Deleting lines that repeat the first 15 characters



  • Hi, @mangoguy, @scott-sumner and All,

    In my initial thread, below, whose Doug spoke of :

    https://notepad-plus-plus.org/community/topic/13147/eliminating-duplicate-identical-lines

    I said :

    The suppression of all the duplicate lines, in a pre-sorted file, can be easily obtained with a Search/Replacement, in Regular expression mode !

    Open your file, containing the sorted list of items

    Open the Replace dialog ( CTRL + H )

    In that text, it’s the word ‘sorted’ which is important ! Indeed, imagine this initial text, NOT sorted :

    pqrst
    pqrst
    pqrst
    uvwxy
    fghij
    fghij
    fghij
    pqrst
    pqrst
    pqrst
    abcde
    abcde
    abcde
    abcde
    fghij
    klmno
    klmno
    klmno
    fghij
    fghij
    

    and my initial regex :

    SEARCH (?-s)(^.+\R)\1+

    REPLACE \1

    Even with a step by step replacement, with the Replace button, it would give :

    pqrst
    uvwxy
    fghij
    pqrst
    abcde
    fghij
    klmno
    fghij
    

    => You can see that it still remains 3 lines fghij, inside, split up on different lines. In a sense, it remains a single line with its two duplicates !

    Now, let’s sort the initial text, first :

    abcde
    abcde
    abcde
    abcde
    fghij
    fghij
    fghij
    fghij
    fghij
    fghij
    klmno
    klmno
    klmno
    pqrst
    pqrst
    pqrst
    pqrst
    pqrst
    pqrst
    uvwxy
    

    After performing the same regex S/R, we, now, get the text :

    abcde
    fghij
    klmno
    pqrst
    uvwxy
    

    And, as expected, it does not contain any duplicate line. So, with an initial sort, this regex, and the derivative regexes, work just fine !

    However, the main drawback is that the original order, of the file, is lost, because of the sort process :-((


    So, against your file, I tried other regexes, which involve a look-ahead and which do not break the file’s order :

    • (?-s)(?:^(.{15}).*\R)(?=.*\1) and EMPTY replacement => Count process give us 1022 occurrences, which is false

    • (?-s)(?:^(.{15}).*\R)(?=(?:.+\R)*\1) => Catastrophic break-down, with only one match ( the entire file contents ! )

    So , what’s about this generic one :

    • (?-s)(?:^(.{15}).*\R)(?=(?:.+\R){0,n}\1)

    Well, but, for instance, 4 lines, in your file, begin with the string 01,12,1215,1012 ( Lines 209988, 208996, 210928 et 210936 ). And the gap between lines 208996 and 210928, for instance, is 1932 !

    So, the number n, in that regex, should be, at least 2000, hence the regex :

    (?-s)(?:^(.{15}).*\R)(?=(?:.+\R){0,2000}\1) => Again, a catastrophic break-down occurred :-((

    Obviously, it is not worth going on, in that direction ! So, Scott, I presume that when a large amount of lines and/or large capturing group areas are involved, we’re going to get, likely, unpredictable results ! We have to change our mind at all about it :-D


    Finally, I found out the right way to get the job done :-)))

    Note that, in the Doug’s file, the key string is the first 15th characters of a line. So, here is the procedure :

    • First, we get rid of possible blank lines, with the regex S/R, below :

    SEARCH ^\h*\R

    REPLACE EMPTY

    => 2 lines should be deleted

    • Then, we add a blank area, of six characters long, right after the 15th character, with the regex S/R, below ( ~23s ) :

    SEARCH (?-s)^.{15}\K

    REPLACE \x20\x20\x20\x20\x20\x20

    • Place the caret at column 19 of the line 1 (IMPORTANT )

    • Now, choose the menu command Edit > Column Editor… ( Alt + C )

    • Select the Number to Insert option

    • Type in 1, as Initial number

    • Type in 1, in the Increase by : option

    • Verify that the chosen format is Dec

    • Check the Leading zeros option ( IMPORTANT )

    • Finally, click on the OK button ( ~36s )

    => After a while, you should get a six-digits column, at position 19, from 000001 to 460726

    • Delete the last line 460726 and the EOL characters of the line 460725

    • Sort the file contents, choosing the menu command Edit > Line Operations > Sort Lines Lexicographically Ascending ( ~17s )

    • Now, perform the main regex, below, clicking on the Replace All button, exclusively ( ~1m 11s )

    SEARCH (?-s)(.{15}).*\R\K(?:\1.*\R)+

    REPLACE EMPTY

    => 29363 replacements occurred and the file, from now on, contains 430346 lines, only !

