Resolve #17 optimize use of '^' / '\A'

The problem was that a regular expression like "^x" would take linear time to is_match a string "aaa...." when all it needed to do was check the first character then bail out. This commit makes it so if the regex begins with '^' (without multiline) or '\A', then it stops iterating over characters as soon as there are no possible matches. Benchmarks: dynamic: anchored_literal_long_non_match: 7492 ns/iter => 435 ns/iter anchored_literal_short_non_match: 917 ns/iter => 431 ns/iter native: anchored_literal_long_non_match: 5020 ns/iter => 50 ns/iter anchored_literal_short_non_match: 375 ns/iter => 50 ns/iter

Author: tmerr

regex_macros/src/lib.rs

@@ -191,6 +191,11 @@ fn exec<'t>(which: ::regex::native::MatchKind, input: &'t str,
  if matched {
  break
  }
+
+ if $prefix_anchor && self.ic != 0 {
+ break
+ }
+
  $check_prefix
  }
  if clist.size == 0 || (!$prefix_anchor && !matched) {

src/vm.rs

@@ -144,12 +144,18 @@ impl<'r, 't> Nfa<'r, 't> {
  break
  }
 
+ // If the expression starts with a '^' we can terminate as soon
+ // as the last thread dies.
+ if self.ic != 0 && prefix_anchor {
+ break;
+ }
+ 
  // If there are no threads to try, then we'll have to start
  // over at the beginning of the regex.
  // BUT, if there's a literal prefix for the program, try to
  // jump ahead quickly. If it can't be found, then we can bail
  // out early.
- if self.prog.prefix.len() > 0 && clist.size == 0 {
+ if self.prog.prefix.len() > 0 {
  let needle = self.prog.prefix.as_bytes();
  let haystack = &self.input.as_bytes()[self.ic..];
  match find_prefix(needle, haystack) {