[SOLVED] PCRE Regular expression match but ignore string in between

[Kyan 12]

Hey everyone

I'm stuck with this regular expression, for matching a iso standard version

what I want to do is match the text in green: ISO 11784:1996/Amd.2:2010(E)
there is also others that look like this: ISO 11789:1999(E), and without the (E) part,

Só I did this one

 (?i)ISO.+?[\d-]+:(?:\d+/Amd.+?:)?[\d-]+(?:\(\w\))*

to match both cases but doesn't ignore the text in red

Can somebody point out what I'm doing wrong here?

 

Edited December 1, 2016 by Kyan
Solved TAG added

[jchd 985]

This should work in your example cases:

(?i)(iso\s+\d+:)(.+:)?(\d{4}(?:\(e\))?)

I've put capturing parenthesis around the three groups to show what they get, but they aren't necessary.

[Kyan 12]

Hi jchd,

It doesn't work, outputs:

Row|Col 0
[0]|ISO 11784:1996(E)
[1]|ISO 11784:
[2]|
[3]|1996(E)

on the line  [0] should have outputted: ISO 11784:2010(E) 

there's also standards with hyphens: ISO 11412-2-1:2001, that's why I used a class

[mikell 635]

Could you simply remove the unwanted part between 2 colons ? (obviously depending on the use)

$str = "ISO 11784:1996/Amd.2:2010(E)" & @crlf & _ 
    "ISO 11789:1999(E)"

Msgbox(0,"", StringRegExpReplace($str, ':.*:', ":") )

 

[Kyan 12]

Hey mikell

Can't be done with the regular expression? I could do that, I do that to remove multiple consecutive spaces after capturing it

(?i)iso\s+\d+:(?:\d{4}.+amd.+:)?\d{4}(?:\(e\))?

This one also doesn't work, it displays 1996

[mikell 635]

It depends on the wanted result, simpler is often better
What is precisely the entered string ? what is precisely the needed result ?

[Kyan 12]

Already said what they are but there it goes an example

Inputs:

ISO 11784:1996/Amd.2:2010(E)
ISO 11784:1996/Amd.1:2007
ISO 11784-1-2:1989(B)

Outputs:
ISO 11784:2010(E)
ISO 11784:2007
ISO 11784-1-2:1989(B)

[mikell 635]

Hmm my previous code works well for this, doesn't it ? 

[Kyan 12]

With stringregexpreplace works fine, but why can't (?:) work properly?

$text="ISO 11784:1996/Amd.2:2010(E)|ISO 11784:1996/Amd.1:2007|ISO 11784-1-2:1989(B)"
$aText = StringSplit($text,"|")
For $x In $aText
    $l = StringRegExp($x,'(?i)iso\s+\d+:(?:\d{4}.+?amd.+?\:)\d{4}(?:\(\w\))?',2)
    If Not @error Then ConsoleWrite($l[0]&@LF)
Next

outputs:

ISO 11784:1996/Amd.2:2010(E)
ISO 11784:1996/Amd.1:2007

EDIT:

This one also doesn't work, is a zero width look behind

$text="ISO 11784:1996/Amd.2:2010(E)|ISO 11784:1996/Amd.1:2007|ISO 11784-1-2:1989(B)"
$aText = StringSplit($text,"|")
For $x In $aText
    $l = StringRegExp($x,'(?i)iso\s+[-\d]+:(?<=\d{4}.+?amd.+?\:)?\d{4}(?:\(\w\))?',2)
    If Not @error Then ConsoleWrite($l[0]&@LF)
Next

@Melba23 I need your expertise pls

Edited November 27, 2016 by Kyan

[mikell 635]

You mean this ?

