There are many ways to do this, but the one I usually prefer is to take advantage of XML entities: what I parse is <!DOCTYPE doc [<!ENTITY real_doc SYSTEM "$doc_file">]><doc>&real_doc;</doc>. The XML string references the real file through the entity.

This way you don't need a temporary file, and you don't touch your original data. Plus you get XML cred, amaze your friends, impress your boss... why would you do it any other way?

You can even check that it works by running this code, that tests it both with XML::Parser and with XML::LibXML:

#!/usr/bin/perl

use strict;
use warnings;
 
use XML::Parser;
use XML::LibXML;

my $doc_file= shift @ARGV;

my $xml=qq{<!DOCTYPE doc [<!ENTITY real_doc SYSTEM "$doc_file">]><doc>
+&real_doc;</doc>};

{ print "XML::Parser:\n";
  my $t= XML::Parser->new( Style => 'Stream')->parse( $xml);
}

{ print "XML::LibXML:\n";
  my $parser = XML::LibXML->new();
  my $doc = $parser->parse_string($xml);
  print $doc->toString;
}
[download]

Neat trick!

For XML parsing, I was thinking I could write a new temp file with a starting tag, write the original xml file into the temp file, and then concat the closing line. Parse it, then erase it.

To me that sounds as sane as you can be with such broken files.

Be sure to also include a <?xml ...> preamble with the right encoding information, otherwise non-ASCII chars will surely break your program.

If your XML file is not huge, you could consider reading it in a variable, prepend/append the open/close root tag and then parse it. Most (if not all) XML parsers allow parsing from a string var