Be Excellent To Each Other

Okay, so I've got a string like this:
The things between the numbers are 'and' or 'or' operators. I can find them using this regex:

How do I match and replace only, say, the 3rd match (in this case, '||')?

Bonus points if you can do it using javascript's string.replace function, but if not I should be able to figure it out from what you tell me.

Not sure you can do this and keep it scalable, to be honest. You can hardcode it, as follows:

.*(?:\|\||&&).*(?:\|\||&&).*(\|\||&&)

The ?: at the start of the parentheses tells the PCRE engine to not store those subpatterns so it'd only match the third one in this case.

At least, I think that should work.

You're probably better off writing some JS/PHP to split the string up into bits using the operators as delimiters, and then process them like that.

Really? Nads. Ta.

Something like:



works in perl. God knows if there are equivalents in PHP.

I'm no regex expert so wait and see if someone else weighs in, but I remember wanting to do similar things in the past and finding that they're just not capable of doing repetitions in the way you want them to.


Which bit is it in that that tells it to only process the third one? Is it the -- ?

Gazchap and I are working along similar lines but I'm apparently more bored at work than he

Shame you didn't want the stuff between your delimiters as that's easy to explode a string to an array using a regexp.

Making this function grab the nth element is an exercise left to the reader.

What I wanted to do was change the nth && or || to it's opposite. I've done it in Javascript - it works, but it's ugly.

Ooops, you have to set $instance =3 somewhere above.

Executing code in regexps is simultaneously beautiful and hideous at the same time.

Just like the rest of Perl I guess

Ha, sorry, I did

Thanks for all your help, chums!

There is nothing like a programming language that looks as if you've accidentally opened an image file in your text editor, but still actually does stuff.

Real nerds code in Brainfuck.

Hello World in Brainfuck:

The lolcats language looks like you accidentally opened a 3 year old's head instead

Pfft. Regex's are for losers. What you want is nested substrings.

I'm not sure what I wanted is possible with substrings, because of the && or || split.

Take that, you substring-using suckers!

*text-processing religious war goes here*

Had to get help with the regex (I was *nearly* there), but this does the job:



Since it's pretty much impossible to know exactly what it is that the regex has matched without iterating through all of the matches, the best I can do is replace the pipe/ampersand set with a placeholder (two full stops), then look at the original string at that position to work out what the function replaced in order to swap it. Splitting the string into an array is probably much faster.

I'm pretty sure a problem like this is non-regular, as it requires a notion of memory over previous matches* and so can't be matched by a standard regular language.

ie: Do it GazChaps way or don't do it at all using a Regex. You're using a screwdriver for a hammer problem. You can smack the nail in with the blunt end but it'll hurt like hell to do so.

* memory as in things wot happened, rather than RAM.


edit: Easiest way is just to loop over the pattern you provided, and only do something on the third match. That's like what, to extra lines? And if you want to make it 20th instead of 3rd you just change one number.

New to regex, I want to replace all characters in a string that are not numeric or are not a line feed character.

Something like [^0-9]\r

But not quite, as that only matches non numeric and a line feed (I think).

How do I use 'or' in regex?
Help?

([^0-9])|(\r) I think. You may not need the brackets in fact.

"|" is or IIRC

That means "all the characters that are anything other than a 0-9, OR are \r". That's not what you want to protect \r too.

[^0-9\r] is the pattern you want, I think. You can kuldge anything you want into square brackets like that. [a-zA-Z0-9] is a common pattern, for example.

Cheers guys, I'll try both of those on Monday.

Stupid Regex.

I'm trying to extract the RTW from the below phrase by selecting everything except that bit and replacing it with nothing:
123456-Mr Smith-RTW(27062013 150729).pdf

(^.*?)-(.*?)- seems to grab "123456-Mr Smith-"
\((.*)\).pdf grabs the "(27062013 150729).pdf" bit, but how can I combine these to grab both at the same time, when the RTW could be any combination of characters.

The first bit could be any characters, but the constants in this changing filename will be the two hyphens, and the datetime in parentheses plus .pdf

Or is it easier to just run the whole thing through a replacement function twice?

String split on "-" and take the right most one, then split on the ( and take the leftmost one. What're those quotes about using regex and having two problems, or everything being a nail?

It's hard to say without some more examples of file names - is it always -AAA that you need? Or can it have numbers in, too? Can the length of that change? What language are you doing it in?

The "loosest" possible way to get those three characters is probably this:



Fiddle with it

Use it once every 10 full tanks and it'll help no end.

No, the 'bit I need' could be any length and any combination of characters.

The only constants are that it will be data hyphen data hyphen NEEDED BIT Brackets Data Brackets.pdf

Pod, I'm trying to run this through an Eval() function in ASPX so I can only use one line, otherwise I'd pre-process the lot using aspx.

I'll have a go with the fiddle, thanks both.

Can the data have brackets or hyphens in?

If not, you need

If so, you're fucked

Fortunately not.

Ooh, so close...

?

Oh, okay:

That's fine for double-barrelled surnames, too.

Or what Squirt said.

In either of those situations, just pull out group 1.

(^.*?)-(.*?)- [but not this bit] \((.*)\).pdf

I kind of need to combine the first and last, where the [but not this bit] can be anything (but not a hyphen or parenth)

Hang on - do you want to find out what the bit before the bracket is, or remove it and glue what's left back together (ie. replace it with nothing)?

In other words, with should it return or ?

I want to turn "123456-Mr Smith-RTW(28062013 141130).pdf" into "RTW"

This is my existing code with the above as PDFFilename:


...that replaces the data in the brackets at the end, including the brackets with nothing, and leaves "123456-Mr Smith-RTW"

I am now trying to get rid of the data hyphen data hyphen before that last bit of data to leave just "RTW" (or whatever string of characters rtw happens to be).

I am aware the inclusion of a Regexreplace routine may now complicate things, sorry for being unclear before.

Take either mine or Squirt's regex, stick on the end and replace it with
rather than

That might not be how ASP does it, but it'll be something like that.

Lookaheads and behinds?

This should do the trick:

Very close if I use Regex.Match instead of Replace, but only matches the anydata bit where it is three characters.

Grim... your seems to be spot on, thank you!

I'm not exactly sure why, mind, as when I paste the regex bit into my regex tester it matches the whole string, but OK!

Thanks to everyone else who melded their mind around this as well

Ahh, can the RTW it be any length?

Then

will do it.

Use mine, not Grim...'s. Mine has question marks in.

The brackets around the .+ mark it out as being a 'group'. So we matched the whole thing, and replaced it with that group (using $1). If there were two groups the next one would be $2, and so on.