Stand-alone parser template.

Stand-alone parser template. 
Uses existing Scanner and ParserLog. 
Generates same preparse-data as existing preparser.

Committed: http://code.google.com/p/v8/source/detail?r=5750

[Mads Ager (chromium), 2010-11-01 14:04:27]

http://codereview.chromium.org/4112012/diff/3001/4001
File src/parser.cc (right):

http://codereview.chromium.org/4112012/diff/3001/4001#newcode47
src/parser.cc:47: #include "preparser.h"
Can you all this after platform.h to alphabetize?

http://codereview.chromium.org/4112012/diff/3001/4001#newcode4937
src/parser.cc:4937: class LogAdapter : public ParserLog {
Why do you need to adapt instead of just changing the interface?

http://codereview.chromium.org/4112012/diff/3001/4001#newcode4945
src/parser.cc:4945: void LogMessage(int start,
blank line before LogMessage?

http://codereview.chromium.org/4112012/diff/3001/4002
File src/parser.h (right):

http://codereview.chromium.org/4112012/diff/3001/4002#newcode184
src/parser.h:184: // Records the occurrence of a function.
Only two of the many methods here have comments. Either add comments to all of
them or don't comment on any of them?

http://codereview.chromium.org/4112012/diff/3001/4002#newcode188
src/parser.h:188: virtual void LogError() { }
Am I missing an implementation of this? I don't see this implemented anywhere,
shouldn't it be?

http://codereview.chromium.org/4112012/diff/3001/4002#newcode204
src/parser.h:204: PartialParserRecorder();
Since you have spaces between the rest of the functions, should you have a blank
line after the constructor?

http://codereview.chromium.org/4112012/diff/3001/4002#newcode231
src/parser.h:231: bool is_recording() {
blank line before?

http://codereview.chromium.org/4112012/diff/3001/4002#newcode243
src/parser.h:243: int prev_start;
Missing underscore: prev_start_

Where is this member used?

http://codereview.chromium.org/4112012/diff/3001/4003
File src/preparser.h (right):

http://codereview.chromium.org/4112012/diff/3001/4003#newcode75
src/preparser.h:75: 
Excessive spacing.

http://codereview.chromium.org/4112012/diff/3001/4003#newcode80
src/preparser.h:80: class PreParser {
Do we have tests that are good enough to catch if the real parser and the
pre-parser do not evolve in the same way?

http://codereview.chromium.org/4112012/diff/3001/4003#newcode253
src/preparser.h:253: template <typename S, typename L>
S -> Scanner
L -> Logger

To improve readability? Also in the rest of the file.

http://codereview.chromium.org/4112012/diff/3001/4004
File test/cctest/test-parsing.cc (right):

http://codereview.chromium.org/4112012/diff/3001/4004#newcode267
test/cctest/test-parsing.cc:267: void LogMessage(int start, int end, const char*
type, const char* arg_opt) {
Blank line before LogMessage?

[Lasse Reichstein, 2010-11-01 15:35:54]

http://codereview.chromium.org/4112012/diff/3001/4001
File src/parser.cc (right):

http://codereview.chromium.org/4112012/diff/3001/4001#newcode47
src/parser.cc:47: #include "preparser.h"
Done.

http://codereview.chromium.org/4112012/diff/3001/4001#newcode4937
src/parser.cc:4937: class LogAdapter : public ParserLog {
Mainly to change the existing parser as little as possible ... for now.
I expect a largish cleanup of the real parser to remove the preparsing
(including the ParserLog class which is the one that does nothing, virtually,
when parsing).
This is, however, a small change that should be fairly easy to do now. I'll
remove the adapter and add the functions directly to PreParserLog.

http://codereview.chromium.org/4112012/diff/3001/4001#newcode4945
src/parser.cc:4945: void LogMessage(int start,
Done, and moved.

http://codereview.chromium.org/4112012/diff/3001/4002
File src/parser.h (right):

http://codereview.chromium.org/4112012/diff/3001/4002#newcode184
src/parser.h:184: // Records the occurrence of a function.
Comments added.

http://codereview.chromium.org/4112012/diff/3001/4002#newcode188
src/parser.h:188: virtual void LogError() { }
I'm pretty sure it's not used. I don't know if it ever was.
I'll remove it.

http://codereview.chromium.org/4112012/diff/3001/4002#newcode204
src/parser.h:204: PartialParserRecorder();
Space added.

http://codereview.chromium.org/4112012/diff/3001/4002#newcode231
src/parser.h:231: bool is_recording() {
Done.

http://codereview.chromium.org/4112012/diff/3001/4002#newcode243
src/parser.h:243: int prev_start;
It's only used in debug mode, to ensure that incoming logs have monotonically
increasing start points.

http://codereview.chromium.org/4112012/diff/3001/4003
File src/preparser.h (right):

http://codereview.chromium.org/4112012/diff/3001/4003#newcode75
src/preparser.h:75: 
Fixed.

http://codereview.chromium.org/4112012/diff/3001/4003#newcode80
src/preparser.h:80: class PreParser {
Probably not.
We should either collect a corpus of JavaScript files that the two can be tested
against, or use a syntax fuzzer and compare the behavior. 
I did some static testing on all our natives files and on a lot of benchmark
files I had available, to test that the old and new preparsing generated the
same preparser data, but I didn't want to add it as a unit-test, since it relied
on JavaScript code that isn't stable.

http://codereview.chromium.org/4112012/diff/3001/4003#newcode253
src/preparser.h:253: template <typename S, typename L>
Done. <S, L> -> <Scanner, Log>.

http://codereview.chromium.org/4112012/diff/3001/4004
File test/cctest/test-parsing.cc (right):

http://codereview.chromium.org/4112012/diff/3001/4004#newcode267
test/cctest/test-parsing.cc:267: void LogMessage(int start, int end, const char*
type, const char* arg_opt) {
Adapter class removed after moving the functions to PartialParserRecorder.

[Lasse Reichstein, 2010-11-02 07:08:51]

Please review changes.

[Mads Ager (chromium), 2010-11-02 07:17:03]

LGTM

http://codereview.chromium.org/4112012/diff/10001/11003
File src/preparser.h (right):

http://codereview.chromium.org/4112012/diff/10001/11003#newcode245
src/preparser.h:245: if (!*ok) return -1; \
Align the two '\'