The first line will be a certain string repeated any amount of times. For example, it could be abcabcabcabc, [];[];[];, etc. It may be cut off; for example: 1231231231. Always find the shortest string; for example, if the line is 22222, then the string is 2, not 22 or 22222 or anything else. The string will always be repeated at least 2 full times.

All subsequent lines will be that pattern offset by any number. For example, it could be:


(offset by 1), or it could be:


(offset by 4).

One of the characters in the input will be wrong. (It is guaranteed not to be on the first line.) For example, in this input:


the 1 on line 3 is the odd one out.


You must output the (zero-based, starting from the top left) coordinates of the odd one out. For example, in the above input, the corresponding output is 4,2. You may also output 4 2, or "4""2", or even [[4],[2]], or any other format, as long as you can tell what the output is supposed to be.

Test cases



Output: 16,2



Output: 8,3


. .

Output: 1,1



Output: 4,2


  • \$\begingroup\$ What characters can be contained in the string? Printable ASCII? ASCII? Unicode? \$\endgroup\$ – Dennis Mar 20 '14 at 22:27
  • \$\begingroup\$ @Dennis Just printable ASCII (that can basically be assumed for any challenge involving strings; otherwise we would have to specify that for almost every challenge :P) \$\endgroup\$ – Doorknob Mar 20 '14 at 22:28
  • \$\begingroup\$ I guessed so. I'm thinking about an approach that would require an unused char, so I thought I'd ask. \$\endgroup\$ – Dennis Mar 20 '14 at 22:31
  • \$\begingroup\$ Should we check case like this: abc/cab/abc - and output 0 2 here? \$\endgroup\$ – user2846289 Mar 21 '14 at 13:24
  • \$\begingroup\$ @VadimR No, since only one character will be wrong. \$\endgroup\$ – Doorknob Mar 21 '14 at 14:13

Bash Perl, 231 229 218 178 164 166 138 106 74 bytes


The script requires using the -n switch, which accounts for two of the bytes.

The idea of appending two copies of all full repetitions of the pattern has been taken from MT0's answer.

In contrast to all other answers, this approach attempts to extract the pattern of the current input line in each iteration; it will fail on the line containing the odd character (and use the pattern of the previous line instead). This is done to include the pattern extraction in the loop, which manages to save a few bytes.

Ungolfed version

#!/usr/bin/perl -n

# The `-n' switch makes Perl execute the entire script once for each input line, just like
# wrapping `while(<>){…}' around the script would do.


# This regular expression matches if `((.*).*)' - accessible via the backreference `\2' -
# is repeated at least once, followed by a single repetition of `\3" - zero or more of the
# leftmost characters of `\2' - followed by the end of line. This means `\1' will contain
# all full repetitions of the pattern. Even in the next loop, the contents of `\1' will be
# available in the variable `$1'.


# Append two copies of `$1' to the current line. For the line, containing the odd
# character, the regular expression will not have matched and the pattern of the previous
# line will get appended.
# Since the pattern is repeated at least two full times, the partial pattern repetition at
# the end of the previous line will be shorter than the string before it. This means that
# the entire line will the shorter than 1.5 times the full repetitions of the pattern, 
# making the two copies of the full repetitions of the pattern at least three times as 
# long as the input lines.

$.-- , die $+[1] if

# If the regular expression below matches, do the following:
#   1. Decrement the variable `$.', which contains the input line number.
#      This is done to obtain zero-based coordinates.
#   2. Print `$+[1]' - the position of the last character of the first subpattern of the
#      regular expression - plus some additional information to STDERR and exit.
#      Notably, `die' prints the (decremented) current line number.


# `(.*)(.)(.*)', enclosed by `^' and a newline, divides the current input line into three
# parts, which will be accesible via the backreferences `\1' to `\3'. Note that `\2'
# contains a single character.
# `.*\1(?!\2).\3' matches the current input line, except for the single character between
# `\1' and `\3' which has to be different from that in `\2', at any position of the line
# containing the pattern repetitions. Since this line is at least thrice as long as
# `\1(?!\2).\3', it will be matched regardless of by how many characters the line has been
# rotated.


For the test case


the output of the golfed version is

16 at script.pl line 1, <> line 2.

meaning that the odd character has the coordinates 16,2.

This blatantly abuses takes advantage of the liberal output format.

Just before exiting, the contents of some of Perl's special variables are:

$_  = lfcodegolfcodegoff\ncodegolfcodegolfcodegolfcodegolf
$1  = lfcodegolfcodego
$2  = f
$3  = f

($n contains the match of the subpattern accessible via the backreference \n.)

