* Changed string-representation of regexps to handle unprintable chars.

Characters in the range 0..31 and 128..65535 are now printed as \x or \u escapes.
This allows tests written in C++ in a simple codepage to represent the output.
Also avoids output containing \0, which ends string comparison.

A small optimization on quantifiers on empty atoms.

[Lasse Reichstein, 2008-11-17 10:36:20]

Small code review, please.

[Christian Plesner Hansen, 2008-11-17 10:54:35]

Lgtm

http://codereview.chromium.org/11203/diff/1/2
File regexp2000/src/ast.cc (right):

http://codereview.chromium.org/11203/diff/1/2#newcode257
Line 257: static void AddChar(StringStream* stream, uc16 character) {
You could just use StringStream::Add("%k", ...).

http://codereview.chromium.org/11203/diff/1/2#newcode320
Line 320: AddChar(stream(), chardata[i]);
Ditto

http://codereview.chromium.org/11203/diff/1/3
File regexp2000/src/ast.h (right):

http://codereview.chromium.org/11203/diff/1/3#newcode1206
Line 1206: virtual bool Is##Name();
I'm right on the edge of thinking that this is overkill since we don't use most
of these functions.

http://codereview.chromium.org/11203/diff/1/6
File regexp2000/src/parser.cc (right):

http://codereview.chromium.org/11203/diff/1/6#newcode477
Line 477: if (atom->IsLookahead() || atom->IsAssertion()) {
I wonder if this is worthwhile?

[Lasse Reichstein, 2008-11-17 11:26:32]

http://codereview.chromium.org/11203/diff/1/2
File regexp2000/src/ast.cc (right):

http://codereview.chromium.org/11203/diff/1/2#newcode257
Line 257: static void AddChar(StringStream* stream, uc16 character) {
"%k" uses the format "\\x%X" for anything outside of 32..255. 
For this, I prefer a format that has a fixed width, so it doesn't run into
nearby letters (i.e., something that can be parsed uniquely again). I also don't
want the range 128..255 to be treated as literals, since the interpretation
might depend on the code-page.

However, it would probably be better to add a new formatting code to
StringStream.

http://codereview.chromium.org/11203/diff/1/6
File regexp2000/src/parser.cc (right):

http://codereview.chromium.org/11203/diff/1/6#newcode477
Line 477: if (atom->IsLookahead() || atom->IsAssertion()) {
It probably isn't. It's optimizing the case where the user has written a stupid
RegExp to begin with. It might be worthwhile if the following translations can
ignore the pathological cases.

[Christian Plesner Hansen, 2008-11-17 11:37:45]

http://codereview.chromium.org/11203/diff/1/2
File regexp2000/src/ast.cc (right):

http://codereview.chromium.org/11203/diff/1/2#newcode257
Line 257: static void AddChar(StringStream* stream, uc16 character) {
Just change %k to do what you want.  Nothing depends on the current behavior.