regex: Remove in-tree version

The regex library was largely used for non-critical aspects of the compiler and
various external tooling. The library at this point is duplicated with its
out-of-tree counterpart and as such imposes a bit of a maintenance overhead as
well as compile time hit for the compiler itself.

The last major user of the regex library is the libtest library, using regexes
for filters when running tests. This removal means that the filtering has gone
back to substring matching rather than using regexes.

Author: alexcrichton

mk/crates.mk

@@ -51,7 +51,7 @@
 
 TARGET_CRATES := libc std flate arena term \
                  serialize getopts collections test rand \
-                 log regex graphviz core rbml alloc \
+                 log graphviz core rbml alloc \
                  unicode rustc_bitflags
 RUSTC_CRATES := rustc rustc_typeck rustc_borrowck rustc_resolve rustc_driver \
                 rustc_trans rustc_back rustc_llvm rustc_privacy
@@ -95,16 +95,15 @@ DEPS_term := std log
 DEPS_getopts := std
 DEPS_collections := core alloc unicode
 DEPS_num := std
-DEPS_test := std getopts serialize rbml term regex native:rust_test_helpers
+DEPS_test := std getopts serialize rbml term native:rust_test_helpers
 DEPS_rand := core
-DEPS_log := std regex
-DEPS_regex := std
+DEPS_log := std
 DEPS_fmt_macros = std
 
 TOOL_DEPS_compiletest := test getopts
 TOOL_DEPS_rustdoc := rustdoc
 TOOL_DEPS_rustc := rustc_driver
-TOOL_DEPS_rustbook := std regex rustdoc
+TOOL_DEPS_rustbook := std rustdoc
 TOOL_SOURCE_compiletest := $(S)src/compiletest/compiletest.rs
 TOOL_SOURCE_rustdoc := $(S)src/driver/driver.rs
 TOOL_SOURCE_rustc := $(S)src/driver/driver.rs
@@ -130,9 +129,8 @@ DOC_CRATES := $(filter-out rustc, \
               $(filter-out rustc_driver, \
               $(filter-out rustc_privacy, \
               $(filter-out log, \
-              $(filter-out regex, \
               $(filter-out getopts, \
-              $(filter-out syntax, $(CRATES))))))))))))
+              $(filter-out syntax, $(CRATES)))))))))))
 COMPILER_DOC_CRATES := rustc rustc_trans rustc_borrowck rustc_resolve \
                        rustc_typeck rustc_driver syntax rustc_privacy
 

src/compiletest/common.rs

@@ -11,7 +11,6 @@ pub use self::Mode::*;
 
 use std::fmt;
 use std::str::FromStr;
-use regex::Regex;
 
 #[derive(Clone, PartialEq, Debug)]
 pub enum Mode {
@@ -101,10 +100,7 @@ pub struct Config {
     pub run_ignored: bool,
 
     // Only run tests that match this filter
-    pub filter: Option<Regex>,
-
-    // Precompiled regex for finding expected errors in cfail
-    pub cfail_regex: Regex,
+    pub filter: Option<String>,
 
     // Write out a parseable log of tests that were run
     pub logfile: Option<Path>,

src/compiletest/compiletest.rs

@@ -22,7 +22,6 @@ extern crate getopts;
 
 #[macro_use]
 extern crate log;
-extern crate regex;
 
 use std::os;
 use std::io;
@@ -33,7 +32,6 @@ use getopts::{optopt, optflag, reqopt};
 use common::Config;
 use common::{Pretty, DebugInfoGdb, DebugInfoLldb, Codegen};
 use util::logv;
-use regex::Regex;
 
 pub mod procsrv;
 pub mod util;
@@ -116,14 +114,7 @@ pub fn parse_config(args: Vec<String> ) -> Config {
     }
 
