仅替换最后一条匹配行（perl单行）

Question

I have multiple files of the form

version 'aaa'
other 'bbb'
another 'ccc'
version 'ddd'
onemore 'eee'

Some have one version , others have multiple; same with the other keys, but the values never repeat. I'm using, as part of a bigger bash function, a perl one-liner to modify values

modify_value() {
  key_to_modify="$1"
  new_value="$2"

  perl -i'' -pe "s|^(\s*)${key_to_modify} .*|\1${key_to_modify} ${new_value}|" "${file}"
}

The indentation on the lines varies and is unpredictable, but should be respected (hence the need for ^(\\s*) ). This function works great to an extent. I can do

modify_value "onemore" "fff"

And it will be correctly replaced in the text file. However, where it breaks down is where I have multiple keys with the same name (such as the aforementioned version ), as this change will be made in all of them. In my particular case, I want the modification to be made always in the last case .

Since values are never repeated, so far what I have is

modify_value() {
  key_to_modify="$1"
  new_value="$2"

  last_key=$(cat "${file}" | grep "^\s*${key_to_modify}" | tail -1 | perl -pe 's/^\s*//')

  perl -i'' -pe "s|^(\s*)${last_key}|\1${key_to_modify} ${new_value}|" "${file}"
}

This works, but is a bit inelegant. Would it be possible to leverage the perl one-liner to act only on the latest occurrence of the match, instead?

Answer 1

You might be tempted to use Tie::File.

# Borodin's solution with the bug fixes I mention below.
perl -MTie::File -e'
   $key  = shift(@ARGV);
   $val  = shift(@ARGV);
   $file = shift(@ARGV);
   tie @f, "Tie::File", $file;
   for (reverse @f) { last if s/^\s*\Q$key\E\s\K.*/$val/; }
' "$1" "$2" "$file"

For small files, Tie::File will provide a solution that's slower than the alternatives and that uses more memory than the alternatives

For large files, Tie::File will provide an abysmally slow solution to this problem, although it will use less memory than loading the entire file into memory.

You really can't do any worse than using Tie::File for this problem.

Here's an alternative:

perl -i -e'
   $key = shift(@ARGV);
   $val = shift(@ARGV);
   my @f = reverse(<>);
   for (@f) { last if s/^\s*\Q$key\E\s\K.*/$val/; }
   print reverse(@f);
' "$1" "$2" "$file"

You could even avoid the double reversing by having the substitution operator find the last match.

# 5.14+
perl -0777 -i -e'
   $key = shift(@ARGV);
   $val = shift(@ARGV);
   print <> =~ s/\A.*^\s*\Q$key\E\s\K[^\n]*/$val/smr;
' "$1" "$2" "$file"

or

perl -0777 -i -e'
   $key = shift(@ARGV);
   $val = shift(@ARGV);
   $_ = <>;
   s/\A.*^\s*\Q$key\E\s\K[^\n]*/$val/sm;
   print;
' "$1" "$2" "$file"

or

perl -0777 -i -pe'
   BEGIN {
      $key = shift(@ARGV);
      $val = shift(@ARGV);
   }
   s/\A.*^\s*\Q$key\E\s\K[^\n]*/$val/sm;
' "$1" "$2" "$file"

If memory is an issue, reverse the input using File::ReadBackwards (or a similarly efficient tool), change the first match, then reverse the output using File::ReadBackwards.

---

These solutions also fix the improper interpolation of $key_to_modify and $new_value into the Perl program (by passing the values as args).

These solutions also fix the improper interpolation of $key_to_modify into the regex (by using \\Q ).

Answer 2

I suggest you use Tie::File , which lets you access a file as an array of lines. Any modifications made to the array are reflected in the file. It has been a core module since version 8 of Perl 5, so it shouldn't need to be installed.

This one-liner works by checking each line of the file from the end to the beginning, and stopping as soon as a match is found. It looks okay, but I'm not in a position to test it at present.

perl -MTie::File -e"tie @f,'Tie::File',\"${file}\"; s/^\s*${key_to_modify}\s\K.*/${new_value}/ and last for reverse @f"