Replacing text from x to y

Neil Schipper

@Mleczny-Sernik Hi. There are a few things about your description that are open to interpretation.

I have a regex solution that is very rigid, and only matches the type of whitespace, and the amount of indentation, shown in your sample.

Since the regex is deleting several lines, I prefer, as a starting point, to delete too little rather than too much.

I will assume the block you want deleted:

always starts exactly with an empty line, then (on next line) a single TAB, then the text “history={”
always ends exactly with: a line that contains one TAB, then text ‘}’, and nothing else

If the files you are processing are human-typed there’s a good chance that variations in whitespace, or commenting, will cause this solution to be incomplete. If the files are machine made, chances are better the solution will meet your need.

Find what: ^\r\n^\thistory={.*?^\t}\r\n
Ensure the “Replace with:” entry is completely empty
Do check the “. matches newline”

Do not rely on it without a lot of testing.

You may wish to enable (Np++ menu) “View - Show Symbol - Show all chars” when examining files before and after applying the regex.

Neil Schipper

@guy038 Hi. Your solution is far, far, far more sophisticated than anything I could have come up with.

However, re:

IF, of course, it contains a well-balanced number of { and } delimiters

what if a goofball, perhaps someone named something like Neil, was the author of one of these files, and in the course of development, he left one of the files to be processed looking something like this:

	history={
		owner = FRA
		add_core_of = CHA
		victory_points = {
			2081 1
		}
		//structures = {  // obsolete name! clean out after testing
		buildings = {
			infrastructure = 2

		}

	}

This would be a problem, I believe.

Terry R

@Neil-Schipper said in Replacing text from x to y:

what if a goofball, perhaps someone named something like Neil, was the author of one of these files, and in the course of development, he left one of the files to be processed looking something like this:

In every solution, whether it be a regex; simple to complex like @guy038 one here; through pythonscript code, through UDL’s, they ALL rely on data integrity.

So if someone comes to the forum seeking help and shows a “sample” of their data, the solution, whatever it may be will be based on that example. We may give a caveat, such as @guy038 did here, the need for balanced delimiters. In the end it is always the OP’s responsibility to provide enough “evidence” that we can trust our solution to act appropriately on the data.

If you were to read back through a lot of the posts in this forum, you will find however that the OP’s generally have a naive view of their data. Often the forum member(s) who are striving to help are the ones to ask the “right” questions about the data and thus jolt the OP enough that they gain a new respect for their data. Only through that process can the solution providers believe enough in their solution to provide it, albeit sometimes with a caveat.

Often solution providers will also suggest running the solution over a copy of the data and vetting the result before fully integrating it into their workflow.

I think your question, whilst having some merit is over thinking the process. OP’s ask for help. Solution providers may ask for additional information and/or examples. A solution is then provided, sometimes with information on how it works and what is required of the data to get valid results and it’s left to the OP to test. Hopefully the OP comes back with a “thank you” (amazing how many times we NEVER get that) so we know it solved their problem, or a gotcha so the process repeats with the additional information loaded in.

So don’t sweat it and get too bogged down in what-ifs when helping someone. You learn to rely on judgement and sometimes in the end the solution doesn’t work through no fault of the person helping.

Terry

guy038

Hi, @mleczny-sernik, @neil-schipper, @terry-r and All,

Neil, to solve the case you mention, we have 3 possibilities :

The first possibility is quite obvious :
- Open the Find dialog ( Ctrl + F )
- Type in a { or a } char in the Find what zone
- Tick the Wrap around option
- Click on the Count button
=> The number of { chars must be identical to the number of } chars ! If not, the program’s logic is broken and I wish you good luck to identify the missing or extra brace !
The second possibility is to decide that an escaped { or } char will be considered as a normal character, different from [{}].
- => Any allowed character, inside a {•••••} section, is represented by the regex (?:\\[{}]|[^{}])
- Then, the search regex to use becomes (?-i)^\h*history=(\{(?:(?:\\[{}]|[^{}])++|(?1))*\})\R+
- Of course, your comment line must then be re-written as //structures = \{ // obsolete name! clean out after testing
The third possibility is to use the following regex which identifies the longest contiguous zone with an identical number of opening brace and ending brace character(s)
- This magic regex is (?:[^{}]*(\{(?:[^{}]++|(?1))*\}))+[^{}]*|[^{}]+
- Do not tick the Wrap around and perform some tests, moving the caret to different locations, in your program, which should help you, to some extent, to spot the guilty brace char !
- Note that a simple text, without any brace char, is also matched by this regex. Logical, this text contains the same number of opening and ending brace chars : 0 ;-))

