RegEx Question

[Michael Mattias]

The PB REGEXPR can get a bit unwieldy what with its "nonsupport" for the "repeat" option available with other flavors. Real Life example:

; Setting two Alpha, 3-5 numeric, 'P', 2-5 numeric; suffix all alpha up to first numeric
FI,621611429,[B][a-z][a-z][0-9][0-9][0-9][0-9]?[0-9]?P[0-9][0-9][0-9]?[0-9]?[0-9]?[/B],[B][a-z]+[/B]

This example actually comes from an "INI" file in which the user would enter his own masks (there are two masks here: first is the two A,3-5 N, P, 2-5 N; second is all alpha).
If a user can do it, so can you. (Yes, I created this particular example for him).

But as far as " it would be nice if a tool could convert it to PB", I have made my opinion very clear on trying to do verb-for-verb, function-for-function conversions from one source code language to another language many times (=stone loser) and that view holds here.

[Michael Mayerhoffer]

>I might ask JG soft if they'd consider adding a PB flavour,

Maybe not. IMO the PB REGEXPR and REGREPL statements are not as feature-rich as are the other flavors.

They play well with other string functions in PB , somewhat levels the playing field so I think, like mid$

Down to brass tacks

While pb don't have bla you may always improvise, a tool to create or convert would be nice. Here is a simple example.


For Example - some have numeric repeat for part of the mask expression , easy to replicate in PB.

Off the top of my head lets say {(z)'x5)'} or some notation like this would equal (zzzzz) 5 z's for exact match

pb use "(" + REPEAT$(5 "z") + ")" or (zzzzz)



a not so easy, because group can be changed during runtime in a script by the user.

define {GROUP} ="ABCDEFG"

{GROUP} - 'G'

group now equals ABCDEF

or {GROUP} + '1'

now ABCDEF1


lot more work in pb , several ways to approach this. Regex might be off the table , it would be nice if a tool could convert it to PB


I am sure you get the point

Mike

added: is this worth 30-50 dollars for a user ?

[Michael Mattias]

>I might ask JG soft if they'd consider adding a PB flavour,

Maybe not. IMO the PB REGEXPR and REGREPL statements are not as feature-rich as are the other flavors.

[Grant Mac]

OK, so it's unique.I believe the Ultra-Edit Regex was very limited, but the now can run Perl which I'm told is far far better. I might ask JG soft if they'd consider adding a PB flavour, though I expect it's probably too small a user-base for them to be interested, relative to the C++, etc user-base. I guess a large part would depend on just how different it was.

[Michael Mattias]

By the way, I have been wondering what 'flavour' (or flavor for you Yanks) of Regex is the PB Regex?

That would be the 'PowerBASIC' flavor. No, it's not Perl, C++, C# or Java as you mention; and for good measure it's also neither Unix nor Ultra-Edit.

[Grant Mac]

Thanks. That sounds interesting Michael. I haven't used the PB Regex as yet, though I do have a program where I've often thought it would be useful, if I ever get back to it!

[Michael Mayerhoffer]

I have already reported the missing Webpage. FWIW , I created my own locally program now for quick testing, nothing more than a wrapper.

[Grant Mac]

By the way, I have been wondering what 'flavour' (or flavor for you Yanks) of Regex is the PB Regex? For example, RegexMagic and RegexBuddy can handle Perl, C++, C#, Java, etc, 194 in total at present. I'm not sure if PB fits close enough to any of those 194, but if so, perhaps JGSoft (who wrote those two programs) might be willing to add PB to the mix. I found RegexMagic really useful for some stuff I was doing a while back, so it would be nice to have it suitable for PB. Also, that link I posted above is now no longer so perhaps the moderators / admin can fix it?

[Michael Mattias]

PB/Win (and maybe PB/CC) help used to have a link to an online REGEXPR tester somewhere at powerbasic.com.

(Under, duh, REGEXPR). May still be there.

[ADDED]
I couldn't resist. PB/Win 10 help does have a link to it. It's under "See also" near the bottom of the REGEXPR page. The tester page even has a little "control character help" right on the screen.

There is no such link under REGREPL.

[Grant Mac]

Obviously this is an old post, but I was looking for stuff on PB's Regex and found this post, so figured other people might also find PB's online Regex Tester useful. There is a clickable link in the Helpfile (at least in PBCC6 there is anyway), but I got the actual link so it can be accessed via your browser too. I'm not sure if the source code for that is around anywhere, perhaps somebody else knows?

http://www.powerbasic.com/bin/regexpr.exe?from=pbcc

[Jeff Blakeney]

This was posted by Kevin Voell 6/11/06:

http://www.powerbasic.com/support/pbforums/showthread.php?t=24927

(That's for PB/Win. )

I just grabbed that to try it out and I just wanted to point out that the source seems to have been pasted into the message twice so you need to remove the second copy before it will compile. PB/WIN v9.01 complained when it hit the second #COMPILE EXE statement.

[Michael Mattias]

Didn't someone post a nifty "regular expression" tester program here once? Where you can just type in your strings and click a button and you get the answer?

Um, yes, someone did.

This was posted by Kevin Voell 6/11/06:

http://www.powerbasic.com/support/pbforums/showthread.php?t=24927

(That's for PB/Win. )

[George Bleck]

OK, Never mind... big DUH on my part...I forgot to account for numbers which is why my original example failed, even after I escaped out the "$", "(", and ")" characters... took me awhile to catch that stupid error. Also the "\s" helps.

Below winds up being the "final" code. I changed the rules a bit and forced the parameter to be inbetween quotes and parenthesis as it looks cleaner in my production work.

