How to redefine "word boundary" in regex, or define a new character type

[leuce]

Hello everyone

I'm processing text files using AutoIt scripts, and in one such process I want to replace all standalone numbers with "00".  In other words, where a number is a "word", I want the number replaced with "00".  

I'm using the regex \b\d+\b.  However, this matches things that I do not consider "word boundaries".  For example, a date 12-Jan-15 is not "three words", in my view of what a "word" is, but Autoit think so.  Similarly, although "x; 123" or "y, 234" are two "words" each in my view, "x;123" or "y,234" are only one word each, but AutoIt thinks "x;123" and "y,234" are two "words" each.

The solution to my problem, I think, is if I could redefine the meaning of "\b", or define some other class that uses a different letter in regex.  Is there a way to do that?  It would simplify my regular expressions later in the script if I could define that earlier in the script.

Thanks

Samuel

[UEZ]

1 hour ago, leuce said:

I'm processing text files using AutoIt scripts, and in one such process I want to replace all standalone numbers with "00".  In other words, where a number is a "word", I want the number replaced with "00".  

Your statement is contradictory.

 

Should be the result this way?

"This is a test created on  12-Jan-15 with 1234 chars." -> "This is a test created on  12-Jan-15 with 00 chars."

 

If not please give an example.

[jchd]

leuce,

It's up to you to define precisely what you consider a "word boundary". For PCRE (AutoIt regex engine), \b has a very precise semantics and I understand it doesn't fit your need. Ask yourself what your "word boundary" really means and try it with that test, then any other text you can come up with:

"Pi is a transcendental value close to 3.1415926 that is as simple as 1+2=3 or rather 1 + 2 = 3."

Is a "stand-alone number" forcibly prefixed and suffixed by a space? Maybe another set of characters, or maybe depending of other rules like the final 3 above followed by a final point?

Formalize your need and thetranslation into regex will certainly follow.

[Skysnake]

You problem is Regex and not AutoIt.  I agree with the sages above, your example is not clear.  Perhaps post a sample of the source text with the replacements as you would EXPECT the outcome to be?

Also, practice your Regex here:

https://regex101.com/

[Malkey]

Here is my solution of my understanding of the problem.

Local $sTest = "334 This is a 1 test created on 12-Jan-15 with 1234 characters (3.456) and x;123.321 and y: 456 and z,789."

ConsoleWrite(StringRegExpReplace($sTest, "(?<=\s|^|\(|;|,|-)\d+(\.\d+)?(?=\s|$|\.|,|\)|-)", "00") & @LF)
#cs
This reg. exp. pattern will match "\d+" an integer or, "\d+(\.\d+)?" a decimal number, only if that number is preceeded by:-
    "\s" a space or a linefeed, or, "^" the number is at the beginning of the string, or, "(" an open bracket, ":" a semi-colon,"'" a coma, or, "-" a dash.
And, the matching number must also be followed by:-
    "\s" a space or a linefeed, or, "$" the number is at the end of the string, or, "\." a dot, "," a coma, ")" a closing bracket, or, "-" a dash.

Returns:-
00 This is a 00 test created on 00-Jan-00 with 00 characters (00) and x;00 and y: 00 and z,00.
#ce

 

[mikell]

On mardi 26 janvier 2016 at 10:11 AM, leuce said:

For example, a date 12-Jan-15 is not "three words", in my view of what a "word" is, but Autoit think so.

Malkey, your code doesn't fit the "explanations" in post #1
Probably it should be *something* like this (but the requirements are really imprecise)

$s = "Pi is a transcendental value close to 3.1415926 that is on 12-Jan-15 as simple as 1+2=3 or rather 10 + 20 = 30."

ConsoleWrite(StringRegExpReplace($s, '(?<=^|\s)\d*\.?\d+(?=[.,;\s]|$)', "00") )