restricted_keyword(p, fieldname) {
                     p.fatal("found " + fieldname + " in binding position");
                 }
                 subpat = @{id: p.get_id(),
@@ -2098,7 +2098,7 @@ fn parse_item_enum(p: parser, attrs: [ast::attribute]) -> @ast::item {
     let mut variants: [ast::variant] = [];
     // Newtype syntax
     if p.token == token::EQ {
-        if is_bad_expr_word(p, id) {
+        if is_restricted_keyword(p, id) {
             p.fatal("found " + id + " in enum constructor position");
         }
         p.bump();

src/librustsyntax/parse/token.rs

@@ -201,48 +201,79 @@ fn is_bar(t: token::token) -> bool {
 }
 
 #[doc = "
-All the valid words that have meaning in the Rust language. Some of these are
-nonetheless valid as identifiers becasue they are unambiguous.
+All the valid words that have meaning in the Rust language.
+
+Rust keywords are either 'contextual' or 'restricted'. Contextual
+keywords may be used as identifiers because their appearance in
+the grammar is unambiguous. Restricted keywords may not appear
+in positions that might otherwise contain _value identifiers_.
 "]
 fn keyword_table() -> hashmap<str, ()> {
     let keywords = str_hash();
-    for bad_expr_word_table().each_key {|word|
+    for contextual_keyword_table().each_key {|word|
+        keywords.insert(word, ());
+    }
+    for restricted_keyword_table().each_key {|word|
         keywords.insert(word, ());
     }
-    let other_keywords = [
+    ret keywords;
+}
+
+#[doc = "Keywords that may be used as identifiers"]
+fn contextual_keyword_table() -> hashmap<str, ()> {
+    let words = str_hash();
+    let keys = [
         "as",
         "bind",
         "else",
         "implements",
         "move",
         "of",
         "priv",
-        "self",
-        "send",
-        "static",
+        "self", "send", "static",
         "to",
         "use",
         "with"
     ];
-    for other_keywords.each {|word|
-        keywords.insert(word, ());
+    for keys.each {|word|
+        words.insert(word, ());
     }
-    ret keywords;
+    words
 }
 
 #[doc = "
-These are the words that shouldn't be allowed as value identifiers,
-because, if used at the start of a line, they will cause the line to be
-interpreted as a specific kind of statement, which would be confusing.
+Keywords that may not appear in any position that might otherwise contain a
+_value identifier_. Restricted keywords may still be used as other types of
+identifiers.
+
+Reasons:
+
+* For some (most?), if used at the start of a line, they will cause the line
+  to be interpreted as a specific kind of statement, which would be confusing.
+
+* `true` or `false` as identifiers would always be shadowed by
+  the boolean constants
 "]
-fn bad_expr_word_table() -> hashmap<str, ()> {
+fn restricted_keyword_table() -> hashmap<str, ()> {
     let words = str_hash();
-    let keys = ["alt", "assert", "be", "break", "check", "claim",
-                "class", "const", "cont", "copy", "crust", "do", "else",
-                "enum", "export", "fail", "false", "fn", "for", "if",
-                "iface", "impl", "import", "let", "log", "loop", "mod",
-                "mut", "native", "new", "pure", "resource", "true",
-                "ret", "trait", "type", "unchecked", "unsafe", "while"];
+    let keys = [
+        "alt",
+        "assert",
+        "be", "break",
+        "check", "claim", "class", "const", "cont", "copy", "crust",
+        "do",
+        "else", "enum", "export",
+        "fail", "false", "fn", "for",
+        "if", "iface", "impl", "import",
+        "let", "log", "loop",
+        "mod", "mut",
+        "native", "new",
+        "pure",
+        "resource", "ret",
+        "true", "trait", "type",
+        "unchecked", "unsafe",
+        "while"
+    ];
     for keys.each {|word|
         words.insert(word, ());
     }