Replacing a digit with exception

  • @guy038 merci.
    I’m going to read this carefully.
    Actually I have to change your following code (?-s)($|\G)[[:xdigit:]]*?\K2 because it does not work if there are already digits in the 8 characteres strings : the search is going to the next line each time it encounters a letter, even if it didn’t red the entire string.
    I have to scratch my head a little bit to try to find a solution, I’ll tell you if I find a solution :)

  • @jean-francois-trehard and All,

    I don’t understand ! In the picture showing the text, below, I added some pure hexadecimal characters, either, in lower or upper case, in many locations, even after the ; Tile # string. Seemingly, all 2 digits, before comments only, are correctly marked !?

    dc.l $02A20620 ; Tile #42
    dc.l $0BaC2021
    dc.l $55111003 ; Tile #9A56
    dc.l $55001111
    dc.l $54000116 ; Tile #1b628
    dc.l $dF022222


    Jean-françois, could you provide some text which breaks the logic down ?



  • @guy038 oh I wrote something wrong :
    “because it does not work if there are already digits in the 8 characteres strings”
    I meant this : “because it does not work if there are already letters in the 8 characters strings”

  • @jean-francois-trehard and All,

    Ah OK, I understood the problem. Actually, you meant that, once an S/R has been processed to replace, for instance, a 2 digit with the Q letter, a second search, for instance, of the 6 digit will not find all occurrences of the 6 digit, because the letter Q is not an hexadecimal number !

    Indeed, the regex must be changed as below :

    SEARCH (\$|\G)\w*?\K2


    BTW, note that we can, also, express the search regex, using the free-spacing mode (?x), for a better readability :

    (?x)   ( \$ | \G )   \w   *?   \K   2

    Now, concerning the explanations, in my previous post, simply change any string “hexadecimal character” with the string “word character” !!




    I also think that my previous explanations need to be reread. I’ll see tomorrow !

  • That is awsome, thank you. This is very complex for a pure beginner.
    Do you know if we can do multiple replacements at once in Notepad++ ?
    Like replacing all my 0 with G, all my 1 with H, etc, in one “Replace all”.
    I’m now checking it :)

  • This is not working >
    Find : (($|\G)\w*?\K0)|(($|\G)\w*?\K1)
    Replace : (?1G)(?2H)

    If I have 01001110, then I get GH1GHGH111GH

  • @Jean-Francois-Trehard said in Replacing a digit with exception:

    Do you know if we can do multiple replacements at once in Notepad++ ?

    If you go with my regex, altered to cater for the different changes to be made you can do it ALL in 1 pass.

    So the new regex is
    Find What:(?-s)(;.*$)|(0)|(1)|(2)|(3)|(4)
    Replace With:(?1\1)(?2G)(?3H)(?4I)(?5J)(?6K)

    This is just an example as you have not provided ALL the changes you want, but the idea is the same. My regex can be extended to cater for as many characters as you want changed in the one run.


  • @Terry-R said in Replacing a digit with exception:


    Actually just thinking that you possibly need to change more than 9 groups (in total), the more correct coding would be:
    So this allows for lots of groups to be identified. In your case you would be using …(?{10}X)(?{11}Y)(?{12}Z) as an example.


  • Hi, @jean-francois-trehard, @terry-r and All,

    No problem too. We can do miracles with regexes ;-))

    This time :

    • We must use a non-capturing group at the very beginning of the regex

    • We must add a second non-capturing group to get the \K feature for search of, either, the digits 0, 1 or 3, only

    • We must add inner real groups in order that :

      • Group 1 is defined when a 0 digit is matched

      • Group 2 is defined when a 1 digit is matched

      • Group 3 is defined when a 2 digit is matched

    And the replacement zone is self-explanatory !

    So, given this text :

    dc.l $02A20620 ; Tile #42
    dc.l $0BaC2021
    dc.l $55111003 ; Tile #9A56
    dc.l $55001111
    dc.l $54000116 ; Tile #1b628
    dc.l $dF022222

    The following regex S/R :

    SEARCH (?:\$|\G)\w*?\K(?:(0)|(1)|(2))

    REPLACE (?1G)(?2H)(?3I)

    will output, in one pass, the expected text, leaving the comments untouched

    dc.l $GIAIG6IG ; Tile #42
    dc.l $GBaCIGIH
    dc.l $55HHHGG3 ; Tile #9A56
    dc.l $55GGHHHH
    dc.l $54GGGHH6 ; Tile #1b628
    dc.l $dFGIIIII

    Again, if we use the (?x) modifier for the free-spacing behavior, the search regex becomes :

    (?x)  (?: \$ | \G )  \w  *?  \K  (?:  (0) | (1) | (2)  )

    However, note that the replacement regex cannot be expressed with the free-spacing mode, but, for this specific replacement, the order of the (?#<letter> blocks does not matter !