BR

guy038

P.S. :

You may test the regex (?:[^{}]*(\{(?:[^{}]++|(?1))*\}))+[^{}]*|[^{}]+ against this text, pasted in a new tab :

{{{{ab{{{cd{{{}}}}}ef}}}}}}
1234  567  89198765  43210x
             0

{{ab{{{{cd{{{ef{{}}}}}gh}}}}ijkl}}}}
12  3456  789  1119876  5432    10xx
               010

{{{{{{ab{cd{ef{{{}}}}}gh}ijkl}}}mn}}}}}
123456  7  8  91119876  5    432  10xxx
               010

{{01ab{cd{ef23gh{ij45kl}mn}op{{qr67st}uvwx}34}yz}}128956}abc
12    3  4      5      4  3  45      4     3  2  10      x

For example , move your cursor at the beginning of each line and click on the Find Next button ! This damn regex is never wrong !!

Neil Schipper

@guy038

Hi Guy,

I played with your bullet 3 regex against your sample text. It works, and it’s very cool.

If you wished (and you do not need to prove yourself as someone who likes challenges – I for sure would not try it) you could try to make it so that on successive Find Next clicks, rather than resume from text after the entire prior match, it could crawl from where prior match started to the next open brace and go from there, allowing user to see the enclosed region shrink bit by bit… Leading to the next challenge: going left bit by bit, and watching the region grow… maybe that’s too many levels of super-cool, I dunno… and then I remember that I/we already get matching brace highlighting (maybe only for some known languages?) (I forget if it’s a native Np++ feature or provided by a plugin.)

I tried the regex in your 2nd bullet and although it still works against the original sample code, it’s not working for me with that extra line I threw in, with the open brace but backslash-escaped. Don’t know why, and I’m not inclined to try to debug it; I’m really a noob in regex flow control.

Your conclusion in Bullet 1 is not strictly correct (unless file were stripped of all comments) but it’s not worth sweating about.

Neil Schipper

@Neil-Schipper said in Replacing text from x to y:

it’s not working for me

Sorry, I should have said why: it’s matching one more level of right brace.

guy038

Hi, @mleczny-sernik, @neil-schipper, @terry-r and All,

Ah…Ok ! So, I improve this magic regex to this version :

(?:[^{}]*(\{(?:[^{}]++|(?1))*\}))+[^{}]*|[^{}]+|}+ I just added the alternative |}+ at the end of the regex

Again, test it against the text below :

Move your caret/cursor right before each ¤ char, first
Note that I indicated, with the ^ character, the char changed in the 5 following lines ¤{{{{ab{{{cd....., in comparison with the first one
Right above each line ¤{{{{ab{{{cd..... :
- Any range -.....- represents a match of the main part (?:[^{}]*(\{(?:[^{}]++|(?1))*\}))+[^{}]*|[^{}]+
- Any range b.....b represents a match of the final part {+

---------------------------b----------------------------------bb------------------------------------bbb--------------------------------------------------------b---
¤{{{{ab{{{cd{{{}}}}}ef}}}}}}{{ab{{{{cd{{{ef{{}}}}}gh}}}}ijkl}}}}{{{{{{ab{cd{ef{{{}}}}}gh}ijkl}}}mn}}}}}{{01ab{cd{ef23gh{ij45kl}mn}op{{qr67st}uvwx}34}yz}}128956}abc


---------------------------------------------------------------b------------------------------------bbb--------------------------------------------------------b---
¤{{{{ab{{{cd{{{}}}}}ef}}}}}{{{ab{{{{cd{{{ef{{}}}}}gh}}}}ijkl}}}}{{{{{{ab{cd{ef{{{}}}}}gh}ijkl}}}mn}}}}}{{01ab{cd{ef23gh{ij45kl}mn}op{{qr67st}uvwx}34}yz}}128956}abc
                           ^

