libsyntax: be more accepting of whitespace in lexer

mk/crates.mk

@@ -86,7 +86,7 @@ DEPS_serialize := std log
 DEPS_term := std log
 DEPS_test := std getopts serialize rbml term native:rust_test_helpers
 
-DEPS_syntax := std term serialize log arena libc rustc_bitflags
+DEPS_syntax := std term serialize log arena libc rustc_bitflags rustc_unicode
 DEPS_syntax_ext := syntax fmt_macros
 
 DEPS_rustc := syntax fmt_macros flate arena serialize getopts rbml rustc_front\

src/etc/unicode.py

@@ -398,7 +398,7 @@ def emit_norm_module(f, canon, compat, combine, norm_props):
         derived = load_properties("DerivedCoreProperties.txt", want_derived)
         scripts = load_properties("Scripts.txt", [])
         props = load_properties("PropList.txt",
-                ["White_Space", "Join_Control", "Noncharacter_Code_Point"])
+                ["White_Space", "Join_Control", "Noncharacter_Code_Point", "Pattern_White_Space"])
         norm_props = load_properties("DerivedNormalizationProps.txt",
                      ["Full_Composition_Exclusion"])
 
@@ -408,7 +408,7 @@ def emit_norm_module(f, canon, compat, combine, norm_props):
         # category tables
         for (name, cat, pfuns) in ("general_category", gencats, ["N", "Cc"]), \
                                   ("derived_property", derived, want_derived), \
-                                  ("property", props, ["White_Space"]):
+                                  ("property", props, ["White_Space", "Pattern_White_Space"]):
             emit_property_module(rf, name, cat, pfuns)
 
         # normalizations and conversions module

src/librustc_unicode/lib.rs

@@ -50,3 +50,8 @@ pub mod str {
 pub mod derived_property {
     pub use tables::derived_property::{Cased, Case_Ignorable};
 }
+
+// For use in libsyntax
+pub mod property {
+    pub use tables::property::Pattern_White_Space;
+}

src/librustc_unicode/tables.rs

@@ -1180,6 +1180,15 @@ pub mod derived_property {
 }
 
 pub mod property {
+    pub const Pattern_White_Space_table: &'static [(char, char)] = &[
+        ('\u{9}', '\u{d}'), ('\u{20}', '\u{20}'), ('\u{85}', '\u{85}'), ('\u{200e}', '\u{200f}'),
+        ('\u{2028}', '\u{2029}')
+    ];
+
+    pub fn Pattern_White_Space(c: char) -> bool {
+        super::bsearch_range_table(c, Pattern_White_Space_table)
+    }
+
     pub const White_Space_table: &'static [(char, char)] = &[
         ('\u{9}', '\u{d}'), ('\u{20}', '\u{20}'), ('\u{85}', '\u{85}'), ('\u{a0}', '\u{a0}'),
         ('\u{1680}', '\u{1680}'), ('\u{2000}', '\u{200a}'), ('\u{2028}', '\u{2029}'), ('\u{202f}',

src/libsyntax/lib.rs

@@ -37,6 +37,7 @@ extern crate term;
 extern crate libc;
 #[macro_use] extern crate log;
 #[macro_use] #[no_link] extern crate rustc_bitflags;
+extern crate rustc_unicode;
 
 extern crate serialize as rustc_serialize; // used by deriving
 

src/libsyntax/parse/lexer/comments.rs

@@ -15,7 +15,7 @@ use codemap::{BytePos, CharPos, CodeMap, Pos};
 use errors;
 use parse::lexer::is_block_doc_comment;
 use parse::lexer::{StringReader, TokenAndSpan};
-use parse::lexer::{is_whitespace, Reader};
+use parse::lexer::{is_pattern_whitespace, Reader};
 use parse::lexer;
 use print::pprust;
 use str::char_at;
@@ -153,7 +153,7 @@ fn push_blank_line_comment(rdr: &StringReader, comments: &mut Vec<Comment>) {
 }
 
 fn consume_whitespace_counting_blank_lines(rdr: &mut StringReader, comments: &mut Vec<Comment>) {
-    while is_whitespace(rdr.curr) && !rdr.is_eof() {
+    while is_pattern_whitespace(rdr.curr) && !rdr.is_eof() {
         if rdr.col == CharPos(0) && rdr.curr_is('\n') {
             push_blank_line_comment(rdr, &mut *comments);
         }

src/libsyntax/parse/lexer/mod.rs

@@ -16,6 +16,7 @@ use ext::tt::transcribe::tt_next_token;
 use parse::token::str_to_ident;
 use parse::token;
 use str::char_at;
+use rustc_unicode::property::Pattern_White_Space;
 