  • \$\begingroup\$ Clever catching of the reply unit. It can be optimized by one byte: ^((.*?)(.*?))(?=\1+\2$) \$\endgroup\$ – Heiko Oberdiek Mar 21 '14 at 10:33
  • \$\begingroup\$ I switched to the language the popular kids are using. Can probably be golfed down further; this is my first Perl script in over a decade... \$\endgroup\$ – Dennis Mar 25 '14 at 22:06
  • 2
    \$\begingroup\$ ...and you're over a decade late if you think perl is what the popular kids are using \$\endgroup\$ – ardnew Mar 26 '14 at 3:32
  • \$\begingroup\$ this answer isn't getting the love it deserves. looks to me like the winner @Doorknob \$\endgroup\$ – ardnew Apr 1 '14 at 3:19

Perl, 212 191 181 168 bytes

  • This version uses an optimized trick for catching the reply unit, learned in Dennis' answer.
  • Optimization by using the property that all lines have equal lengths.
  • Line end is also needed for the last line, otherwise chomp instead of chop should be used.
  • Optimizations of ardnew's comment added.

Old version, 212 bytes:

$_=<>;chop;/^(.+?)\1+(??{".{0,".(-1+length$1).'}'})$/;$;=$1;while(<>){$x=$;x length;chop;$x=~/\Q$_\E/&&next;for$i(0..-1+length$_){for$r(split//,$;){$b=$_;$b=~s/(.{$i})./$1$r/;$x=~/\Q$b\E/&&exit print$i,$",$.-1}}}

Ungolfed version:

$_ = <>;  # read first line
# The repeat unit \2 consists of \3 and \4,
# and the start part \2 can be added at the end (as partial or even full unit).
$x = $1 x 4; # $x is long enough to cover each following line

# Old version:
# /^(.+?)\1+(??{ ".{0," . (-1 + length $1) . '}' })$/;
# $a = $1; # $a is the repeat unit.
# The unit is caught by a non-greedy pattern (.+?) that is
# repeated at least once: \1+
# The remaining characters must be less than the unit length.
# The unit length is known at run-time, therefore a "postponed"
# regular expression is used for the remainder.

# process the following lines until the error is found
while (<>) {
    # old version:
    # $x = $a x length;
    # $x contains the repeated string unit, by at least one unit longer
    # than the string in the current line
    chop; # remove line end of current line
    $x =~ /\Q$_\E/ && next;
          # go to next line, if current string is a substring of the repeated units;
          # \Q...\E prevents the interpretation of special characters
    # now each string position $x is checked, if it contains the wrong character:
    for $i (0 .. y///c - 1) {  # y///c yields the length of $_
        for $r (split //, $x) { #/ (old version uses $a)
            # replace the character at position $i with a
            # character from the repeat unit
            $b = $_;
            $b =~ s/(.{$i})./$1$r/;
            $x =~ /\Q$b\E/
               && die $i, $", $. - 1, $/;
               # $" sets a space and the newline is added by $/;
               # the newline prevents "die" from outputting line numbers
  • \$\begingroup\$ Great solution and comments, I need to learn more regex ;) \$\endgroup\$ – Newbrict Mar 21 '14 at 4:13
  • 1
    \$\begingroup\$ the first chop is unnecessary- should be removed. the final exit print can be replaced with die (add ,$/ to hide the extra stuff (if required)). also length$_ can be replaced with y///c \$\endgroup\$ – ardnew Mar 25 '14 at 19:44
  • \$\begingroup\$ @ardnew: Many thanks, I have removed the first chop, because $ matches before the newline at the end of the string. Hiding the extra stuff of die via the added newline seems necessary to me. Also y///c is much shorter than length$_ and one byte shorter than length without the unnecessary $_. \$\endgroup\$ – Heiko Oberdiek Mar 26 '14 at 12:47
  • 1
    \$\begingroup\$ @ardnew: I had forgotten about die's verbosity. It even includes prints the line number! I'll use that in my next update. \$\endgroup\$ – Dennis Mar 26 '14 at 16:47

C, 187 bytes


  • Don't use input strings longer than 98 characters :)

Golfed version

char s[99],z[99],*k,p,i,I,o,a;c(){for(i=0;k[i]==s[(i+o)%p];i++);return k[i];}main(){for(gets(k=s);c(p++););for(;!(o=o>p&&printf("%d,%d\n",I,a))&&gets(k=z);a++)while(o++<p&&c())I=I<i?i:I;}

Ungolfed version

char s[99],z[99],*k,p,i,I,o,a;

    return k[i];

    for(;!(o=o>p&&printf("%d,%d\n",I,a)) && gets(k=z);a++)
           while(o++ < p && c())

Python, 303 292

g=s[:[all((lambda x:x[1:]==x[:-1])(s[n::k])for n in R(k))for k in R(1,l)].index(True)+1]*l*2
while 1:
 z=[map(lambda p:p[0]==p[1],zip(t,g[n:l+n]))for n in R(l)]
 any(all(y)for y in z)or exit("%d,%d"%(max(map(lambda b:b.index(False),z)),m))

Input goes through stdin. I'll explain it if there's any demand, but it doesn't look like I'm going to win anyway.