    • Then, move the middle column number, at beginning of each line, with the regex S/R, below ( ~31s ) :

    SEARCH (?-s)^(.+?)\x20+(\d+)

    REPLACE \2\x20\x20\x20\1

    • Now, execute a last sort operation Edit > Line Operations > Sort Lines Lexicographically Ascending ( ~15s )

    • Finally get rid of the line numbers, at beginning of lines, along with the space characters, with the regex S/R, below ( ~15s ) :

    SEARCH ^\d+\x20+

    REPLACE EMPTY

    Et voilà !

    To sump up, this procedure, while going downwards, throughout the file contents, keeps, only, each first occurrence of any duplicate line, with identical first 15th characters, as well as any single line, of course !

    Cheers,

    guy038

    P.S. :

    To easily understand my method, above, copy/paste this short initial text, below, in a new tab :

    pqrstSmith
    pqrstJones
    pqrstTaylor
    uvwxyBrown
    fghijWilliams
    fghijWilson
    fghijJohnson
    pqrstDavies
    pqrstRobinson
    pqrstWright
    abcdeThompson
    abcdeEvans
    abcdeWalker
    abcdeWhite
    fghijRoberts
    klmnoGreen
    klmnoHall
    klmnoWood
    fghijJackson
    fghijClarke
    

    Then, I decided that two lines are identical if their first 5 characters are equal !

    So, we add, first, a six blank characters column, after the 5th character of each line

    SEARCH (?-s)^.{5}\K

    REPLACE \x20\x20\x20\x20\x20\x20

    pqrst      Smith
    pqrst      Jones
    pqrst      Taylor
    uvwxy      Brown
    fghij      Williams
    fghij      Wilson
    fghij      Johnson
    pqrst      Davies
    pqrst      Robinson
    pqrst      Wright
    abcde      Thompson
    abcde      Evans
    abcde      Walker
    abcde      White
    fghij      Roberts
    klmno      Green
    klmno      Hall
    klmno      Wood
    fghij      Jackson
    fghij      Clarke
    

    After the Column Editor operation, at column 9, we get :

    pqrst   01   Smith
    pqrst   02   Jones
    pqrst   03   Taylor
    uvwxy   04   Brown
    fghij   05   Williams
    fghij   06   Wilson
    fghij   07   Johnson
    pqrst   08   Davies
    pqrst   09   Robinson
    pqrst   10   Wright
    abcde   11   Thompson
    abcde   12   Evans
    abcde   13   Walker
    abcde   14   White
    fghij   15   Roberts
    klmno   16   Green
    klmno   17   Hall
    klmno   18   Wood
    fghij   19   Jackson
    fghij   20   Clarke
    

    And, after an ascending sort :

    abcde   11   Thompson
    abcde   12   Evans
    abcde   13   Walker
    abcde   14   White
    fghij   05   Williams
    fghij   06   Wilson
    fghij   07   Johnson
    fghij   15   Roberts
    fghij   19   Jackson
    fghij   20   Clarke
    klmno   16   Green
    klmno   17   Hall
    klmno   18   Wood
    pqrst   01   Smith
    pqrst   02   Jones
    pqrst   03   Taylor
    pqrst   08   Davies
    pqrst   09   Robinson
    pqrst   10   Wright
    uvwxy   04   Brown
    

    Then, the regex S/R, where we just change number 15 by number 5, suppresses all the duplicate lines, but one :

    SEARCH (?-s)(.{5}).*\R\K(?:\1.*\R)+

    REPLACE EMPTY

    And give us :

    abcde   11   Thompson
    fghij   05   Williams
    klmno   16   Green
    pqrst   01   Smith
    uvwxy   04   Brown
    

    Now, the following regex S/R, swap the line numbers area and the key area ( the first 5th characters )

    SEARCH ^(?-s)(.+?)\x20+(\d+)