 
 
'----------------------------------------------------------------------------(')
 
#INCLUDE "win32api.inc"
 
'----------------------------------------------------------------------------(')
 
FUNCTION GetCommands( strSearchBuffer AS STRING, strArray( ) AS STRING ) AS LONG
 LOCAL lngMatchCount AS LONG
 LOCAL lngMatchLength AS LONG
 LOCAL lngMatchPosition AS LONG
 LOCAL lngSearchPosition AS LONG
 LOCAL strRegEx AS STRING
 ERASE strArray( )
 ' Enumerate the following mask into an array:  $CommandText("Parameter")
 strRegEx = "\s\x24[a-z0-9]+\x28\q(.*)\q\x29"
 DO
  REGEXPR strRegEx IN strSearchBuffer AT lngSearchPosition TO lngMatchPosition, lngMatchLength
  IF lngMatchPosition THEN
   ' We have a match!
   ' Increment the match count, resize the array accordingly,
   ' then save the match to the new last slot in the array
   INCR lngMatchCount
   REDIM PRESERVE strArray( lngMatchCount - 1 )
   strArray( lngMatchCount - 1 ) = MID$( strSearchBuffer, lngMatchPosition, lngMatchLength )
   ' Change the search position and keep trying
   lngSearchPosition = lngMatchPosition + lngMatchLength
   ITERATE
  END IF
  EXIT LOOP
 LOOP
 FUNCTION = lngMatchCount
END FUNCTION
 
'----------------------------------------------------------------------------(')
 
FUNCTION PBMAIN( )
 DIM strArray( 0 ) AS STRING
 LOCAL lngArrayIndex AS LONG
 LOCAL lngMatchCount AS LONG
 LOCAL strTestString AS STRING
 strTestString = "C:\MyPath\MyEXE.exe " & _
   "/Switch1=Command1(""Parameter1"") " & _
   "/Switch2=$Command2("""") " & _
   "/Switch3=$Command3(""Parameter3"") " & _
   "/Switch4=$(""Parameter4"")"
 MSGBOX "Test string:" & $CRLF & $CRLF & strTestString
 lngMatchCount = GetCommands( strTestString, strArray( ))
 IF lngMatchCount THEN
  MSGBOX FORMAT$( lngMatchCount, "###  " ) & "Item(s) found"
  FOR lngArrayIndex = LBOUND( strArray ) TO UBOUND( strArray )
   MSGBOX FORMAT$( lngArrayIndex + 1, "###  " ) & strArray( lngArrayIndex )
  NEXT lngArrayIndex
 ELSE
  MSGBOX "No commands were present"
 END IF
END FUNCTION

[George Bleck]

Mike, thanks for your assistance. Unfortunately you were a bit TOO literal, I guess as hard as I tried to be descriptive I still failed.

The patterns I am looking to enumerate must:

(1) Start with a "$"
(2) Have one or more letters/numbers following the "$"
(3) Have a "(" following all the previous characters
(4) Have any combination of letters/numbers (or nothing) following the "("
(5) End with a ")"

Examples
$Command(Parameter) is valid
$Word(MoreWords) is valid
$(SomethingSpecial) is NOT valid
$BlankParameter() is valid
MissingSomething(SomeText) is NOT valid

I am not worried with the looping to search for them, that much I can handle. I am just having a hard time nailing the basic RegEx. It works perfect in RegExBuddy but I just can't get it to work in PB.

[Michael Mattias]

If you are looking for the dollar sign '$' and/or parens '()' you will have to escape them since they are both metacharacters in PB regular expressions

You can try something like

 StrRegEx =    "\S"   _           ' shortest (else you could get the whole line)
                & "/Switch"      _    ' literal
                &  "0-9"          _   ' exactly one numeric digit 
                &  "="            _   '  literal  
                & "\x24"         _   ' dollar sign in accepted format
                & "Command"  _   '  literal
                & "0-9"          _   '  exactly one numeric digit
                & "\x28"         _    ' left paren
                & "(.+)"            _    ' any characters, tagged (ranges may not be tags)
               & "\x29"          _   ' rightParem
               & "\w"                 ' word break 

  ReplaceMask  =  "\1"       ' first tagged group

 REGREPL strRegEx IN strTestString with ReplaceMask TO  iPos, paramvalue

.. and maybe do it in a loop as shown in this demo...

REGEXPR and REGREPL demo January 16, 2002

The above is probably not exactly what you want but I think by escaping the metatcharacters you will be off to a good start.

(that demo has multiple examples in it, too).

Didn't someone post a nifty "regular expression" tester program here once? Where you can just type in your strings and click a button and you get the answer?

MCM

[George Bleck]

Can someone help me with a simple RegEx. I am looking to return each $Command(Parameter).

According to RegExBuddy I could just use \$\w+\(\w*\) but that's not valid PB RegEx syntax so I expanded (and simplified) the \w token.

 
FUNCTION PBMAIN( )
 LOCAL lngLen AS LONG
 LOCAL lngPos AS LONG
 LOCAL strRegEx AS STRING
 LOCAL strTestString AS STRING
 strRegEx = "\$[a-z]+\([a-z]*\)"
 strTestString = "/Switch1=$Command1(Parameter1) /Switch2=$Command2(Parameter2)"
 REGEXPR strRegEx IN strTestString TO lngPos, lngLen
 MSGBOX _
 "Position = " & FORMAT$( lngPos ) & $CRLF & _
 "Length   = " & FORMAT$( lngLen ) & $CRLF & _
 "Text     = " & MID$( strTestString, lngPos, lngLen )
END FUNCTION