     let filter = if !matches.free.is_empty() {
-        let s = matches.free[0].as_slice();
-        match regex::Regex::new(s) {
-            Ok(re) => Some(re),
-            Err(e) => {
-                println!("failed to parse filter /{}/: {:?}", s, e);
-                panic!()
-            }
-        }
+        Some(matches.free[0].clone())
     } else {
         None
     };
@@ -145,7 +136,6 @@ pub fn parse_config(args: Vec<String> ) -> Config {
                                        .as_slice()).expect("invalid mode"),
         run_ignored: matches.opt_present("ignored"),
         filter: filter,
-        cfail_regex: Regex::new(errors::EXPECTED_PATTERN).unwrap(),
         logfile: matches.opt_str("logfile").map(|s| Path::new(s)),
         runtool: matches.opt_str("runtool"),
         host_rustcflags: matches.opt_str("host-rustcflags"),
@@ -374,18 +364,24 @@ fn extract_gdb_version(full_version_line: Option<String>) -> Option<String> {
           if full_version_line.as_slice().trim().len() > 0 => {
             let full_version_line = full_version_line.as_slice().trim();
 
-            let re = Regex::new(r"(^|[^0-9])([0-9]\.[0-9])([^0-9]|$)").unwrap();
-
-            match re.captures(full_version_line) {
-                Some(captures) => {
-                    Some(captures.at(2).unwrap_or("").to_string())
+            // used to be a regex "(^|[^0-9])([0-9]\.[0-9])([^0-9]|$)"
+            for (pos, c) in full_version_line.char_indices() {
+                if !c.is_digit(10) { continue }
+                if pos + 2 >= full_version_line.len() { continue }
+                if full_version_line.char_at(pos + 1) != '.' { continue }
+                if !full_version_line.char_at(pos + 2).is_digit(10) { continue }
+                if pos > 0 && full_version_line.char_at_reverse(pos).is_digit(10) {
+                    continue
                 }
-                None => {
-                    println!("Could not extract GDB version from line '{}'",
-                             full_version_line);
-                    None
+                if pos + 3 < full_version_line.len() &&
+                   full_version_line.char_at(pos + 3).is_digit(10) {
+                    continue
                 }
+                return Some(full_version_line[pos..pos+3].to_string());
             }
+            println!("Could not extract GDB version from line '{}'",
+                     full_version_line);
+            None
         },
         _ => None
     }
@@ -408,18 +404,26 @@ fn extract_lldb_version(full_version_line: Option<String>) -> Option<String> {
           if full_version_line.as_slice().trim().len() > 0 => {
             let full_version_line = full_version_line.as_slice().trim();
 
-            let re = Regex::new(r"[Ll][Ll][Dd][Bb]-([0-9]+)").unwrap();
-
-            match re.captures(full_version_line) {
-                Some(captures) => {
-                    Some(captures.at(1).unwrap_or("").to_string())
-                }
-                None => {
-                    println!("Could not extract LLDB version from line '{}'",
-                             full_version_line);
-                    None
-                }
+            for (pos, l) in full_version_line.char_indices() {
+                if l != 'l' && l != 'L' { continue }
+                if pos + 5 >= full_version_line.len() { continue }
+                let l = full_version_line.char_at(pos + 1);
+                if l != 'l' && l != 'L' { continue }
+                let d = full_version_line.char_at(pos + 2);
+                if d != 'd' && d != 'D' { continue }
+                let b = full_version_line.char_at(pos + 3);
+                if b != 'b' && b != 'B' { continue }
+                let dash = full_version_line.char_at(pos + 4);
+                if dash != '-' { continue }
+
+                let vers = full_version_line[pos + 5..].chars().take_while(|c| {
+                    c.is_digit(10)
+                }).collect::<String>();
+                if vers.len() > 0 { return Some(vers) }
             }
+            println!("Could not extract LLDB version from line '{}'",
+                     full_version_line);
+            None
         },
         _ => None
     }

src/compiletest/errors.rs

@@ -9,32 +9,26 @@
 // except according to those terms.
 use self::WhichLine::*;
 
-use std::ascii::AsciiExt;
 use std::io::{BufferedReader, File};
-use regex::Regex;
 
 pub struct ExpectedError {
     pub line: uint,
     pub kind: String,
     pub msg: String,
 }
 