    REPLACE \2\x20\x20\x20\1

    11   abcde   Thompson
    05   fghij   Williams
    16   klmno   Green
    01   pqrst   Smith
    04   uvwxy   Brown
    

    And, after a last ascending sort operation, we get :

    01   pqrst   Smith
    04   uvwxy   Brown
    05   fghij   Williams
    11   abcde   Thompson
    16   klmno   Green
    

    Finally, a last regex S/R, below, delete all the line numbers, at beginning of lines :

    SEARCH ^\d+\x20+

    REPLACE EMPTY

    pqrst   Smith
    uvwxy   Brown
    fghij   Williams
    abcde   Thompson
    klmno   Green
    


  • Thank you again.

    My bad for thinking a nonsorted file would not be a problem.

    Despite multiple attempts, every time I try to use the column editor, despite putting the caret “^” at column 19 on row 1, the incremental numerical column appears at position 1 and not position 19.

    Any thoughts as to what I am doing wrong?

    Thank you,
    Doug



  • @mangoguy

    What do you mean by ^ for the caret? I know that sometimes ^ is referred to as the caret character, but it is not in Notepad++ so I’m confused.

    Anyway, Here’s what I do and it works to insert at col 19:

    • move caret to line 1 col 19
    • press Alt+c to get Column Editor window
    • tick Number to Insert
    • specify Initial number of 0
    • specify Increase by of 1
    • specify an empty field for Repeat
    • tick Leading zeros
    • specify Dec for Format
    • click OK

    Notepad++ places incrementing numbers in col 19 (and beyond) throughout the length of my document.

    Are you doing something very different from this?



  • This post is deleted!


  • Thank you for the clarification. It worked perfectly with the file exactly as instructed. Thank you!

    With another pre-sorted file which has no duplicates or blank lines, when I perform the main regex to find and remove duplicate lines
    (?-s)(.{15}).\R\K(?:\1.\R)+

    the replace box returns: “Replace All:1 occurrence was replaced” no matter how many times I repeat the replace. If there are no duplicates I would expect a report of 0 occurrences found.

    The file is found at
    https://mangoguy.sharefile.com/d-s7b2d2a8b3fb459cb

    Thank you,
    Doug



  • @mangoguy said:

    Replace All:1 occurrence was replaced

    Formatting note: Your regular expression was stated as (?-s)(.{15}).\R\K(?:\1.\R)+ but I think you really meant (?-s)(.{15}).*\R\K(?:\1.*\R)+ as per one of @guy038 's regexes above. In the future, wrap any exact text you want to post here in ` (backticks) to hopefully avoid any confusion. For example, if you type in `hello` it should appear here as hello without any special characters having trouble. You can also start a new line with four spaces and then your text to provide some data that won’t be specially interpreted.

    I see the same behavior as you when trying this regex replacement on your newest data file. Note that the file is NOT modified by this replacement (disk icon on its tab remains blue after the “replacement” occurs…starting point was a freshly loaded DATA2.txt file). I’m at a loss to explain this (why it is saying “1 replacement”). This thread has brought out some really odd things!

    Note that it IS possible to see non-zero replacements listed and have a file NOT be modified (try a Find-what of ^ and a Replace-with of $0, also Reg exp search mode), but this is very different from your replacement action.



  • Hi, @mangoguy, @scott-sumner and All,

    To begin with, Doug, I was a bit surprised that, both, the numbers, at column 19 and the first 15th characters look equally sorted, in your Data2.txt file ! So I hope that you understood that the first sort must be performed, after the use of the Column Editor. Indeed, these numbers are just added in order to get the original order back, after the suppression of all the duplicate lines ! Just a remark :-))

    Now, mangoguy and others, keep in mind that, when a rather complicated regex is applied, against an important file, a complete failure may occur, with only 1 match which represents, simply, the selection of all the file contents :-((

    So, I began to investigate this problem, more deeply ! First of all, I verified that the first 15th characters, of your Data2.txt file, had absolutely no duplicate And, like Scott and you, I noticed that the regex (?-s)(.{15}).*\R\K(?:\1.*\R)+, wrongly selects the whole file, after a while, instead of finding 0 result


    At this point, I simply thought about reducing the file to reach the upper value, beyond we get into trouble. It happened, that, with my old Win XP laptop, the limit is 67,000 lines about. For this value, you get the correct result : no match. But, for instance, with 67,100 lines, we get the non-correct one match !

