Recent changes to feature-requests

Improve token regexp towards JavaScript's full RegExp syntax

Olivier Mengué — Thu, 30 Jul 2009 00:43:04 -0000

JS/CC should use the same regexp syntax as JavaScript ast this is what the user is familiar with.
In particular:
- '\xxx' support should be dropped as it is ambiguous and is not in JavaScript
- '\uXX' and '\UXXXX' should be added for full Unicode support
- '.' should have the same effect (not match end of line)

__parse() should return %% of the top rule

Olivier Mengué — Thu, 30 Jul 2009 00:27:11 -0000

Currently __parse() returns an error count. There is no clean way to return %% from the top rule to the __parse() caller. The only way is to use an ugly global variable. And this makes parsing non thread safe.

Instead __parse() could return an object containing:
- src: the parsed string
- errors: the error list
- result: %% of the top grammar rule
For an attempt to implement this around the current (0.30) JS/CC generated code, see http://registry-files-tools.googlecode.com/svn-history/r3/trunk/msreg.par

Redesigning the Web Environment using a Web Framework

beachcoder — Wed, 29 Jul 2009 13:30:19 -0000

Like the project RXBuild (http://ddlsmurf.github.com/rxbuild/regex.html), based on JS/CC and written by Eric Doughty-Papassideris, an existing Framework (e.g. - like in this case the Yahoo User Interface Library (http://developer.yahoo.com/yui/)) should be used to make the web environment looking more neaty and making it more dynamic (the current version is very static!)).

Unget Input

beachcoder — Wed, 29 Jul 2009 13:25:32 -0000

It must be possible to unget the last input sequence and re-read it within the next parsing action.
This must be provided by a function or method in the parser templates (using the new PCB-feature!)

Multiple token definition in lexer

beachcoder — Wed, 29 Jul 2009 13:23:34 -0000

This is a request from a current support request. JS/CC must be enabled to support multiple tokens for the same input sequence.
Which token is returned must be decided by the semantic code.

Current request is implementing a Phyton-like language, where blocks are defined using tabs.

Example:
'\t'+ INDENT, DEDENT [* if( test for mode ) %token = INDENT; else return %token = DEDENT; *]

The feature of returning a different token must be implemented to support ALL available token!

Implementing default reductions

beachcoder — Tue, 14 Apr 2009 11:35:26 -0000

There must be a a way of providing default reduce actions as it is done in BISON - This is fundamentally required for better error recovery based on the error token (0.32 has very bad bugs!).

Some piece of code:
---------------------------
! ' |\t'

"ZZ"
"YY"
';'
;

slist : slist stmt ';' [* print("slist stmt"); *]
| stmt ';' [* print("stmt"); *]
| ~ ';' [* print("error"); *]
;

stmt : "ZZ" stmt
| "ZZ"
;

_dbg_withtrace = true;
var err_off = new Array();
var err_la = new Array();

var str = "ZZ ZZ YY ZZ ZZ";
print( str );

print( "Errors: " + __parse( str, err_off, err_la ) );
//print( err_off[0] + "#" + err_la[0].join() );