+#[derive(PartialEq, Show)]
+enum WhichLine { ThisLine, FollowPrevious(uint), AdjustBackward(uint) }
+
 /// Looks for either "//~| KIND MESSAGE" or "//~^^... KIND MESSAGE"
 /// The former is a "follow" that inherits its target from the preceding line;
 /// the latter is an "adjusts" that goes that many lines up.
 ///
 /// Goal is to enable tests both like: //~^^^ ERROR go up three
 /// and also //~^ ERROR message one for the preceding line, and
 ///          //~| ERROR message two for that same line.
-
-pub static EXPECTED_PATTERN : &'static str =
-    r"//~(?P<follow>\|)?(?P<adjusts>\^*)\s*(?P<kind>\S*)\s*(?P<msg>.*)";
-
-#[derive(PartialEq, Show)]
-enum WhichLine { ThisLine, FollowPrevious(uint), AdjustBackward(uint) }
-
 // Load any test directives embedded in the file
-pub fn load_errors(re: &Regex, testfile: &Path) -> Vec<ExpectedError> {
+pub fn load_errors(testfile: &Path) -> Vec<ExpectedError> {
     let mut rdr = BufferedReader::new(File::open(testfile).unwrap());
 
     // `last_nonfollow_error` tracks the most recently seen
@@ -50,7 +44,7 @@ pub fn load_errors(re: &Regex, testfile: &Path) -> Vec<ExpectedError> {
     rdr.lines().enumerate().filter_map(|(line_no, ln)| {
         parse_expected(last_nonfollow_error,
                        line_no + 1,
-                       ln.unwrap().as_slice(), re)
+                       ln.unwrap().as_slice())
             .map(|(which, error)| {
                 match which {
                     FollowPrevious(_) => {}
@@ -63,30 +57,39 @@ pub fn load_errors(re: &Regex, testfile: &Path) -> Vec<ExpectedError> {
 
 fn parse_expected(last_nonfollow_error: Option<uint>,
                   line_num: uint,
-                  line: &str,
-                  re: &Regex) -> Option<(WhichLine, ExpectedError)> {
-    re.captures(line).and_then(|caps| {
-        let adjusts = caps.name("adjusts").unwrap_or("").len();
-        let kind = caps.name("kind").unwrap_or("").to_ascii_lowercase();
-        let msg = caps.name("msg").unwrap_or("").trim().to_string();
-        let follow = caps.name("follow").unwrap_or("").len() > 0;
+                  line: &str) -> Option<(WhichLine, ExpectedError)> {
+    let start = match line.find_str("//~") { Some(i) => i, None => return None };
+    let (follow, adjusts) = if line.char_at(start + 3) == '|' {
+        (true, 0)
+    } else {
+        (false, line[start + 3..].chars().take_while(|c| *c == '^').count())
+    };
+    let kind_start = start + 3 + adjusts + (follow as usize);
+    let letters = line[kind_start..].chars();
+    let kind = letters.skip_while(|c| c.is_whitespace())
+                      .take_while(|c| !c.is_whitespace())
+                      .map(|c| c.to_lowercase())
+                      .collect::<String>();
+    let letters = line[kind_start..].chars();
+    let msg = letters.skip_while(|c| c.is_whitespace())
+                     .skip_while(|c| !c.is_whitespace())
+                     .collect::<String>().trim().to_string();
 
-        let (which, line) = if follow {
-            assert!(adjusts == 0, "use either //~| or //~^, not both.");
-            let line = last_nonfollow_error.unwrap_or_else(|| {
-                panic!("encountered //~| without preceding //~^ line.")
-            });
-            (FollowPrevious(line), line)
-        } else {
-            let which =
-                if adjusts > 0 { AdjustBackward(adjusts) } else { ThisLine };
-            let line = line_num - adjusts;
-            (which, line)
-        };
+    let (which, line) = if follow {
+        assert!(adjusts == 0, "use either //~| or //~^, not both.");
+        let line = last_nonfollow_error.unwrap_or_else(|| {
+            panic!("encountered //~| without preceding //~^ line.")
+        });
+        (FollowPrevious(line), line)
+    } else {
+        let which =
+            if adjusts > 0 { AdjustBackward(adjusts) } else { ThisLine };
+        let line = line_num - adjusts;
+        (which, line)
+    };
 