    Note that using the similar regex (?-s)(.{15}).*\R\K(?:\1.*\R), without the + sign, at its end, this limit increases to 68,830 lines about !


    So I was wondering : Could it be that the lack of matches, with the necessity to scan great amount of data, causes that false positive ? So, strangely, I decided to add false positives every 65,000 lines about, as below :

    ---------------
    ---------------
    

    So, I added these two lines of 15 dashes, at lines 65,000, 130,000, 195,000, 260,000, 325,000, 390,000 and 455,000. In addition, I duplicated the first line as well as the last line of the file.

    If my intuition was correct, the regex would match, of course, all the second lines of dashes ( false positives ) but also, the first duplicate, in line 2 and the second duplicate, at end of file. This would prove that the search process can go on, normally, throughout an important file ! I ran a Find All in Current Document process and… Bingo ! I obtained the Find Result panel, below, with the expected results :

    Search "(?-s)(.{15}).*\R\K(?:\1.*\R)+" (9 hits in 1 file)
      new 1 (9 hits)
    	Line 2: 01,02,2013,1000   000001   ,22.107,22.513,20.976,21.151,0
    	Line 65003: ---------------
    	Line 130002: ---------------
    	Line 195002: ---------------
    	Line 260002: ---------------
    	Line 325002: ---------------
    	Line 390002: ---------------
    	Line 455002: ---------------
    	Line 458420: 12,31,2015,2559   458404   ,3.270,3.270,3.538,3.527,0
    

    Therefore, it seems that a too important gap, between two successive matches, causes the complete failure of the regex search process !? I just hope that, for most of users, this gap of 65000 lines about( perhaps, we’d better speak about bytes ! ), noted with my outdated laptop, can really be greater :-))


    Instead of adding some false positives, in huge files, we could, also, search for a string, which would occur every x lines ! For instance, starting with the Data2.txt file, I build a file, made of five times Data2.txt : I just changed the first character of each line, taking, successively, 3 and 4, then 5 and 6,… instead of 0 and 1, in order to keep a list of lines, without any duplicate :-)

    This file contained 126,274,854 bytes and 2,292,022 lines. So, I decided that, in addition to the detection of duplicates, with the regex (?-s)(.{15}).*\R\K(?:\1.*\R)+, I would search for lines 50,000, 100,000, and so on…, with the regex (5|0)0000\x20 To that purpose, I just used the list of numbers, at column 19, copied five times !

    So the final regex is , simply, the two alternatives : (?-s)(.{15}).*\R\K(?:\1.*\R)+|(5|0)0000\x20. Again, I clicked on the Find All in Current Document button and, …after 6m 49s( Waoooou ! ) , the Find Result displayed, at last :

