How to replace third repetition of kate with diane in each line-CodePudding

How to replace third repetition of kate with diane in each line only if william appears in sentence.

kate must be implicit - for example "kate's" is not a valid rep

For example:

kate prince william's wife is the second kate after his mother kate

will be replaced with:

kate prince william's wife is the second kate after his mother diana

but the following one will not:

kate’s prince william's wife is the second kate after his mother kate

CodePudding user response：

So given this input:

kate prince william's wife is the second kate after his mother kate
kate's prince william's wife is the second kate after his mother kate

This remove and replace solution works:

parse.sed

/william/ {
  s/\bkate's\b/XXXXXX/g
  s/\bkate\b/diana/3
  s/XXXXXX/kate's/g
}

Run it like this:

sed -Ef parse.sed infile

Output:

kate prince william's wife is the second kate after his mother diana
kate's prince william's wife is the second kate after his mother kate

CodePudding user response：

Using sed

$ sed -E '/william/s/(^| )kate( |$)/\1diane\2/3' input_file
kate prince william's wife is the second kate after his mother diane
kate’s prince william's wife is the second kate after his mother kate

CodePudding user response：

The most difficult part of this problem is trying to match sentences instead of lines. If you allow a "sentence" to mean "a string delimited by '.'" (which is a terrible definition, and fails to capture many common instances), you might get away with:

perl -0056 -pe 's{\bkate\b}{  $c == 3 ? "Diana" : $&}ige; $c=0' input-file

The -0056 causes perl to treat . (056 is the octal representation of .) as the record separator, so each "sentence" is a single record. Again, this abuses the definition of "sentence" and misses common cases. The i flag causes the matches to be case insensitive. Normally, one would use \b to match word boundaries, but since you don't want kate's to count as a match, you will need to modify this slightly. You haven't precisely defined what you want to consider a valid match, so perhaps:

$ b="(?:(?<=\w)(?=\W|\z|')|(?:(?<=\W)|(?<=\A)|')(?=\w))"
$ perl -0056 -pe 's{'"$b"'kate'"$b"'}{  $c == 3 ? "$1Diana$2" : $&}ige; $c=0' input