Perl, 157 154

Edit: -3 thanks to ardnew' suggestion.

<>=~/^(((.*?).*?)\2+)\3$/;$p=$2;$n=$+[1];while(<>){s/.{$n}/$&$&/;/(\Q$p\E)+/g;$s=$p;1while/./g*$s=~/\G\Q$&/g;print$n>--($m=pos)?$m:$m-$n,$",$.-1,$/if pos}

It took me some time (on and off, of course, not 5 days ;-) ), and idea about algorithm was initially elusive (though I felt it was there), but finally (and suddenly) it became all clear.

If string length is multiple of pattern length, and even if the string doesn't start with the beginning of the pattern, concatenating string with itself will produce pattern in place of concatenation (imagine infinite repetition of a word on circular ribbon - the place of welding is not important). So, the idea is to trim the line to multiple unit length and concatenate original to it. The result, even for the string containing the wrong character, is guaranteed to match to pattern at least once. From there on it's easy to find position of offending character.

First line is shamelessly borrowed from Heiko Oberdiek's answer :-)

<>=~/^(((.*?).*?)\2+)\3$/;      # Read first line, find the repeating unit
$p=$2;                          # and length of whole number of units.
$n=$+[1];                       # Store as $p and $n.
while(<>){                      # Repeat for each line.
    s/.{$n}/$&$&/;              # Extract first $n chars and
                                # append original line to them.
    /(\Q$p\E)+/g;               # Match until failure (not necessarily from the
                                # beginning - doesn't matter).
    $s=$p;                      # This is just to reset global match position
                                # for $s (which is $p) - we could do without $s,
                                # $p.=''; but it's one char longer.
                                # From here, whole pattern doesn't match -
    1while/./g*$s=~/\G\Q$&/g;   # check by single char.
                                # Extract next char (if possible), match to 
                                # appropriate position in a pattern (position 
                                # maintained by \G assertion and g modifier).
                                # We either exhaust the string (then pos is 
                                # undefined and this was not the string we're
                                # looking for) or find offending char position.

    print$n>--($m=pos)?$m:$m-$n,$",$.-1,$/if pos
  • 1
    \$\begingroup\$ nice work. i think you can replace /.{$n}/;$_=$&.$_; with s/.{$n}/$&$&/; \$\endgroup\$ – ardnew Mar 25 '14 at 19:52

JavaScript (ES6) - 147 133 136 Characters


Expects the string to be tested to be in the variable s and outputs the result to the console.

var repetitionRE = /^(((.*).*)\2+)\3\n/;
                                        // Regular expression to find repeating sequence
                                        // without any trailing sub-string of the sequence.
var sequence = repetitionRE.exec(s)[1]; // Find the sequence string.
s.split('\n')                           // Split the input into an array.
   ( row, index ) =>                    // Anonymous function using ES6 arrow syntax
     var testStr = row + '᛫'+ sequence + sequence;
                                        // Concatenate the current row, a character which won't
                                        // appear in the input and two copies of the repetitions
                                        // of the sequence from the first line.
     var match = /^(.*)(.)(.*)᛫.*\1(?!\2).\3/.exec(testStr);
                                        // Left of the ᛫ finds sub-matches for a single
                                        // character and the sub-strings before and after.
                                        // Right of the ᛫ looks for any number of characters
                                        // then the before and after sub-matches with a
                                        // different character between.
      if ( match )
       console.log( match[1].length, index );
                                        // Output the index of the non-matching character
                                        // and the row.

Test Case 1



16 2

Test Case 2



8 3

Test Case 3

s="...\n. .\n..."


1 1

Test Case 4



4 2

Test Case 5



0 1

Test Case 6



6 1
  • \$\begingroup\$ Sadly, this approach fails if, e.g., s="xyxy\nyyxy". For the second line, match[4] will be yy; it should be just y. \$\endgroup\$ – Dennis Mar 27 '14 at 3:05
  • \$\begingroup\$ Re-worked and shortened by 14 characters. \$\endgroup\$ – MT0 Mar 27 '14 at 20:44
  • \$\begingroup\$ Very nice! I had tried the very same second regex at some point, but I appended the minimal pattern twice instead of the maximal one (and, thus, failed miserable). One minor issue: The first regex will report abab ast the pattern of ababaababa; you need to use ^…$. \$\endgroup\$ – Dennis Mar 27 '14 at 23:37
  • \$\begingroup\$ /^…\n/ works or /^…$/m \$\endgroup\$ – MT0 Mar 28 '14 at 0:05
  • 1
    \$\begingroup\$ It may not need the leading ^ (at least it doesn't for any of the 6 test case I've listed - but there is probably a counter example where it does so I've left it in). \$\endgroup\$ – MT0 Mar 28 '14 at 0:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.