Don't insert .*? at the beginning of anchored expressions.

Don't insert .*? at the beginning of anchored expressions.

[Lasse Reichstein, 2009-01-22 15:29:52]

LGTM, with nitpicks.

http://codereview.chromium.org/18497/diff/1/2
File src/ast.cc (right):

http://codereview.chromium.org/18497/diff/1/2#newcode264
Line 264: }
It's probably useless in practice, but a capture or a positive look-ahead with
an anchored body is also anchored.

http://codereview.chromium.org/18497/diff/1/2#newcode439
Line 439: : alternatives_(alternatives) {
Could you add an "ASSERT(alternatives->length() > 1);" here too.

http://codereview.chromium.org/18497/diff/1/2#newcode453
Line 453: ASSERT(nodes->length() > 0);
Asserting that it's > 1 would be even better. The parser avoids creating an
alternative if there is only one node in it (ditto for disjunctions).

http://codereview.chromium.org/18497/diff/1/5
File test/mjsunit/ascii-regexp-subject.js (right):

http://codereview.chromium.org/18497/diff/1/5#newcode31
Line 31: 
Is this "test" falsifiable?
Should it timeout if we aren't optimized? If we just test that "^" works, we
should have some asserts, and no need for a loop.
It looks more like a benchmark than a test. I suggest dropping it or creating a
benchmark.

[Christian Plesner Hansen, 2009-01-23 07:41:18]

http://codereview.chromium.org/18497/diff/1/2
File src/ast.cc (right):

http://codereview.chromium.org/18497/diff/1/2#newcode264
Line 264: }
That's pretty obscure but I've added it anyway.

The same thing holds for the term following an initial  assertion but that's not
worth the trouble to implement.

http://codereview.chromium.org/18497/diff/1/2#newcode439
Line 439: : alternatives_(alternatives) {
Added

http://codereview.chromium.org/18497/diff/1/2#newcode453
Line 453: ASSERT(nodes->length() > 0);
Fixed

http://codereview.chromium.org/18497/diff/1/5
File test/mjsunit/ascii-regexp-subject.js (right):

http://codereview.chromium.org/18497/diff/1/5#newcode31
Line 31: 
It is indeed more like a benchmark but I'll leave it to Erik to create that. 
I'm keeping the test because something, even if it's pretty poor, is better than
nothing.  Also, our friend the arm simulator makes this test falsifiable on the
build bot.

[Erik Corry, 2009-01-23 13:26:17]

LGTM

http://codereview.chromium.org/18497/diff/1/2
File src/ast.cc (right):

http://codereview.chromium.org/18497/diff/1/2#newcode248
Line 248: return type() == RegExpAssertion::START_OF_INPUT;
Hey, if you're adding obscure cases then this should return true if the
successor returns true.

http://codereview.chromium.org/18497/diff/1/5
File test/mjsunit/ascii-regexp-subject.js (right):

http://codereview.chromium.org/18497/diff/1/5#newcode31
Line 31: 
On 2009/01/22 15:29:52, Lasse Reichstein wrote:
> Is this "test" falsifiable?

Sure it's falsifiable.  That's how we found this bug :-)