Regex + stackoverflow with Dolphin Pro 5.1.4

Hi,

first of all, I would like to congratulate OA on their great product.
I am looking forward to Dolphin 6.0!

Now the problem I am dealing with. I installed the regex package
(ported from Chris Uppal) on Dolphin Pro 5.1.4 and getting a stack overflow
while evaluating the following expression:
 
| theMatcher |
theMatcher := '^([a-zA-Z0-9_]([a-zA-Z0-9_*:
]|\t)*)(\s*)\((\([^()]*\)|[^()])*\)((
|\n|\t)|(//.*\n)|(/\*([^*]|(\*+[^*/]))*\*+/))*(\s*){'
"^^^^^^ remove any line break in the string ^^^^^^"
asRegex.
theMatcher matches: 'void testFunc(  sint16 *rHWE
                       , uint8 stPwrStgDiaDis
                       , sint16 facBattCorr
                       , uint8 Pwmoutpin1
                       , uint8 Pwmoutpin2
                       , uint16 PwrStgSCB
                       , uint16 PwrStgSCG
                       , uint16 PwrStgOL
                       , uint16 FanCtlrOT
                       , uint16 FanCtlrEE
                       , uint16 Fan1CtlrErr1
                       , uint16 Fan1CtlrErr2
                       , uint16 Fan2CtlrErr1
                       , uint16 Fan2CtlrErr2
                       , uint16 *ErrIntFanCtlrOT
                       , uint16 *ErrIntFanCtlrEE
                       , uint16 *ErrIntFan1CtlrErr1
                       , uint16 *ErrIntFan1CtlrErr2
                       , uint16 *ErrIntFan2CtlrErr1
                       , uint16 *ErrIntFan2CtlrErr2
                       , uint16 *TmrFanCtlrErr
                       , uint16 *TmrFanCtlrNoErr
                       , sint16 tiErrInt
                       , const uint16 *PwmOutCorrStruct
                       , uint16 *stPwrStgErr
                       , uint16 *stPwrStgTested
                       , uint32 tiDiffDSM
                        )
{'

The regex is used to detect c function in source files. To avoid the error
just remove some function parameters. Is it possible to tweak the stack
size so i can stay with my regex?


Thanks

Pierre

Pierre Schwarz wrote:

I can't help with the stack problem, I'm afraid.

I think that your example should have a space (or \s) at the end of the third
line of the regexp ?   Without that the regexp parser won't accept it.
Assuming that change, I find that the test you posted works fine (#matches:
answers false).  I'm using 5.1.4 on WinXP Pro, so I don't know what the
difference between our systems is.

BTW, as an aside, I don't think you'll have much luck creating a regexp that
can reliably recognise C or C++ declarations.  It's just too complicated for
classic regexps, you need something much more like a full parser.  E.g. I don't
/think/ that your expression would work with

   extern void set_callback( void (*callback)() ) ;

(or whatever the damnable syntax is !)   Of course, you may not need that level
of generality.

    -- chris

"Pierre Schwarz" <[hidden email]> wrote in message
news:izpi5iqj9cwk$.epm7vhdivhea$.[hidden email]...
> Hi,
>
> first of all, I would like to congratulate OA on their great product.
> I am looking forward to Dolphin 6.0!

Thanks.

> Now the problem I am dealing with. I installed the regex package
> (ported from Chris Uppal) on Dolphin Pro 5.1.4 and getting a stack
overflow
> while evaluating the following expression:
>  ... [snip] ...
> The regex is used to detect c function in source files. To avoid the error
> just remove some function parameters. Is it possible to tweak the stack
> size so i can stay with my regex?
>

I wouldn't want to denigrate Chris' package, but as a port of Vassili
Bykov's regex implementation it was originally designed on the assumption of
an unbounded stack. Although you can increase the stack size in Dolphin
(see the comment in the Process class>>initialize method), I suspect you may
find that this will provide only temporary relief, and that you will still
manage to blow the stack with more larger or more complex inputs, or with a
different regex.

At OA we use a wrapper around the Windows Scripting regexp component (used
to implement regexp in Microsoft's JScript and VBScript). This doesn't
suffer from such issues, and is also considerably faster. The component will
be installed on any machine which meets Dolphin's own prerequisite
requirements. The package (extracted from Dolphin 6) can be downloaded from:

http://www.object-arts.com/Lib/Downloads/RegExp.zip

I tested it out with your example...

theMatcher := IRegExp2 new.
theMatcher pattern: ^([a-zA-Z0-9_]([a-zA-Z0-9_*: ..... etc...
theMatcher test: ...etc...

And it returned false. This is either due to differences in the regexp
syntax (see first link below) or perhaps an error in your regexp.

I've pasted some other useful links below.

Regards

Blair
----------------

1) "Regular Expression Syntax"
http://msdn.microsoft.com/library/default.asp?url=/library/en-us/script56/html/js56jsgrpRegExpSyntax.asp
2) "Introduction to Regular Expressions"
http://msdn.microsoft.com/library/default.asp?url=/library/en-us/script56/html/js56reconIntroductionToRegularExpressions.asp
3) "Microsoft Beefs Up VBScript with Regular Expressions"
http://msdn.microsoft.com/library/en-us/dnclinic/html/scripting051099.asp.
4) KB818802 "HOW TO: Use Regular Expressions in Microsoft Visual Basic 6.0"
http://support.microsoft.com/default.aspx?scid=kb%3Ben-us%3B818802
5) Microsoft Windows Script Downloads
http://msdn.microsoft.com/library/default.asp?url=/downloads/list/webdev.asp

Thanks for the package. I generated also an interface with the ActiveX
Wizard on the "Microsoft VBScript Regular Expressions 5.5" component
(Thanks to Andy for the hint).

> I tested it out with your example...
>
> theMatcher := IRegExp2 new.
> theMatcher pattern: ^([a-zA-Z0-9_]([a-zA-Z0-9_*: ..... etc...
> theMatcher test: ...etc...
>
> And it returned false. This is either due to differences in the regexp
> syntax (see first link below) or perhaps an error in your regexp.
>

Yes, the syntax divers from the ActiveX control and the regex package.

Thanks for the links.

At last, thanks to the regulars in this newsgroup. I appreciate the work
and time they put into this newsgroup. With DSDN from Ian it is a prefect
source of wisdom.

Pierre

<snip>

Chris Uppal wrote:
Yes, i know, it not possible to use a regex to detect every C declarations.
In my case i need to detect only very simple function calls e.g.
 
"void functionName(void) {..."

With a litte rework on the regex and the use of the ActiveX component Blair
mentioned, i can parse 900 source files in 10 seconds. So, at the moment i
can stay with the ActiveX solution.

But for further requirements, has anybody a full c parser running in
smalltalk (e.g. /SmaCC/)?

Pierre