    Search "(?-s)(.{15}).*\R\K(?:\1.*\R)+|(5|0)0000\x20" (47 hits in 1 file)
      new 1 (47 hits)
    	Line 2: 01,02,2013,1000   000001   ,22.107,22.513,20.976,21.151,0
    	Line 50001: 02,11,2014,2536   050000   ,0.357,0.380,0.270,0.310,0
    	Line 100001: 03,24,2014,1115   100000   ,5.494,5.191,5.494,5.299,0
    	Line 150001: 05,05,2017,1346   150000   ,0.301,0.301,0.270,0.289,0
    	Line 200001: 06,13,2013,1107   200000   ,0.519,0.588,0.516,0.588,0
    	Line 250001: 07,23,2013,1437   250000   ,0.070,0.064,0.073,0.071,0
    	Line 300001: 09,04,2013,1158   300000   ,2.314,2.368,2.314,2.362,0
    	Line 350001: 10,06,2017,1031   350000   ,0.201,0.138,0.201,0.151,0
    	Line 400001: 11,08,2012,1254   400000   ,1.263,1.253,1.284,1.284,0
    	Line 450001: 12,21,2012,1043   450000   ,3.838,3.815,3.858,3.823,0
    	Line 508405: 22,11,2014,2536   050000   ,0.357,0.380,0.270,0.310,0
    	Line 558405: 23,24,2014,1115   100000   ,5.494,5.191,5.494,5.299,0
    	Line 608405: 25,05,2017,1346   150000   ,0.301,0.301,0.270,0.289,0
    	Line 658405: 26,13,2013,1107   200000   ,0.519,0.588,0.516,0.588,0
    	Line 708405: 27,23,2013,1437   250000   ,0.070,0.064,0.073,0.071,0
    	Line 758405: 29,04,2013,1158   300000   ,2.314,2.368,2.314,2.362,0
    	Line 808405: 30,06,2017,1031   350000   ,0.201,0.138,0.201,0.151,0
    	Line 858405: 31,08,2012,1254   400000   ,1.263,1.253,1.284,1.284,0
    	Line 908405: 32,21,2012,1043   450000   ,3.838,3.815,3.858,3.823,0
    	Line 966809: 42,11,2014,2536   050000   ,0.357,0.380,0.270,0.310,0
    	Line 1016809: 43,24,2014,1115   100000   ,5.494,5.191,5.494,5.299,0
    	Line 1066809: 45,05,2017,1346   150000   ,0.301,0.301,0.270,0.289,0
    	Line 1116809: 46,13,2013,1107   200000   ,0.519,0.588,0.516,0.588,0
    	Line 1166809: 47,23,2013,1437   250000   ,0.070,0.064,0.073,0.071,0
    	Line 1216809: 49,04,2013,1158   300000   ,2.314,2.368,2.314,2.362,0
    	Line 1266809: 50,06,2017,1031   350000   ,0.201,0.138,0.201,0.151,0
    	Line 1316809: 51,08,2012,1254   400000   ,1.263,1.253,1.284,1.284,0
    	Line 1366809: 52,21,2012,1043   450000   ,3.838,3.815,3.858,3.823,0
    	Line 1425213: 62,11,2014,2536   050000   ,0.357,0.380,0.270,0.310,0
    	Line 1475213: 63,24,2014,1115   100000   ,5.494,5.191,5.494,5.299,0
    	Line 1525213: 65,05,2017,1346   150000   ,0.301,0.301,0.270,0.289,0
    	Line 1575213: 66,13,2013,1107   200000   ,0.519,0.588,0.516,0.588,0
    	Line 1625213: 67,23,2013,1437   250000   ,0.070,0.064,0.073,0.071,0
    	Line 1675213: 69,04,2013,1158   300000   ,2.314,2.368,2.314,2.362,0
    	Line 1725213: 70,06,2017,1031   350000   ,0.201,0.138,0.201,0.151,0
    	Line 1775213: 71,08,2012,1254   400000   ,1.263,1.253,1.284,1.284,0
    	Line 1825213: 72,21,2012,1043   450000   ,3.838,3.815,3.858,3.823,0
    	Line 1883617: 82,11,2014,2536   050000   ,0.357,0.380,0.270,0.310,0
    	Line 1933617: 83,24,2014,1115   100000   ,5.494,5.191,5.494,5.299,0
    	Line 1983617: 85,05,2017,1346   150000   ,0.301,0.301,0.270,0.289,0
    	Line 2033617: 86,13,2013,1107   200000   ,0.519,0.588,0.516,0.588,0
    	Line 2083617: 87,23,2013,1437   250000   ,0.070,0.064,0.073,0.071,0
    	Line 2133617: 89,04,2013,1158   300000   ,2.314,2.368,2.314,2.362,0
    	Line 2183617: 90,06,2017,1031   350000   ,0.201,0.138,0.201,0.151,0
    	Line 2233617: 91,08,2012,1254   400000   ,1.263,1.253,1.284,1.284,0
    	Line 2283617: 92,21,2012,1043   450000   ,3.838,3.815,3.858,3.823,0
    	Line 2292022: 92,31,2015,2559   458404   ,3.270,3.270,3.538,3.527,0
    

    As you can see, the duplicate line 2 and the second duplicate, at line 2,292,022, were correctly found and reported !


    Conclusion :

    Apparently, when a too important amount of text separates two consecutive occurrences of the regex search, it breaks the normal process, getting, wrongly, a single selection of all file contents !? So, Mangoguy, as no duplicate exists in your data2.txt file, it’s obvious that we’re going into trouble as soon as your file exceeds a certain size limit !

