How to replace text at a special place in special rows?

[ErwinSchmidt17]

Hello, I want replace text (for example 5000x5000 with 4500x4500) only in the rows with the text projectID:1234

I got the answer how to do this at
https://community.notepad-plus-plus.org/topic/19742/how-to-replace-only-in-rows-with-a-special-text: 5000x5000(?=.*?projectID:1234)

But now I want to have a further condition set:

I want to replace the text only for a special platform.

My file is like this (pf1=platform1, pf2=platform2…):

{name:11.png,filename:c:\img\11\11.png,projectID:1234},pf1:{spname:11.png,spfilename:11.png},pf2:{spname:11.png,spfilename:11.png},pf3:{…}
{name:11.png,filename:c:\img\11\11.png,projectID:1234},pf1:{spname:11.png,spfilename:11.png},pf2:{spname:33.png,spfilename:33.png},pf3:{…}
{name:11.png,filename:c:\img\11\11.png,projectID:3456},pf1:{spname:11.png,spfilename:11.png},pf2:{spname:11.png,spfilename:11.png},pf3:{…}
{name:11.png,filename:c:\img\11\11.png,projectID:4567},pf1:{spname:11.png,spfilename:11.png},pf2:{spname:11.png,spfilename:11.png},pf3:{…}
…

Now for example I want to replace 11 with 22 only where the projectID is 1234 AND only for platform2 (pf2 in file)
After this I have this file:

{name:11.png,filename:c:\img\11\11.png,projectID:1234},pf1:{spname:11.png,spfilename:11.png},pf2:{spname:22.png,spfilename:22.png},pf3:{…}
{name:11.png,filename:c:\img\11\11.png,projectID:1234},pf1:{spname:11.png,spfilename:11.png},pf2:{spname:33.png,spfilename:33.png},pf3:{…}
{name:11.png,filename:c:\img\11\11.png,projectID:3456},pf1:{spname:11.png,spfilename:11.png},pf2:{spname:11.png,spfilename:11.png},pf3:{…}
{name:11.png,filename:c:\img\11\11.png,projectID:4567},pf1:{spname:11.png,spfilename:11.png},pf2:{spname:11.png,spfilename:11.png},pf3:{…}
…

The first part I can do with this regular expression:

find: 11(?=.*?projectID:1234)
replace with: 22
(regular expression checked)

But this condition alone would change all incidences of 11 with 22 in a row where the text projectID:1234 is in the row.
But I want to only have the 11 changed to 22 in the parenthesis of pf2:{spname:11.png,spfilename:11.png}

So how can this been done - only change rows with a special projectID and as second condition only in the parenthesis of pf2?

[Ekopalypse]

@ErwinSchmidt17

from the given example I assume the following will do what you want to achieve.

find what:(?<=projectID:1234}).*(?<=pf2).*?\K11(?=.*pf3)
replace with:22

And you must press replace all until npp reports that no more matches are found.

[Terry R]

@ErwinSchmidt17 said in How to replace text at a special place in special rows?:

So how can this been done

Building on @Ekopalypse solution I’d extend the regex slightly to change both occurances of the 11 at the same time if the conditions are met. Thus my Replace function would be
Find What:(?-s)^((.+)?projectID:1234(.+)?pf2:\{[^}]+?)11([^}]+?)11
Replace With:\122\422

From your example it would appear the 11 will ALWAYS occur twice in the “pf2” section (between { and }) if it does exist at all. So my regex prevents running into the next section “pf3” with the use of [^}]. This means take any characters as long as they are NOT the }. So you will only need to press the “Replace All” button once to have all the occurances replaced.

If unsure of my regex you can also use it to bookmark the lines containing the conditions. Then you could step through the lines to confirm. Alternatively you could use the “Replace” button to change 1 line at a time. Use the "Find button at the start, then the replace button from there on and you will see each line selected in turn.

Another method of achieving the same result, but taking more steps to do so are:

1. add a line number (with leading zeroes) at the start of each line (Edit, Column editor)
2. Bookmark the lines containing the first condition.
3. Cut these lines out and insert into another empty file.
4. Continue with editing these lines based on the second condition.
5. Copy these lines back to the original file and sort numerically, thus the line number at start puts them back into the correct sequence.
6. Remove the line numbers at the start of a line.
7. Finished. You see, by breaking down the problem into several smaller steps you can achieve the desired result.

Terry

[Ekopalypse]

I recently found out that there is a difference in handling when using \1 vs. $1.
In this example \122 the 22 to would be added to the captured text from \1 but $122 wouldn’t work.
I assume it treats this as the 122nd capturing group.
Any idea why this is so?

