[Solved] StringRegExp() recursion.

[MvGulik]

Just wondering if its possible to split strings like this "[1,0],[[2,0],[[0,0]]],[[3,0]],etc" with StringRegExp() into:

array[0]="[1,0]"

array[2]="[[2,0],[[0,0]]]"

array[3]="[[3,0]]"

---

[edit] small title change.

Edited March 1, 2010 by MvGulik

[SmOke_N]

If the number is consistent ( always follows the [[0,0]]] rule for 2 dimensions ( of course we're really left to assume what it is you're trying to do), then yes... it's possible.

[MvGulik]

no fixed dimention depth.

no fixed dimention size.

only fixed thing is that every [..] part is a one dimentional array.

so [[2,0],[[1,2,3]]] is equivalent to: array[ array[2,0], array[ array[1,2,3] ] ]

trying to see if StringRegExp() can unwrap this while keeping the right [..] balance.

[memoryoverflow]

It is possible, but a bit complex, since it needs things like look ahead & recursion and the square brackets \[, \] make it look confusing. I don't think that someone, who doesn't have a similar pattern ready 'on his shelf', can write it in a few minutes, even with some experience.

[Heron]

If the number is consistent ( always follows the [[0,0]]] rule for 2 dimensions ( of course we're really left to assume what it is you're trying to do), then yes... it's possible.

I don't think it's possible, at least not with one statement. You'll have to use, a limited count, of submatches. Indeed we don't know what your aim is, if you can change the notation you could do it with StringSplit().

- Heron -

[MvGulik]

Complex was kinda expected.

But if its possible, than I could give it at leased a other attempt.

---

just interest in how far StringRegExp() can be pushed in this case.

I already have a alternative function that is doing the job.

Edited October 21, 2009 by MvGulik

[Nutster]

Regular expression are primarily set up to find linear patterns, not nested ones. I will do a little research to see if I can figure this out.

[Authenticity]

Hope it's correct:

#include <Array.au3>

Local $sString = "[1,0],[[2,0],[[0,0]]],[[3,0]]"
Local $avMatches = StringRegExp($sString, '(?x)(\[ (?: [^[\]]++ | (?R) )* \])', 3)

If IsArray($avMatches) Then _ArrayDisplay($avMatches)

[MvGulik]

Oww, nice one.

Can almost read how it works without additional doc.

Think its perfect.

Actually, it is.

Time to take a closer look at it.

Thanks for the help and info.

[SmOke_N]

Hope it's correct:

#include <Array.au3>

Local $sString = "[1,0],[[2,0],[[0,0]]],[[3,0]]"
Local $avMatches = StringRegExp($sString, '(?x)(\[ (?: [^[\]]++ | (?R) )* \])', 3)

If IsArray($avMatches) Then _ArrayDisplay($avMatches)

You know, now that I see your string in a code box, I see the pattern ( without looking at your code ). Then when I look at your pattern, I am thankful I was actually working before I attempted it (Very nice). Edited October 21, 2009 by SmOke_N
ebonics

[Authenticity]

This one almost always matches the things surrounding pattern recursion. It's publicly visible on PCRE's man page. The way it differs from a function call's nested arguments is that it's now a single dimension array, but again, the only difference is that you're trying to match nested array elements to an arbitrary depth.

Edit: I guess a simple change would make it visible:

(1,0),((2,0),((0,0))),((3,0))

Another non recursive way is to increase a counter for the opening parentheses and decreasing the counter when the closing parentheses are encountered. Like the calculator, 1+(2+3*(4-5)), the second it encounters the opening ( it increases a variable to 1 then to 2, then decreases to 1 and then 0, so it's a valid expression. The rest is the language elements tokenization I guess.

Edited October 21, 2009 by Authenticity

[MvGulik]

mm, I need to read better.

RECURSIVE PATTERNS

...

This PCRE pattern solves the nested parentheses problem (assume the

PCRE_EXTENDED option is set so that white space is ignored):

\( ( [^()]++ | (?R) )* \)

[MvGulik]

Trying to get recursion to work on something longer than one character.

Tried to substitute (?>[^U]+) with different things, like (?!REM+), but none of them did any capturing so far.

$sSource = 'aa X bb X cc Z bb Z aa X bb X cc Z bb Z aa'
        $sPattern = "(?x)  (?'pn'  X  (?:  (?>[^XZ]+) | (?&pn)  )*  Z  )" ;; YES

        $sSource = 'aa UV bb UV cc ZY bb ZY aa UV bb UV cc ZY bb ZY aa'
        $sPattern = "(?x)  (?'pn'  UV  (?:  (?>[^UZ]+) | (?&pn)  )*  ZY  )" ;; YES

;~      $sSource = 'aa UV bb UV cc UY bb UY aa UV bb UV cc UY bb UY aa'
;~      $sPattern = "(?x)  (?'pn'  UV  (?:  (?>[^U]+) | (?&pn)  )*  UY  )" ;; YES: ???

        $sSource = 'aa REMstart bb REMstart cc REMend bb REMend aa REMstart bb REMstart cc REMend bb REMend aa'
        $sPattern = "(?x)  (?'pn'  REMstart  (?:  (?>[^R]+) | (?&pn)  )*  REMend  )" ;; YES: but unsave.

;~      $sSource = 'aa REMstart bb REMstart cc REMend bb REMend aa REMstart bb REMstart cc REMend bb REMend aa'
;~      $sPattern = "(?x)  (?'pn'  REMstart  (?:  (?...???... ) | (?&pn)  )*  REMend  )" ;; ...

Target is of course #cs...#ce.

[GEOSoft]

No wonder I can never get (\R) to work. It should be (?R)

[Authenticity]

#include <Array.au3>

$sSource = 'aa REMstart bb REMstart cc REMend bb REMend ee REMstart ff REMstart gg REMend hh REMend ii'
$sPattern = "(?x)  (REMstart  (?: (?>(?s).*?(?=REMstart|REMend)) | (?1)  )*  REMend  )" ;; ...

$aMatches = StringRegExp($sSource, $sPattern, 3)
_ArrayDisplay($aMatches)

One thing I'm asking now is why #cs followed by another #cs requires two #ce? Isn't anything between #cs to #ce must be commented out?

[MvGulik]

One thing I'm asking now is why #cs followed by another #cs requires two #ce?

Not sure if your asking me, as its sound more like a questions for a Dev's. As in "why do Autoit block comments support nesting?".

I'm just following the behavior of Autoit in this case. (Help: The #comments-start and #comments-end directives can be nested.)

Isn't anything between #cs to #ce must be commented out?

?, It is. But only between matching #cs and #ce pairs.

Edit: actually from #cs to next matching #ce OR end of document.

Edited December 25, 2009 by MvGulik

[GEOSoft]

Unless it's been changed again End of document isn't valid for #cs. I used to do that and then I started getting errors about it and I had to end all the #ce's back in at the end of the document. I haven't tried it again since and it's probably not a good practice anyway.

[MvGulik]

End of document isn't valid for #cs.

Yep.

It was more a observed syntax highlighting behavior note.

- - -

On the Regular recursive pattern part.

Although code wise speaking I worked my way around it. I'm still interested if its possible, or not, to do direct recursive pattern matching for words.

[Authenticity]

So lookahead does not fit the situation?

[MvGulik]

So lookahead does not fit the situation?

Oops. I mistook the code in your previous message as a quote or something related to your question, and not as a possible solution. sorry about that.

Your code does what I asked for. So it fits.

Thanks.