Perl非贪婪正则表达式

Question

So I finally got my boss to approve the use of perl for this purpose as opposed to sed.

Here's the basic quandry.

I have lines like this:

<div class="SectionText">Sometext</div><div class="SectionText">Some more text</div>

It's terribly messy, but I didn't write it. Either way, there are a goodly number of pages like this and they need to be changed to this format:

<p>Sometext</p><p>Some more text</p>

Obviously this needs to be non-greedy. Now here's the line I've come up with to help with this:

perl -nle "s/(.*)<div class=\"SectionText\">(.*?)<\/div>(.*)/\1<p>\2<\/p>\3/ig; print $1" "somefile.html" > otherfile.html

However, this does nothing and all of the SectionText tags still remain.

Answer 1

Be aware that regular expressions are far from ideal for processing HTML. The proper way is to use a parser and manipulate the DOM, but you can get away with regexes for simple and well-behaved situations. Just be aware further down the line that this is a weak point in your design and may cause unexpected problems.

There is no need to capture and restore text outside the area to be edited. Simply replace the <div> element with a <p> element with the same content. There is also no need to escape double quotes or slashes as long as you choose different delimiters.

It is also wrong to use \\1 , \\2 etc. in the replacement string. $1 , $2 etc. belong here, and you would have been warned of this if you had used -w on the command line.

This should work for you

perl -pe 's|<div class="SectionText">(.*?)</div>|<p>$1</p>|ig' somefile.html > otherfile.html

Answer 2

See HTML::TreeBuilder ::XPath , and HTML::Element for output methods.

my $t = HTML::TreeBuilder::XPath
    ->new_from_content('<div class="SectionText">Sometext</div><div class="SectionText">Some more text</div>');
for ($t->findnodes('//div[@class="SectionText"]')) {
    $_->tag('p');
    $_->attr(class => undef);
}

To make it 100% correct, the class attribute value should be split on white-space, the class name SectionText removed, and then the attribute value reassembled. I think in your case you can get away with just deleting the class attribute as in the code above.