[Terry R]

@Ekopalypse said in How to replace text at a special place in special rows?:

$122 wouldn’t work

From the little I have read I would think the \ means take “as few numbers following as possible”, whereas the $ might be "more greedy, sort of like the * and + quantifiers within regex. I think it did see that we can also use ( and ) to force the system to recognise what we actually want.

This might be another case of becoming a bit more particular such as the (?-s) which I admit is still something I HAVE to think about, otherwise I forget. Especially as I know my own environment and rarely use it “at home”.

Perhaps @guy038 can supply the “answer”?

Terry

[PeterJones]

@Terry-R , @Ekopalypse , and all interested parties:

From my understanding (as reflected in the Searching: capture groups and backreferences and substitutions docs), the \ℕ notation only accepts single-digit. In the search string, you can use one of the more verbose backreferences (\gℕ, \g{ℕ}, \g<ℕ>, \g'ℕ', \kℕ, \k{ℕ}, \k<ℕ> or \k'ℕ').

The Boost 1.70 docs confirm: back references says that in the \ℕ form, ℕ must be in the range 1-9; the \g-variants (especially with \g{ℕ} allow for higher ℕ values (the \k-variants are supposed to be for named groups, but they also work for numbered groups at last experimentation). The Boost replacement: placeholder sequences only mentions $ℕ and ${ℕ}, and do not mention \ℕ-notation at all in the substitution/replacement syntax.

My rule of thumb is to exclusively use $ℕ or ${ℕ} in replacements, and I use the ${ℕ} more often, because it is 100% unambiguous for reading (even though it is harder to type on my keyboard).

[Alan Kilborn]

I’m fairly certain it works like this:

\122 --> group 1 and then a literal 22

$122 --> group 12 and then a literal 2

${122} --> group 122

Obviously, the group numbers need to exist for it to work.

The good habit to get into is to always use the curly brace syntax, so, for example from the current discussion: ${1}22

[Ekopalypse]

@Alan-Kilborn said in How to replace text at a special place in special rows?:

I’m fairly certain it works like this:

I don’t think so.
Having the text

0123456789ABCDEF

and doing
find what:(.)(.)(.)(.)(.)(.)(.)(.)(.)(.)(.)(.)(.)(.)(.)(.)
replace with:$122
doesn’t result in B2
but
replace with:$12 2
results in B 2

[Ekopalypse]

what we could do in addition is
replace with:($12)2

[Ekopalypse]

I can’t seem to get the \g notation to work either.
replace with:\g{12}2
results in g{12}2

[PeterJones]

@Ekopalypse said in How to replace text at a special place in special rows?:

what we could do in addition is
replace with:($12)2

Equivalently: replace with: ${12}2

I can’t seem to get the \g notation to work either.
replace with:\g{12}2

In Boost, \g-notation is only listed in the SEARCH section, not in the REPLACE section (fixed typo).

[PeterJones]

@Ekopalypse said in How to replace text at a special place in special rows?:

\g notation

To confirm: I can successfully search 0123456789ABCDEFB using the regex (.)(.)(.)(.)(.)(.)(.)(.)(.)(.)(.)(.)(.)(.)(.)(.)\g12, and it matches (because the 17th character matches the 12th backref)

[Alan Kilborn]

@Ekopalypse

Haha, well, that’s why I said “fairly” certain.

Actually, I cheated: Long ago I read about RegexBuddy here on the Community (at least I think it was here), and purchased a license. It has proved invaluable.

Here’s what it told me for this case:

I should have cited RB a few minutes ago when I posted, but I wanted to see if there was agreement/disagreement first.

Very rarely have I found any discrepancies between RB and N++, but this may be one of those cases.

It is interesting that RB doesn’t say “Insert the character string 22 literally” in the second and third lines of its output, but breaks it into 2 parts…hmmm…

[Ekopalypse]

@Alan-Kilborn

maybe that is implementation detail (!?)

[ErwinSchmidt17]

Thank you for all your answers.