-        debug!("line={} which={:?} kind={:?} msg={:?}", line_num, which, kind, msg);
-        Some((which, ExpectedError { line: line,
-                                     kind: kind,
-                                     msg: msg, }))
-    })
+    debug!("line={} which={:?} kind={:?} msg={:?}", line_num, which, kind, msg);
+    Some((which, ExpectedError { line: line,
+                                 kind: kind,
+                                 msg: msg, }))
 }

src/compiletest/runtest.rs

@@ -99,7 +99,7 @@ fn run_cfail_test(config: &Config, props: &TestProps, testfile: &Path) {
     }
 
     let output_to_check = get_output(props, &proc_res);
-    let expected_errors = errors::load_errors(&config.cfail_regex, testfile);
+    let expected_errors = errors::load_errors(testfile);
     if !expected_errors.is_empty() {
         if !props.error_patterns.is_empty() {
             fatal("both error pattern and expected errors specified");

src/grammar/verify.rs

@@ -13,14 +13,11 @@
 extern crate syntax;
 extern crate rustc;
 
-extern crate regex;
-
 #[macro_use]
 extern crate log;
 
 use std::collections::HashMap;
 use std::io::File;
-use regex::Regex;
 
 use syntax::parse;
 use syntax::parse::lexer;
@@ -167,15 +164,19 @@ fn count(lit: &str) -> usize {
 }
 
 fn parse_antlr_token(s: &str, tokens: &HashMap<String, token::Token>) -> TokenAndSpan {
-    let re = Regex::new(
-      r"\[@(?P<seq>\d+),(?P<start>\d+):(?P<end>\d+)='(?P<content>.+?)',<(?P<toknum>-?\d+)>,\d+:\d+]"
-    ).unwrap();
-
-    let m = re.captures(s).expect(format!("The regex didn't match {}", s).as_slice());
-    let start = m.name("start").unwrap_or("");
-    let end = m.name("end").unwrap_or("");
-    let toknum = m.name("toknum").unwrap_or("");
-    let content = m.name("content").unwrap_or("");
+    // old regex:
+    // \[@(?P<seq>\d+),(?P<start>\d+):(?P<end>\d+)='(?P<content>.+?)',<(?P<toknum>-?\d+)>,\d+:\d+]
+    let start = s.find_str("[@").unwrap();
+    let comma = start + s[start..].find_str(",").unwrap();
+    let colon = comma + s[comma..].find_str(":").unwrap();
+    let content_start = colon + s[colon..].find_str("='").unwrap();
+    let content_end = content_start + s[content_start..].find_str("',<").unwrap();
+    let toknum_end = content_end + s[content_end..].find_str(">,").unwrap();
+
+    let start = &s[comma + 1 .. colon];
+    let end = &s[colon + 1 .. content_start];
+    let content = &s[content_start + 2 .. content_end];
+    let toknum = &s[content_end + 3 .. toknum_end];
 
     let proto_tok = tokens.get(toknum).expect(format!("didn't find token {:?} in the map",
                                                               toknum).as_slice());

src/liblog/directive.rs

@@ -8,7 +8,6 @@
 // option. This file may not be copied, modified, or distributed
 // except according to those terms.
 
-use regex::Regex;
 use std::ascii::AsciiExt;
 use std::cmp;
 
@@ -34,7 +33,7 @@ fn parse_log_level(level: &str) -> Option<u32> {
 ///
 /// Valid log levels are 0-255, with the most likely ones being 1-4 (defined in
 /// std::).  Also supports string log levels of error, warn, info, and debug
-pub fn parse_logging_spec(spec: &str) -> (Vec<LogDirective>, Option<Regex>) {
+pub fn parse_logging_spec(spec: &str) -> (Vec<LogDirective>, Option<String>) {
     let mut dirs = Vec::new();
 
     let mut parts = spec.split('/');
@@ -80,17 +79,7 @@ pub fn parse_logging_spec(spec: &str) -> (Vec<LogDirective>, Option<Regex>) {
         });
     }});
 
-    let filter = filter.map_or(None, |filter| {
-        match Regex::new(filter) {
-            Ok(re) => Some(re),
-            Err(e) => {
-                println!("warning: invalid regex filter - {:?}", e);
-                None
-            }
-        }
-    });
-
-    return (dirs, filter);
+    (dirs, filter.map(|s| s.to_string()))
 }
 
 #[cfg(test)]