 use std::borrow::Cow;
 use std::char;
@@ -546,10 +547,10 @@ impl<'a> StringReader<'a> {
                 let c = self.scan_comment();
                 debug!("scanning a comment {:?}", c);
                 c
-            }
-            c if is_whitespace(Some(c)) => {
+            },
+            c if is_pattern_whitespace(Some(c)) => {
                 let start_bpos = self.last_pos;
-                while is_whitespace(self.curr) {
+                while is_pattern_whitespace(self.curr) {
                     self.bump();
                 }
                 let c = Some(TokenAndSpan {
@@ -1435,7 +1436,7 @@ impl<'a> StringReader<'a> {
     }
 
     fn consume_whitespace(&mut self) {
-        while is_whitespace(self.curr) && !self.is_eof() {
+        while is_pattern_whitespace(self.curr) && !self.is_eof() {
             self.bump();
         }
     }
@@ -1460,7 +1461,7 @@ impl<'a> StringReader<'a> {
     }
 
     fn consume_non_eol_whitespace(&mut self) {
-        while is_whitespace(self.curr) && !self.curr_is('\n') && !self.is_eof() {
+        while is_pattern_whitespace(self.curr) && !self.curr_is('\n') && !self.is_eof() {
             self.bump();
         }
     }
@@ -1591,11 +1592,10 @@ impl<'a> StringReader<'a> {
     }
 }
 
-pub fn is_whitespace(c: Option<char>) -> bool {
-    match c.unwrap_or('\x00') { // None can be null for now... it's not whitespace
-        ' ' | '\n' | '\t' | '\r' => true,
-        _ => false,
-    }
+// This tests the character for the unicode property 'PATTERN_WHITE_SPACE' which
+// is guaranteed to be forward compatible. http://unicode.org/reports/tr31/#R3
+pub fn is_pattern_whitespace(c: Option<char>) -> bool {
+    c.map_or(false, Pattern_White_Space)
 }
 
 fn in_range(c: Option<char>, lo: char, hi: char) -> bool {

src/libsyntax/util/parser_testing.rs

@@ -10,11 +10,11 @@
 
 use ast;
 use parse::{ParseSess,PResult,filemap_to_tts};
-use parse::new_parser_from_source_str;
+use parse::{lexer, new_parser_from_source_str};
 use parse::parser::Parser;
 use parse::token;
 use ptr::P;
-use str::char_at;
+use std::iter::Peekable;
 
 /// Map a string to tts, using a made-up filename:
 pub fn string_to_tts(source_str: String) -> Vec<ast::TokenTree> {
@@ -87,69 +87,62 @@ pub fn strs_to_idents(ids: Vec<&str> ) -> Vec<ast::Ident> {
 
 /// Does the given string match the pattern? whitespace in the first string
 /// may be deleted or replaced with other whitespace to match the pattern.
-/// this function is Unicode-ignorant; fortunately, the careful design of
-/// UTF-8 mitigates this ignorance.  In particular, this function only collapses
-/// sequences of \n, \r, ' ', and \t, but it should otherwise tolerate Unicode
-/// chars. Unsurprisingly, it doesn't do NKF-normalization(?).
+/// This function is relatively Unicode-ignorant; fortunately, the careful design
+/// of UTF-8 mitigates this ignorance. It doesn't do NKF-normalization(?).
 pub fn matches_codepattern(a : &str, b : &str) -> bool {
-    let mut idx_a = 0;
-    let mut idx_b = 0;
+    let mut a_iter = a.chars().peekable();
+    let mut b_iter = b.chars().peekable();
+
     loop {
-        if idx_a == a.len() && idx_b == b.len() {
-            return true;
-        }
-        else if idx_a == a.len() {return false;}
-        else if idx_b == b.len() {
-            // maybe the stuff left in a is all ws?
-            if is_whitespace(char_at(a, idx_a)) {
-                return scan_for_non_ws_or_end(a,idx_a) == a.len();
-            } else {
-                return false;
+        let (a, b) = match (a_iter.peek(), b_iter.peek()) {
+            (None, None) => return true,
+            (None, _) => return false,
+            (Some(&a), None) => {
+                if is_pattern_whitespace(a) {
+                    break // trailing whitespace check is out of loop for borrowck
+                } else {
+                    return false
+                }
             }
-        }
-        // ws in both given and pattern:
-        else if is_whitespace(char_at(a, idx_a))
-           && is_whitespace(char_at(b, idx_b)) {
-            idx_a = scan_for_non_ws_or_end(a,idx_a);
-            idx_b = scan_for_non_ws_or_end(b,idx_b);
-        }
-        // ws in given only:
-        else if is_whitespace(char_at(a, idx_a)) {
-            idx_a = scan_for_non_ws_or_end(a,idx_a);
-        }
-        // *don't* silently eat ws in expected only.
-        else if char_at(a, idx_a) == char_at(b, idx_b) {
-            idx_a += 1;
-            idx_b += 1;
-        }
-        else {
-            return false;
+            (Some(&a), Some(&b)) => (a, b)
+        };
+
+        if is_pattern_whitespace(a) && is_pattern_whitespace(b) {
+            // skip whitespace for a and b
+            scan_for_non_ws_or_end(&mut a_iter);
+            scan_for_non_ws_or_end(&mut b_iter);
+        } else if is_pattern_whitespace(a) {
+            // skip whitespace for a
+            scan_for_non_ws_or_end(&mut a_iter);
+        } else if a == b {
+            a_iter.next();
+            b_iter.next();
+        } else {
+            return false
         }
     }
+
+    // check if a has *only* trailing whitespace
+    a_iter.all(is_pattern_whitespace)
 }
 