[guy038]

Hello, @ErwinSchmidt17, @terry-r, @ekopalypse, @peterjones, @alan-kilborn and All,

Sorry, to be late as I’m on a family vacation right now, for the better part of August ;-))

---

Quickly, about solutions to @ErwinSchmidt17’s problem, I would say :

SEARCH (?-s)(^.+projectID:1234.+pf2|\G).*?\K11(?=.*pf3)

REPLACE 22

Thus, the test data, below, containing 4 names 11.png, in the pf2 section :

{name:11.png,filename:c:\img\11\11.png,projectID:1234},pf1:{spname:11.png,spfilename:11.png},pf2:{spname:11.png,spfilename:11.png,bla_blah:11.png,test:11.png},pf3:{spname:11.png,spfilename:11.png},pf4:{...}

would be changed as :

{name:11.png,filename:c:\img\11\11.png,projectID:1234},pf1:{spname:11.png,spfilename:11.png},pf2:{spname:22.png,spfilename:22.png,bla_blah:22.png,test:22.png},pf3:{spname:11.png,spfilename:11.png},pf4:{...}

---

Now, about the different syntaxes, related to groups, back-references and subroutine calls, I did some tests and here are my conclusions, not definitive, of course :

In search regexes, the possible syntaxes, with Boost regex library, are :

• Unnamed group is defined with surrounding parentheses : (.....)

• Named group is defined with the one of the syntaxes :

  • (?<Name>.....)

  • (?'Name'.....)

• Absolute back-reference, to an unnamed group N, is defined with one of the syntaxes :

  • \N    ( with 1 <= N <= 9 )

  • \gN    \g{N}    \g<N>    \g'N'    ( with 1 <= N <= Max )

  • \kN    \k{N}    \k<N>    \k'N'    ( with 1 <= N <= Max )

• Relative back-reference, to an unnamed group X, is defined with one of the syntaxes :

  • \g-X    \g{-X}    \g<-X>    \g'-X'    ( with 1 < X <= Max )

  • \k-X    \k{-X}    \k<-X>    \k'-X'    ( with 1 < X <= Max )

• Absolute subroutine call, to an unnamed group N, is defined with the syntax :

  • (?N)    ( with 0 <= N < Max )

• Relative subroutine call, to an unnamed group of relative number X, is defined with one of the syntaxes :

  • (?-X)    ( with 1 < X <= Max )

  • (?+X)    ( with 1 < X <= Max )

• Absolute back-reference, to a named group Name, is defined with one of the syntaxes :

  • \g{Name}    \g<Name>    \g'Name'

  • \k{Name}    \k<Name>    \k'Name'

• Absolute subroutine call, to a named group Name, is defined with one of the syntaxes :

  • (?&Nom)

  • (?P>Nom)

---

Remarks :

• For all the relative syntaxes above, the Max value is the greatest group of the overall regex

• For all the absolute syntaxes, I suppose that the Max value is 2,147,483,647, as it’s the same value in replacement, too !

• The names of named groups are word characters, non beginning with a digit

• The (?0) is a subroutine call to the overall regex and is, implicitly, a recursive subroutine call !

Summary example :

To find a four-letters word palindrome, you can use, either, one of these 23 syntaxes :

\b(\w)(\w)\2\1\b

\b(\w)(\w)\g2\g1\b
\b(\w)(\w)\g{2}\g{1}\b
\b(\w)(\w)\g<2>\g<1>\b
\b(\w)(\w)\g'2'\g'1'\b

\b(\w)(\w)\k2\k1\b
\b(\w)(\w)\k{2}\k{1}\b
\b(\w)(\w)\k<2>\k<1>\b
\b(\w)(\w)\k'2'\k'1'\b

\b(\w)(\w)\g-1\g-2\b
\b(\w)(\w)\g{-1}\g{-2}\b
\b(\w)(\w)\g<-1>\g<-2>\b
\b(\w)(\w)\g'-1'\g'-2'\b

\b(\w)(\w)\k-1\k-2\b
\b(\w)(\w)\k{-1}\k{-2}\b
\b(\w)(\w)\k<-1>\k<-2>\b
\b(\w)(\w)\k'-1'\k'-2'\b

\b(?<First>\w)(?'Second'\w)\g{Second}\g{First}\b
\b(?<First>\w)(?'Second'\w)\g<Second>\g<First>\b
\b(?<First>\w)(?'Second'\w)\g'Second'\g'First'\b

\b(?'First'\w)(?<Second>\w)\k{Second}\k{First}\b
\b(?'First'\w)(?<Second>\w)\k<Second>\k<First>\b
\b(?'First'\w)(?<Second>\w)\k'Second'\k'First'\b

Test them against this text :

adda – a type of lizard
Adda – a river in Italy; a river in Wales
Anna – a girl’s name
Beeb – an informal name for the BBC
boob – a blunder; a breast
deed – various common meanings
goog – an egg (Australian slang)
immi – a Swiss unit of volume
keek – to peep
kook – a crazy person
naan – a type of Indian bread
noon – midday
Otto - a proper name
peep – various common meanings
poop – a raised deck at the stern of a ship; various other meanings
toot – the sound made by a horn or whistle

Now, as a subroutine call is, basically, a reference to the regex itself, included in a group and NOT the last value of this group like in back-references, the 5 following syntaxes are strictly equivalent to the simple regex \b\w{4}\b and looks for a four-letters word :

\b(\w)(\w)(?2)(?1)\b
\b(\w)(\w)(?-1)(?-2)\b
\b(?+2)(?+1)(\w)(\w)\b

\b(?<First>\w)(?'Second'\w)(?&Second)(?&First)\b
\b(?'First'\w)(?<Second>\w)(?P>Second)(?P>First)\b

Test them, again, on the same sample text, above !

