Question

I have the following Perl script with is meant to indent a XML file correctly:

@files = glob "*.xml";

undef $/;
for $file (@files) {
    $indent = 0;
    open FILE, $file or die "Couldn't open $file for reading: $!";
    $_ = readline *FILE;
    close FILE or die "Couldn't close $file: $!";

    # Remove whitespace between > and < if that is the only thing separating them
    s/(?<=>)\s+(?=<)//g;

    # Indent
    s{  # Capture a tag <$1$2$3>,
        # a potential closing slash $1
        # the contents $2
        # a potential closing slash $3
        <(/?)([^/>]+)(/?)> 

        # Optional white space
        \s*

        # Optional tag.
        # $4 contains either undef, "<" or "</"
        (?=(</?))?
    }
    {
        # Adjust the indentation level.
        # $3: A <foo/> tag. No alteration to indentation.
        # $1: A closing </foo> tag. Drop one indentation level
        # else: An opening <foo> tag. Increase one indentation level
        $indent +=
            $3 ?  0 :
            $1 ? -1 :
                  1;

        # Put the captured tag back into place
        "<$1$2$3>" .
        # Two closing tags in a row. Add a newline and indent the next line
    ($1 and defined($4) and ($4 eq "</") ? "\n" . ("  " x $indent) : 
$4                                   ? "\n" . ("  " x $indent) :
                                       ""
        )
    # /g repeat as necessary
    # /e Execute the block of perl code to create replacement text
    # /x Allow whitespace and comments in the regex
    }gex;

    open FILE, ">", $file or die "Couldn't open $file for writing: $!";
    print FILE or die "Couldn't write to $file: $!";
    close FILE or die "Couldn't close $file: $!";
}

First, it’s indenting my tabs, and I wanted two whitespaces. Also, it’s producing tags in the same indentation to be on the same line, instead of in the next line, but with the same indent:

<?xml version="1.0" encoding="iso-8859-1"?><!DOCTYPE kit SYSTEM "tc.dtd"><kit><contact/><description>

where it is supposed to be:

  <?xml version="1.0" encoding="iso-8859-1"?>
  <!DOCTYPE kit SYSTEM "tc.dtd">
  <kit>
      <contact/>
      <description>

…

I acknowledge there are Perl tools to indent XML, such as XML-Tidy but due to tc.dtd tag, I always get an error complaining about unsolvable dependencies on the tc.dtd file, while I just care about the indentation of the same (formatting), not the dependencies itself.
What’s wrong with my Perl regex?

Answer 1

You can use the tool xmllint which doesn’t necessarily validate. Example:

Input (badly formatted):

<?xml version="1.0" encoding="iso-8859-1"?> <!DOCTYPE kit SYSTEM "tc.dtd"> <kit>
  <contact/>      <description>chunky bacon</description> </kit>

Run xmllint --format file.xml and you get:

<?xml version="1.0" encoding="iso-8859-1"?>
<!DOCTYPE kit SYSTEM "tc.dtd">
<kit>
  <contact/>
  <description>chunky bacon</description>
</kit>