Perl regex capture groups and nth occurence

Admin 1
Viewed 153

I am learning perl regex, and try to combine capture groups and specifying nth occurence of a string.

Say I have the following:

title="alpha" lorem ipsum lorem ipsum name="beta" Morbi posuere metus purus name=delta Curabitur ullamcorper finibus consectetur name=sigma

I want to change the title attribute to the string that follows nth name=, e.g. sigma, while keeping all the content in between. Also, name= may have double quotes such as name="beta" or name=sigma.

1st occurence of name=:

title="beta" lorem ipsum lorem ipsum Morbi posuere metus purus name=delta Curabitur ullamcorper finibus consectetur name=sigma

2nd occurence of name=:

title="sigma" lorem ipsum lorem ipsum name="beta" Morbi posuere metus purus name=delta Curabitur ullamcorper finibus consectetur

I use:

find . -type f -exec perl -pi -w -e 's/(title=)"?[^"\s]*"?(.*) name="?([^"\/]+)"?/$1"$3"$2/' \{\} \;

This works for the first occurence of name=.

I cannot figure how to modify this to specify the nth occurence of name=. I know the basics of specifying nth occurence (such as replace second abc by xyz), ...

s/abc/ ++$count == 2 ? "xyz" : "abc" /eg

... but have trouble integrating this to my code above. How to specify nth name= and move its following capture group in place of title attribute?

perl regex
4 Answers

You can use a pattern to set a manual quantifier in the {n} part and optionally repeat key=value pairs to get to the one you are interested in.

(title=)"?[^\s="]+"?(\h+(?:.*?[^\s=]+=[^\s=]+){0}.*?)[^\s=]+="?([^\s="]+)"?\h*
                                              ^^^

The pattern matches:

  • (title=)"?[^\s="]+"? Capture group 1, match title= and match the value that you don't want to keep after the replacement
  • ( Capture group 2
    • \h+ Match 1+ spaces
    • (?:.*?[^\s=]+=[^\s=]+){0} n times repeat a preceding key=value pair
  • .*? Match any character as least as possible
  • ) Close group 2
  • [^\s=]+= Match 1+ times any char except a whitespace char or =, then match the = for the key part
  • "?([^\s="]+)"? Capture 1+ chars other than a whitespace char = or " in group 3 between optional double quotes
  • \h* Match optional trailing spaces

See a regex demo for 0 repetitions, 1 repetition and 2 repetitions.

Running the pattern in the command for {0} {1} and {2}

find . -type f -exec perl -pi -w -e 's/(title=)"?[^\s="]+"?(\h+(?:.*?[^\s=]+=[^\s=]+){0}.*?)[^\s=]+="?([^\s="]+)"?\h*/$1"$3"$2/' \{\} \;

Changes the line a file to:

title="beta" lorem ipsum lorem ipsum Morbi posuere metus purus name=delta Curabitur ullamcorper finibus consectetur name=sigma

title="delta" lorem ipsum lorem ipsum name="beta" Morbi posuere metus purus Curabitur ullamcorper finibus consectetur name=sigma

title="sigma" lorem ipsum lorem ipsum name="beta" Morbi posuere metus purus name=delta Curabitur ullamcorper finibus consectetur
user_54249880

You may use this perl solution:

# 1st occurrence 
perl -pe 's/(title=)"?[^"\s]*"?((?:.*?\h+name=){0}.*?)\h+name="?([^"\s]+)"?/$1"$3"$2/'

title="beta" lorem ipsum lorem ipsum Morbi posuere metus purus name=delta Curabitur ullamcorper finibus consectetur name=sigma

# 2nd occurrence
perl -pe 's/(title=)"?[^"\s]*"?((?:.*?\h+name=){1}.*?)\h+name="?([^"\s]+)"?/$1"$3"$2/'

title="delta" lorem ipsum lorem ipsum name="beta" Morbi posuere metus purus Curabitur ullamcorper finibus consectetur name=sigma


# 3rd occurrence 
perl -pe 's/(title=)"?[^"\s]*"?((?:.*?\h+name=){2}.*?)\h+name="?([^"\s]+)"?/$1"$3"$2/'

title="sigma" lorem ipsum lorem ipsum name="beta" Morbi posuere metus purus name=delta Curabitur ullamcorper finibus consectetur

Here (?:.*?\h+name=){N} match N occurrences of sub-pattern that is any text followed by 1+ whitespaces followed by text name=. This subpattern is repeated number {N} as shown in examples.

Online Code Demo

user_5482250

Can simplify by doing multiple passes, instead of one conquer-all regex

$N = 1;                          # for the first match
$cnt = 0;                        # silence warnings ($cnt used once)
while (/name="?([^"\s]*)"?/g) { 
    if (++$cnt == $N) {          # get to N-th match 
        $n=$1;                   # store it
        s{name="?$n"?}{};        # remove
        last 
     }   
}; 
s{title=("?\K[^"\s]*)"?}{$n"}    # rewrite title with name

A full example

perl -pwE'        
    $N=shift//1; $cnt = 0;
    while (/name="?([^"\s]*)"?/g) { 
        if (++$cnt == $N) { $n=$1; s{name="?$n"?}{}; last }  
    }; 
    s{title=("?\K[^"\s]*)"?}{$n"}
' file.txt 2

where for testing I use file.txt with the line from the question,

title="alpha" lorem ipsum lorem ipsum name="beta" Morbi posuere metus purus name=delta Curabitur ullamcorper finibus consectetur name=sigma

The command-line input 2 makes it seek the second "name." It prints

title="delta" lorem ipsum lorem ipsum name="beta" Morbi posuere metus purus  Curabitur ullamcorper finibus consectetur name=sigma

The whole thing can be written on one-line if for some reason needed.

This searches for the same pattern twice (in while condition and in its body) so it's inefficient in that sense. It's not too bad though, since the patterns are rather straightforward and this can be optimized if it mattered, while such an inefficiency can be felt only if done a lot or with very complex patterns. But it is two regexes with the same pattern, not pleasing to look at.

The gain is the (comparative) simplicity, where all patterns seek an isolated simple phrase.

user_46533790

Instead of doing everything in one regex, proceed in steps:

perl -lwpe '$n = 2;
            @m=/name="?([^" ]+)"?/g;
            s/title="[^"]+"/title="$m[$n-1]"/;
            s/ name="?\Q$m[$n-1]\E"?//'
  1. Extract all the names into the @m array;
  2. Replace the title by the wanted name;
  3. Remove the name definition.

Note: It's not clear to me why you say sigma is the 2nd name. I'd say it's the 3rd one with delta being the 2nd one.

user_10306750
Related