---------------------------b------------------------------------------------------------------------bbb--------------------------------------------------------b---
¤{{{{ab{{{cd{{{}}}}}ef}}}}}}{{ab{{{{cd{{{ef{{}{}}}gh}}}}ijkl}}}}{{{{{{ab{cd{ef{{{}}}}}gh}ijkl}}}mn}}}}}{{01ab{cd{ef23gh{ij45kl}mn}op{{qr67st}uvwx}34}yz}}128956}abc
                                              ^

---------------------------b----------------------------------bb----------------------------------bbbbb--------------------------------------------------------b---
¤{{{{ab{{{cd{{{}}}}}ef}}}}}}{{ab{{{{cd{{{ef{{}}}}}gh}}}}ijkl}}}}{{{}{{ab{cd{ef{{{}}}}}gh}ijkl}}}mn}}}}}{{01ab{cd{ef23gh{ij45kl}mn}op{{qr67st}uvwx}34}yz}}128956}abc
                                                                   ^

---------------------------b----------------------------------bb------------------------------------bbb------------------------------------------------bb------b---
¤{{{{ab{{{cd{{{}}}}}ef}}}}}}{{ab{{{{cd{{{ef{{}}}}}gh}}}}ijkl}}}}{{{{{{ab{cd{ef{{{}}}}}gh}ijkl}}}mn}}}}}{{01ab{cd}ef23gh{ij45kl}mn}op{{qr67st}uvwx}34}yz}}128956}abc
                                                                                                                ^

---------------------------------------------------------------------------------------------------bbbb--------------------------------------------------------b---
¤{{{{ab{{{cd{{{}}}}}ef}}}}}{{{ab{{{{cd{{{ef{{}{}}}gh}}}}ijkl}}}}{{{}{{ab{cd{ef{{{}}}}}gh}ijkl}}}mn}}}}}{{01ab{cd}ef23gh{ij45kl}mn}op{{qr67st{uvwx}34}yz}}128956}abc
                           ^                  ^                    ^                                            ^                           ^

In this last example, the regex automatically advances to the next well-balanced range of characters ! I note, with an lower-case letter s ( for skipped ), the character(s) skipped :

---------------------------------------------------------------------------------------------------bb---bbs--------------------------ss--ssss--------------s------
¤{{{{ab{{{cd{{{}}}}}ef}}}}}{{{ab{{{{cd{{{ef{{}{}}}gh}}}}ijkl}}}}{{{}{{ab{cd{ef{{{}}}}}gh}ijkl}}}mn}}}abc}}{{01ab{cd}ef23gh{ijkl}mn}op{{12{{{{34{}{qr}st{uv}{{34}yz

Now, regarding the second bullet of my previous post (?-i)^\h*history=(\{(?:(?:\\[{}]|[^{}])++|(?1))*\})\R+, if I escape the opening brace, in you comment line, as a convention which changes the brace as an ordinary character, giving this history section :

history={
		owner = FRA
		add_core_of = CHA
		victory_points = {
			2081 1
		}
		//structures = \{  // obsolete name! clean out after testing
		buildings = {
			infrastructure = 2

		}

	}

The regex does match all the section ! So could you show me an example where the regex fails to match this history block ?

Finally, my answer, in bullet 1 was, indeed, rather vague and I suppose that most IDE has tools to identify and correct the unmatched blocks of programming languages ! But running first my simple work-around means necessarily an error when numbers are different, isn’t it ?

Best Regards,

guy038

Mleczny Sernik

Thanks everyone for help :) It worked

Neil Schipper

@guy038 said in Replacing text from x to y:

The regex does match all the section ! So could you show me an example where the regex fails to match this history block ?

I’m seeing it match past end of state’s } (includes all trailing newlines) rather than past history’s as desired:

v7.9.5 64-bit

guy038

@neil-schipper,

Ah, yes ! I needed to indicate two consecutive \ characters in the regex ( in order to search for a literal \ char ! )

So the correct regex is rather :

(?-i)^\h*history=(\{(?:(?:\\\[{}]|[^{}])++|(?1))*\})\R+

BR

guy038

Neil Schipper

@guy038 said in Replacing text from x to y:

So the correct regex

Confirmed. In hindsight I should’ve spotted it myself – every instance of backslash-leftbracket on this site should be treat with suspicion and caution.