wx27 has asked for the wisdom of the Perl Monks concerning the following question:

I have a similar problem to http://www.perlmonks.org/?node_id=647133 but can't get the \G solution to work. I have data coming in that has a variable number of data pairs, with pipe delimiting. It should be <name>foo</name><value>bar</value>

The current feed comes in as: 

<name>test</name><value>431|alpha|123|bravo|542|charlie|412</value>

[download]
so I am transforming this via: 
perl -pe "do{s/(\<value\>.+?)\|(.+?)\|(.*?)(?=\<\/value\>)/$1\<\/value
+\>\<name\>$2\<\/name\>\<value\>$3/gi;} while /\|/;"

[download]
which gives me the correct output of 
<name>test</name><value>431</value><name>alpha</name><value>123</value
+><name>bravo</name><value>542</value><name>charlie</name><value>412</
+value>

[download]
When I try to recode using Oha's solution with \G, it seems to only process the line once 
perl -pe "s/(\<value\>|\G)(.+?)\|(.+?)\|(.*?)(?=\<\/value\>)/$1$2\<\/v
+alue\>\<name\>$3\<\/name\>\<value\>$4/g"

<name>test</name><value>431</value><name>alpha</name><value>123|bravo|
+542|charlie|412</value>

[download]

What am I doing wrong here?

Replies are listed 'Best First'.
Re: Recursive substitution difficulties
by ikegami (Patriarch) on Mar 15, 2010 at 20:49 UTC
    Your pattern with less leaning toothpicks (removed needless escapes of "<" and ">", and used an alternate delimiter to avoid escaping "/"): 
    s{(<value>|\G)(.+?)\|(.+?)\|(.*?)(?=</value>)}
 {$1$2</value><name>$3</name><value>$4}sg;

    [download]

    What it should match for each pass of /g: 

    Pass 1: <name>(test)</name><value>(431)|
Pass 2: (alpha)|(123)|
Pass 3: (bravo)|(542)|
Pass 4: (charlie)|(412)</value>

    [download]

    What it does match for each pass of /g: 

    Pass 1: <value>(431)|(alpha|123|bravo|542|charlie|412)

    [download]

    Fix: 

    s{(?:<name>(.*?)</name><value>|\G([^|]*?)\|)([^|]*?)(?:\||</value>)}{
   "<name>" . (defined($1)?$1:$2) . "</name><value>$3</value>"
}seg;

    [download]

    A much more robust fix: 

    s{<name>(.*?)</name><value>(.*?)</value>}{
   my $s = "|$1|$2";
   $s =~ s{\|([^|]*)\|([^|]*)}{<name>$1</name><value>$2</value>}sg;
   $s
}seg;

    [download]
[reply]
[d/l]
[select]
Re: Recursive substitution difficulties
by ikegami (Patriarch) on Mar 15, 2010 at 20:27 UTC 
    s{(<name>.*?</name>)<value>(.*?)</value>}{
    $1 . join '', map "<value>$_</value>", split /\|/, $2
}seg;

    [download]

    Better version for 5.10+: 

    s{<name>.*?</name>\K<value>(.*?)</value>}{
    join '', map "<value>$_</value>", split /\|/, $1
}seg;

    [download]

    Update: Oops, didn't pay enough attn to the desired output.

[reply]
[d/l]
[select]
Re: Recursive substitution difficulties
by wx27 (Initiate) on Mar 15, 2010 at 20:52 UTC

    If it helps, the original data is coming in from a tab-delimited file, and I have control over getting that data from 

    [tab]test|431|alpha|123|bravo|542|charlie|412[tab]

    [download]
    to any intermediate step necessary.

    Right now, the field above is being dumped into my input XML file as: 

    <name>test</name><value>431|alpha|123|bravo|542|charlie|412</value>

    [download]

[reply]
[d/l]
[select]
      My earlier solution can easily be adapted: 
      s{\t([^\t]+)\t}{
   my $s = "|$1";
   $s =~ s{\|([^|]*)\|([^|]*)}{<name>$1</name><value>$2</value>}sg;
   $s
}eg;

      [download]
[reply]
[d/l]

        Thanks. That looks pretty straightforward. Is it possible to use that all on a command line invocation? Right now I have something like: 

        perl -pe "s/^(.*?)\t(.*?)\|(.*?)\t(.*?)(?=\n)/<wrapper1>$1</wrapper1><
+wrapper2>$2\|$3</wrapper2><wrapper3>$4</wrapper3>/g;" inputfile

        [download]
        and then a follow-up to expand wrapper2 and the delimited fields within.

[reply]
[d/l]

Back to Seekers of Perl Wisdom