    So, the replacement regex (?3I)(?2H)(?1G) would be correct, as well !



  • @guy038, @Terry-R,
    Thanks to both of you, this working great now !
    Have a nice day :)

  • Hello, @jean-francois-trehard and All,

    In wanting to explain, my regex S/R, provided in my previous post, I realized that we should add an hypothesis :

    • You must move the caret to a blank line, located before the text to be processed by the S/R

    So, let’s start with the last version :

    SEARCH (?:\$|\G)\w*?\K(?:(0)|(1)|(2))

    REPLACE (?1G)(?2H)(?3I)

    • Globally, the regex (?:\$|\G)\w*?\K(?:(0)|(1)|(2)), searches, from after a literal $ symbol OR after the end of the previous search, if any, for the smallest range, even empty, of words characters, till a 0, 1 or 2 digit, then only selects this last digit and replaces it, respectively, with the G, H or I letters

    • Note that the \G assertion forces the regex engine to ask itself : does the current match attempt, immediately follows the previous match ? In case of a positive answer, the current match attempt is a possible match, so far !

    • As this regex only looks for consecutive words chars, it’s easy to understand why the characters, located, in the second part of a line, after the # sign, are not concerned !

    Now, if you’re still in the fog, let’s go into a little more detail and start the search, for instance, with caret on an empty line, right before the first line dc.l $02220620 ; Tile #42

    Beware, everything below is a bit “off-putting” but, in any case, it happens like that if you break the process down into smaller basic actions !

    • The regex engine first searches for a range of consecutive words characters till a 0, 1 or 2 digit, either, after a literal $ or after the end of the previous search. As no search has been performed, so far, \G syntax matches, by default, at current position, the beginning of the empty line, which is a zero-length string

    • As no word char exists in that empty line, the regex engine skips the EOL chars and moves to the beginning of the next line dc.l $02220620 ; Tile #42. You could say : it should match the dc string, which are word chars, also ? No, no ! Because the initial location, in the empty line, and the dc location are not contiguous. Indeed, there is a gap of the two chars : \r and \n )

    • Thus, the \G assertion is not true, presently. So the regex engine tries to match the first alternative and skips to the literal $ and the next 0 digit to search for

    • The \K feature cancels any match, so far and resets the regex engine working position => So, it only matches this first 0 digit. Remember that this configuration is possible as the range of chars before digit 0, 1 or 2 to search for, may be empty

    • As the group 1 is defined, when matching the 0 digit, the regex engine replaces it with the string G

    • Now, as no more $ symbol exists, the regex engine needs to use the second alternative, the \G assertion, which represents the location right between the letter G and the next range of word chars till a 0, 1 or 2 digit. So it just matches the first 2 digit, right after and, again, only selects that digit 2

    • As the group 3 is defined when matching the 2 digit, the regex engine replaces it with the string I

    • The second, third and fourth 2 digit, of current line, as well as the second 0 digit, are matched, in the same way as above, and replaced, consequently, with the appropriate letter

    • Then, the regex engine matches the 62 digit, right after the previous G letter, which verifies the \G assertion and selects only the fourth digit 2 of current line, due to the \K syntax

    • Again, this 2 digit is replaced with a I letter, as group 3 is defined when matching the 2 digit

    • The regex engine advances one position and matches the last 0 digit of current line and replaces it with the G letter

    • Now, things become interesting : the regex engine must find a next range of consecutive word chars, possibly empty, ending with, either, a 0, 1 or 2 digit. Obviously, this next range of contiguous word chars is the string 42, located some chars after the end of our previous search ! So the \G assertion is not verified anymore and, as there is no other $ symbol, either, the overall search fails. Thus, the regex engine skips the remaining chars of that first line, after the third 0 digit, which has been changed into G and moves to the beginning of the second line where the process resumes !

    • And, when a line just ends with a first range of word chars, without any comments zone, it, necessarily, will search for a literal $ symbol, as the \G feature cannot be true, due to the gap produced by the EOL characters of current line !

    As you may notice, the key point, in this kind of data, is that, the several ranges of words characters are not juxtaposed. So, the \G assertion forces, automatically, the process to cancel any further search after examination of each first range of consecutive words chars of each line, following a $ symbol ;-))

    Note also that, due to the \K syntax, inside this search regex, you cannot use a step by step replacement with several clicks on the Replace button. However, you can use the Find Next button, to get the different matches

    Wow ! Glad to see that you’re still there, after these long explanations ;-)) Thank you for your patience and full attention !

    Best Regards,


  • This post is deleted!

Log in to reply