$text="ISO 11784:1996/Amd.2:2010(E)|ISO 11784:1996/Amd.1:2007|ISO 11784-1-2:1989(B)"
$aText = StringSplit($text,"|")
For $x In $aText
    $l = StringRegExp($x,'(?i)(iso[^:]+:)(?:\d{4}.+?amd.+?:)*(\d{4})(\(\w\))?',3)
    If Not @error Then ConsoleWrite($l[0]&$l[1]&@LF)
Next

Hard way. Removing the unwanted parts is much easier

$text="ISO 11784:1996/Amd.2:2010(E)|ISO 11784:1996/Amd.1:2007|ISO 11784-1-2:1989(B)"
$aText = StringSplit($text,"|")
For $x In $aText
    $l = StringRegExpReplace($x, '(?m):.*(?=:)|\(\w\)$', "")
    If Not @error Then ConsoleWrite($l & @LF)
Next

 

Edited November 27, 2016 by mikell
typo... sorry

[Kyan 12]

yes, but that changes how I was capturing strings previously, I was outputting one liners with these (using StringRegExp flag 2):

BS( EN ISO)?.+?[\d-]+:[\d-]+(\(\w\))?|DIN.+?[\d-]+:[\d-]+(\(\w\))?|EN\s+?NP\s+?ISO.+?[\d-]+?:[\d-]+(\(\w\))?|EN\s+?ISO.+?[\d-]+:[\d-]+(\(\w\))?|ISO.+?[\d-]+:[\d-]+(\(\w\))?

the thing is that I found those "Amd..." that I only wanted the year preceeded by "Amd.\d:"
 

Sorry if this is kinda complicated, but it's kinda hard to understand how  a non capturing group is captured.

Edited November 27, 2016 by Kyan

[mikell 635]

2 minutes ago, Kyan said:

it's kinda hard to understand how  a non capturing group is captured

Kinda weird 

[jchd 985]

Matching part X then non-capturing group Y then matching part Z is still matching all of XYZ. Unless you use *Replace and glue together parts X and Z by capturing them, you can't make Y vanish mysteriously. Hence my proposal : https://regex101.com/r/jezYus/1

Edited November 27, 2016 by jchd

[InunoTaishou 179]

My attempt

#include <Array.au3>

Global $aInput[] = ["ISO 11784:1996/Amd.2:2010(E)", _
        "ISO 11784:1996/Amd.1:2007", _
        "ISO 11784-1-2:1989(B)"]

For $i = 0 To UBound($aInput) - 1
    Local $aRegex = StringRegExp($aInput[$i], "(?i)(iso.*:).*:(.*)", 3)
    If (Not @Error) Then ConsoleWrite(_ArrayToString($aRegex, "", -1, -1, "") & @LF)
Next

I'm confused by

Quote

ISO 11784-1-2:1989(B)

In your example, the input and the output are the same for this value.

[Kyan 12]

First of all, thanks to all of you for helping me out.

@jchd wow, great website, so your proposal is to match them in groups then use StringRegExpReplace to place them correctly? Can you give me a advice how should I include other standards variants in the first capture group? There're BS EN ISO... EN ISO.., EN NP ISO...and sometimes my OCR fails and translates ISO to IS0 so I tried with this but doesn't look good: ((?:BS\h+)?(?:EN\h+)?(?:NP\h+)?IS[0o])

@InunoTaishou don't be confused, because that's a correct input-output

[jchd 985]

See the accordingly modified regexp and examples using the revised link: https://regex101.com/r/jezYus/2

[Kyan 12]

thank you @jchd! I made it to match other standards than ISO (DIN) and the letter between parenthesis isn't always a (E), that's a revision letter, so I placed a \w instead of "e" https://regex101.com/r/jezYus/3

[jchd 985]

Ha, yes I saw that mentionned days ago but forgot to include the variation in the regexp.

The point of importance is that you get the idea and can now fine-tune the expression to more cases, if need arises.

[Kyan 12]

@jchd, there's some way that I start matching from the end of the document?

[jchd 985]

In case your source has a regexp-able trailer after the "right" standard reference, yes you can probably, but I'm afraid you're going to encounter too many variations there. Maybe in his case you can display the scan output and highlight the possible captures then have a user (you ?) select the correct one.

If the wanted reference is always the last one in the source text, you can always try prefixing the pattern with .* which will greedily skip everything up to the last part of the pattern.