    In other words, if, in huge files, you get a lot of occurrences, throughout the file contents, this should help the search process to correctly finish the job :-))

    Best Regards,

    guy038



  • So @guy038’s results and conclusions are interesting. I decided to see what would happen if a Pythonscript-based search was conducted. To that end I came up with:

    matches = []
    def match_found(m): matches.append(m.span(0))
    editor.research(r'(?-s)(.{15}).*\R\K(?:\1.*\R)+', match_found)
    for (start, _) in matches: print editor.lineFromPosition(start) + 1
    print 'done'
    

    With that script and the DATA2.txt file, I found that with 67025 lines in the file I would see “done” printed in the PS console window, but with one more line, 67026, I would get this:

    Traceback:
        editor.research(r'(?-s)(.{15}).*\R\K(?:\1.*\R)+', match_found)
    <type 'exceptions.RuntimeError'>:  The complexity of matching the regular expression exceeded predefined bounds.  Try refactoring the regular expression to make each choice made by the state machine unambiguous.  This exception is thrown to prevent "eternal" matches that take an indefinite period time to locate.
    

    This seems consistent with @guy038’s findings that somewhere between 67000 and 67100 lines there is a “problem”.

    So I think the meaning of all this is that Notepad++ is not a great tool for the OP’s task. :-(

    No one wants to be trying to solve one problem, only to encounter problems with the method they are using to solve that problem. Thus, I’d advise, if this is a recurring need, to have a serious look at the short bit of standard Python (or rewrite in your language of choice) that I provided much earlier in this thread. :-D



  • Hello, @mangoguy, @scott-sumner and All,

    I’m extremely confused, Indeed ! I did an important and beginner mistake, in my previous regex, that I was testing, intensively :-(( My God, of course ! The RIGHT regex is (?-s)^(.{15}).*\R\K(?:\1.*\R)+ and NOT the regex (?-s)(.{15}).*\R\K(?:\1.*\R)+ :-))

    Do you see the difference ? Well, it’s just the anchor ^, after the modifier (?-s) !

    Indeed, let’s try again the wrong regex :

    Assuming the test list, below :

    91,02,2013,1000   000001   ,22.107,22.513,20.976,21.151,0
    13,1000   000002   ,20.976,21.724,20.620,21.336,0
    13,1000   000003   ,21.344,22.116,21.336,21.918,0
    13,1000   000004   ,21.918,21.918,20.797,20.797,0
    

    So, first, the caret is right before the 9 digit, of the first line and the fifteen characters 91,02,2013,1000 cannot be found elsewhere. Then, as no anchor ^ ( beginning of line ) exists, the regex engine goes ahead one position between the digits 9 and 1 of the first line. Again, as the fifteen characters 1,02,2013,1000b do not exist further on, the regex engine goes ahead one position, examining, now the string ,02,2013,1000bb

    … till the fifteen characters 13,1000bbb00000, which can be found, this time, at beginning of lines 2, 3 and 4 ! Just imagine the work to accomplish for 458,404 lines of the Data2.txt file :-((

    ( Note : the lowercase letter b, above, stands for a space character )

    To easily see the problem, just get rid of the \K syntax, forming the regex (?-s)(.{15}).*\R(?:\1.*\R)+. If you click on the Find Next button, it selects, after test on positions 1, 2,…and 8, from the two last digits of year 2013 till the end of text. But, if you’re using the regex (?-s)^(.{15}).*\R(?:\1.*\R)+, with the anchor ^, it correctly gets the identical lines 2, 3 and 4, regarding theirs first 15 characters !


    So, Doug, to sump up, using the right regex (?-s)^(.{15}).*\R\K(?:\1.*\R)+, against your Data2.txt file, does not find any occurrence ( ~5s ), that is the expected result, as we know, by construction, that the 458,404 lines of this file, are all different :-)

    Best Regards,

    guy038



  • @guy038

    Yea, wow, I totally didn’t see the missing ^ as well. Of course, as our local regex guru I don’t normally question @guy038’s regexes, but there is no excuse for a second pair of eyes (mine) not noticing/questioning this. Looking back over my posts in this thread, I really added nothing of value and totally wish I hadn’t participated at all. :-(


Log in to reply