-/// Given a string and an index, return the first usize >= idx
-/// that is a non-ws-char or is outside of the legal range of
-/// the string.
-fn scan_for_non_ws_or_end(a : &str, idx: usize) -> usize {
-    let mut i = idx;
-    let len = a.len();
-    while (i < len) && (is_whitespace(char_at(a, i))) {
-        i += 1;
+/// Advances the given peekable `Iterator` until it reaches a non-whitespace character
+fn scan_for_non_ws_or_end<I: Iterator<Item= char>>(iter: &mut Peekable<I>) {
+    while lexer::is_pattern_whitespace(iter.peek().cloned()) {
+        iter.next();
     }
-    i
 }
 
-/// Copied from lexer.
-pub fn is_whitespace(c: char) -> bool {
-    return c == ' ' || c == '\t' || c == '\r' || c == '\n';
+pub fn is_pattern_whitespace(c: char) -> bool {
+    lexer::is_pattern_whitespace(Some(c))
 }
 
 #[cfg(test)]
 mod tests {
     use super::*;
 
-    #[test] fn eqmodws() {
+    #[test]
+    fn eqmodws() {
         assert_eq!(matches_codepattern("",""),true);
         assert_eq!(matches_codepattern("","a"),false);
         assert_eq!(matches_codepattern("a",""),false);
@@ -160,5 +153,22 @@ mod tests {
         assert_eq!(matches_codepattern("a   b","a b"),true);
         assert_eq!(matches_codepattern("ab","a b"),false);
         assert_eq!(matches_codepattern("a   b","ab"),true);
+        assert_eq!(matches_codepattern(" a   b","ab"),true);
+    }
+
+    #[test]
+    fn pattern_whitespace() {
+        assert_eq!(matches_codepattern("","\x0C"), false);
+        assert_eq!(matches_codepattern("a b ","a   \u{0085}\n\t\r  b"),true);
+        assert_eq!(matches_codepattern("a b","a   \u{0085}\n\t\r  b "),false);
+    }
+
+    #[test]
+    fn non_pattern_whitespace() {
+        // These have the property 'White_Space' but not 'Pattern_White_Space'
+        assert_eq!(matches_codepattern("a b","a\u{2002}b"), false);
+        assert_eq!(matches_codepattern("a   b","a\u{2002}b"), false);
+        assert_eq!(matches_codepattern("\u{205F}a   b","ab"), false);
+        assert_eq!(matches_codepattern("a  \u{3000}b","ab"), false);
     }
 }

src/test/run-pass/parser-unicode-whitespace.rs

@@ -0,0 +1,22 @@
+// Copyright 2015 The Rust Project Developers. See the COPYRIGHT
+// file at the top-level directory of this distribution and at
+// http://rust-lang.org/COPYRIGHT.
+//
+// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
+// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
+// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
+// option. This file may not be copied, modified, or distributed
+// except according to those terms.
+
+
+// Beware editing: it has numerous whitespace characters which are important.
+// It contains one ranges from the 'PATTERN_WHITE_SPACE' property outlined in
+// http://unicode.org/Public/UNIDATA/PropList.txt
+//
+// The characters in the first expression of the assertion can be generated
+// from: "4\u{0C}+\n\t\r7\t*\u{20}2\u{85}/\u{200E}3\u{200F}*\u{2028}2\u{2029}"
+pub fn main() {
+assert_eq!(4+
+
+7   * 2…/‎3‏*
2
, 4 + 7 * 2 / 3 * 2);
+}