XML::Comma FAQ - frequently asked questions about XML::Comma


XML::Comma is an information management platform. It was designed to be used as a core tool for developing Very Large Websites. Comma specifies an XML-based document definition format, encourages Perl code to be embedded in these definitions, and specifies an API for manipulating documents and document collections. Comma includes functionality to store sets of documents in ordered, extensible ways, and integrates with a relational database to index, sort and retrieve collections of documents.


How is XML::Comma licensed and distributed?

XML::Comma is Free Software released under the GNU General Public License. For more information about the license, please see

Comma is distributed as a Perl module. The most recent version is always available at

Is there an XML::Comma web site?


Is there an XML::Comma mailing list?

Yes. Please see

What documentation is available?

The XML::Comma User's Guide describes the Comma API in detail. It is available in HTML and PDF. The Guide can be found in the Comma/docs directory of the distribution, or at:


Some indexing.t tests are failing on a new XML::Comma install using MySQL; what's going on?

Comma relies on MySQL's "USE DATA LOCAL" function, which has been disabled by default in MySQL versions 3.23.48 and greater. This change to MySQL breaks a few of the indexing tests. MySQL can be recompiled with the option "--enable-local-infile" to turn back on the LOCAL functionality. Or mysqld can be started with the argument "--local-infile=1". See the following piece of MySQL documentation:

You may also need to pass 'mysql_local_infile=1' as an argument in the data source name you use to connect (via the DBI modules) to MySQL. The dsn is specified in the Comma/ file.

I'm trying to use the SimpleC parser. Why am I getting all of these "BEGIN failed--compilation aborted" error messages when I try to start Apache?

As stated in mod_perl documentation at, there is no file associated with a handler from mod_perl's perspective, so $0 is set to /dev/null.

If a pre-compiled version of the SimpleC parser is not found by Inline, Inline tries (using the FindBin module) to figure out its "current working directory." FindBin croaks when $0 is set to /dev/null, which causes Apache to abort the startup process.

You can get a compiled version of the parser in the right spot by simply invoking

  perl -MXML::Comma -e ''

from a command line. Comma asks Inline to install its compiled files below XML::Comma->tmp_directory.

If you are worried about your pre-compiled parser being clobbered between apache restarts (for example because your tmp_directory is /tmp), you can add a

  BEGIN { $0 = "/path/to/handler" }

block to your handler script, which will prevent FindBin from croaking, and allow Inline's compilation step to proceed normally.


What's the difference between $foo->bar() and $foo->element('bar')->get()?

This is really a question about Comma's "shortcut" syntax, which defines a mapping between method calls and the elements of a given container.

There is a section on the shortcut syntax in the guide that goes into more detail, but here is a quick list of the seven possible ways a shortcut can can be resolved.

  $x->foo ( [@args] )                  # becomes:
  $x->method('foo, @args)              # if there is a method "foo"
  $x->element('foo')                   # singular, nested
  $x->elements('foo')                  # plural, nested
  $x->element('foo')->get()            # singular, non-nested, no @args 
  $x->element('foo')->set(@args)       # singular, non-nested, with @args
  $x->elements_group_get('foo')        # plural, nested, no @args
  $x->elements_group_set('foo', @args) # plural, nested, with @args

How do I tell how many instances of a given plural element exist?

Because the elements() method always tries to return a "list" of elements -- which means that it returns an array in list context and a reference to an array in scalar context -- you have to do a bit of extra work to determine how many instances of a plural element exist. The most concise (and the recommended) way to do this is:

  my $how_many_foos = scalar ( @{$doc->elements('foo')} );

How do I check whether a given element exists, without auto-creating it?

Comma elements automatically get created when you ask for them. So the following code first makes a new Doc, then (behind the scenes) makes a new Element object for you:

  XML::Comma::Doc->new ( type=>'Some_Def' )->element ( "foo" );

This automatic creation (auto-vivication, in Perl lingo) is almost always what you want. But there are cases where you need to check whether an Element already exists before you try to manipulate it. For example, you might have a nested element that has some required children -- and you probably only want to create that element when you're sure you're ready to populate it fully.

The idiom to do this turns out to be almost the same as the idiom to check how many instances of a plural element exist, given above:

  if ( @{$doc->elements('foo')} ) {
    $doc->foo()->child ( 'some value' );

What does the BAD_CONTENT error that talks about '& found that isn't part of an entity reference' mean?

You have a problem that boils down to something along these lines:

  $doc->element('some_element')->set ( "foo & bar" );

Element content must be legal XML -- so no <, >, or & characters are allowed. These special characters must be "escaped" by replacing them with their entity codes (respectively <, >, or &). The Comma::Util::XML_basic_escape() and Comma::Util::XML_basic_unescape() methods are available, as are shortcut flags for the element set() and get() methods:

  $doc->element('some_element()->set ( "foo & bar", escape=>1 );
  $doc->element('some_element')->get ( unescape=>1 );
  $doc->some_element ( "foo & bar", escape=>1 );

Note that there is no way to pass the unescape flag in the shortcut-get syntax (so there are three examples above, rather than four). It is fair to construe this as a problem with the API.

What's wrong with 'while($iterator++){}' ?

There is a bug in Perl (both versions 5.6.1 and 5.8.0) that leads to a memory leak in Comma code like this:

  my $iterator = $index->iterator();
  while ( $iterator++ ) {


  if ( $iterator++ ) {

If you use the 'while($iterator++){}' or 'if($iterator++){}', then your Iterators objects won't ever get garbage-collected. This is very often not a problem; any stand-alone script will be fine, the Iterators will get properly DESTROYed when the script exits. But code like the above running inside, for example, a web application, can be a problem.

This works fine:

  my $iterator = $index->iterator();
  while ( ++$iterator ) {

The pre-increment may seem a little counter-intuitive, but the Iterator class is written to Do The Right Thing for this very common case. And the pre-increment doesn't trigger the memory leak. And this is fine, too:

  my $iterator = $index->iterator();
  while ( $iterator ) {

What does an error that ends 'sh: /tmp/log.comma: Permission denied' mean?

Comma writes a line about all un-caught errors to a log file. The location of the log file is controlled by a setting in -- the default is /tmp/log.comma/. This file probably needs to be writable by any processes that use the Comma framework. In most installations, the file is made world-writable (which should tell you that the Comma log system isn't intended to be used as part of any security auditing or similar framework -- you should write additional code to handle any secure reporting that an application might need.)


Is XML::Comma fast?

Sure. We don't know of any faster way to develop (or to add new features to) large-scale applications that manipulate collections of hundreds of thousands of pieces of messy-but-structured information. We use it every day, and so do many, many people who access the web sites we build.

Oh, wait: you meant, "does it run fast?" Well, that's in the eye of the beholder. Comma's bottleneck is the parsing and object-ifying of XML files. The power and flexibility that the API gives you comes at some cost -- a hand-coded, special-purpose implementation could well be faster for any single usage.

However, we've worked hard to make Comma fast enough to be really, really, useful. For example, Comma's "Inline" parser is about twice as fast as the general-use XML parsers against which we've benchmarked it (because Comma documents aren't allowed to make use of all parts of the XML specification). An experienced designer of large-scale internet systems will easily be able to structure and tune a Comma-based system to serve hundreds of thousands of dynamic pages a day on mid-range x86 boxes.