Important :

• All the syntaxes, above, are valid in search part ONLY !

• Because of the multiple equivalent syntaxes, for groups, back-references and subroutine calls, it is useful to define, for search regexes, a single, minimal syntax, covering the majority of cases :

Hence, the table, below, with my preferences :

    •============================•=============================•===================•====================•
    |           GROUP            |          REFERENCE          |  ABSOLUTE number  |  RELATIVE number   |
    •============================•=============================•===================•====================•
    |                            |       BACK-REFERENCE        |  \N   or   \g{N}  |       \g{-X}       |
    |  (.....)          UNNAMED  |                             |                   |                    |
    |                            |       SUBROUTINE CALL       |       (?N)        |  (?-X)  or  (?-X)  |
    •----------------------------•-----------------------------•-------------------•--------------------•
    |                            |       BACK-REFERENCE        |     \g<Name>      |        N/A         |
    |  (?<Name>.....)     NAMED  |                             |                   |                    |
    |                            |       SUBROUTINE CALL       |     (?&Name)      |        N/A         |
    •============================•=============================•===================•====================•

---

In replacement regexes, , with Boost regex library, you can use the following syntaxes :

• Absolute reference, to an unnamed group N, is defined with either :

  • \N    ( with 1 <= N <= 9 )

  • $N    ( with 0 <= N <= 2,147,483,647 )

  • ${N}    ( with 0 <= N <= 2,147,483,647 )

• Absolute reference, to an named group Name, is defined with the syntax :

  • $+{Name}

Remarks :

• The $0 or $& syntaxes refer to the overall regex, itself

• If number N is superior to the number of back-references, in the search regex, these syntaxes return an empty string

• If a named reference $+{name} does not exist in search regex, it also returns an empty string

• If, in the replacement regex, a digit follows a $N syntax, it’s preferable to use the ${N} form !

• The $00...00N and ${00...00N} syntaxes are equivalent to, respectively, the $N and ${N} syntaxes

• So, the single minimal syntaxes, in replacement, seems to be :

    •============================•==================•===================•
    |           GROUP            |     REFERENCE    |  ABSOLUTE number  |
    •============================•==================•===================•
    |  (.....)          UNNAMED  |  BACK-REFERENCE  |       ${N}        |
    •----------------------------•------------------•-------------------•
    |  (?<Name>.....)     NAMED  |  BACK-REFERENCE  |     $+{Name}      |
    •============================•==================•===================•

Best Regards,

guy038

[guy038]

Hi, All,

Out of curiosity, do you know how I could determine that the maximum number of group is 2,147,483,647 ?

Well, I began the test using this simple regex S/R :

SEARCH (?-s).

REPLACE --${300}--

When replacing a single character, it returns the string ----. So, the S/R seemed valid and, as the group 300 did not exist, it just wrote the empty string as replacement of this group.

Then, I, successively, changed the replacement zone with :

• --${3000}--    =>    ----

• --${30000}--    =>    ----

• --${3000000000}--    =>    --$3000000000}--

As I suspected that the limit should have a relation to powers of 2, I searched for the largest power of 2, below 3,000,000, which is 2^31 = 2,147,483,647 !

Indeed :

SEARCH (?-s).

REPLACE --${2147483647}--

=> The ---- output

and :

SEARCH (?-s).

REPLACE --${2147483648}--

=> The --${2147483648}-- output

Of course, I do understand that this limit is quite theoretical ! Just imagine a regex containing 2,147,483,647 different groups… Brrrrr

Best Regards,

guy038

[Ekopalypse]

@guy038 - sounds like a 32bit integer limitation.

[Alan Kilborn]

@Ekopalypse

sounds like a 32bit integer limitation.

Or an implementation detail. ;-)
BTW, trying this in RegexBuddy, it reports “group 2147483647” but if you go one higher it reports “group -2147483648”.

[Ekopalypse]

@Alan-Kilborn

:-D