!/usr/bin/perl -w
 4C3HOH1 - XML::Tidy.pm created by Pip Stuart <Pip@CPAN.Org>
   to tidy XML documents as parsed XML::XPath objects.

NAME

XML::Tidy - tidy indenting of XML documents

VERSION

This documentation refers to version 1.2.54HJnFa of 
XML::Tidy, which was released on Sun Apr 17 19:49:15:36 2005.

SYNOPSIS

  use XML::Tidy;

  # create new   XML::Tidy object from         MainFile.xml
  my $tidy_obj = XML::Tidy->new('filename' => 'MainFile.xml');

  # Tidy up the indenting
     $tidy_obj->tidy();

  # Write out changes back to MainFile.xml
     $tidy_obj->write();

DESCRIPTION

This module creates XML document objects (with inheritance from
L<XML::XPath>) to tidy mixed-content (ie. non-data) text node
indenting.  There are also some other handy member functions to
compress && expand your XML document object (into either a
compact XML representation or a binary one).

2DO

- maybe add support for binary char && short ints && single-
          precision floats in bcompress()

- fix reload() from messing up unicode escaped &XYZ; components like
          Copyright &#xA9; ->  && Registered &#xAE; -> 

-     What else does Tidy need?

USAGE

new()

This is the standard Tidy object constructor.  It can take
the same parameters as an L<XML::XPath> object constructor to
initialize the XML document object.  These can be any one of:

  'filename' => 'SomeFile.xml'
  'xml'      => $variable_which_holds_a_bunch_of_XML_data
  'ioref'    => $file_InputOutput_reference
  'context'  => $existing_node_at_specified_context_to_become_new_obj

reload()

The reload() member function causes the latest data contained in
a Tidy object to be re-parsed which re-indexes all nodes.
This can be necessary after modifications have been made to nodes
which impact the tree node hierarchy because L<XML::XPath>'s find()
member preserves state info which can get out-of-sync.  reload() is
probably rarely useful by itself but it is needed by strip() &&
prune() so it is exposed as a method in case it comes in handy for
other uses.

strip()

The strip() member function searches the Tidy object for all
mixed-content (ie. non-data) text nodes && empties them out.
This will basically unformat any markup indenting.  strip() is
used by compress() && tidy() but it is exposed because it could be
worthwhile by itself.

tidy()

The tidy() member function can take a single optional parameter as
the string that should be inserted for each indent level.  Some
examples:

  # Tidy up indenting with default two  (2) spaces per indent level
     $tidy_obj->tidy();

  # Tidy up indenting with         four (4) spaces per indent level
     $tidy_obj->tidy('    ');

  # Tidy up indenting with         one  (1) tab    per indent level
     $tidy_obj->tidy("\t");

The default behavior is to use two (2) spaces for each indent
level. The Tidy object gets all mixed-content (ie. non-data)
text nodes reformatted to appropriate indent levels according to
tree nesting depth.

NOTE: There seems to be a bug in L<XML::XPath> which does not allow
finding XML processing instructions (PIs) properly so they have been
commented out of tidy().  This means that tidy() unfortunately
removes processing instructions from files it operates on.  I hope
this shortcoming can be repaired in the near future.  tidy() also
disturbs some XML escapes in whatever ways L<XML::XPath> does.  It
has also been brought to my attention that these modules also strip
CDATA tags from XML files / data they operate on.  Even though
CDATA tags don't seem very common, I wish they were easy to support.
Hopefully the vast majority of files will work fine.

compress()

The compress() member function calls strip() on the Tidy object
then creates an encoded comment which contains the names of elements
&& attributes as they occurred in the original document.  Their
respective element && attribute names are replaced with just the
appropriate index throughout the document.

compress() can accept a parameter describing which node types to
attempt to shrink down as abbreviations.  This parameter should be
a string of just the first letters of each node type you wish to
include as in the following mapping:

  e = elements
  a = attribute keys
  v = attribute values *EXPERIMENTAL*
  t = text      nodes  *EXPERIMENTAL*
  c = comment   nodes  *EXPERIMENTAL*
  n = namespace nodes  *not-yet-implemented*

Attribute values ('v') && text nodes ('t') both seem to work fine
with current tokenization.  I've still labeled them EXPERIMENTAL
because they seem more likely to cause problems than valid element
or attribute key names.  I have some bugs in the comment node
compression which I haven't been able to find yet so that one should
be avoided for now.  Since these three node types ('vtc')
all require tokenization, they are not included in default compression
('ea').  An example call which includes values && text would be:

  $tidy_obj->compress('eatv');

The original document structure (ie. node hierarchy) is preserved.
compress() significantly reduces the file size of most XML documents
for when size matters more than immediate human readability.
expand() performs the opposite conversion.

expand()

The expand() member function reads any XML::Tidy::compress comments
from the Tidy object && uses them to reconstruct the document
that was passed to compress().

bcompress('BinaryOutputFilename.xtb')

The bcompress() member function stores a binary representation of
any Tidy object.  The format consists of:

  0) a null-terminated version string
  1) a byte specifying how many bytes later indices will be
  2) the number of bytes from 1 above to designate the total string  count
  3) the number of null-terminated strings from 2 above
  4) the number of bytes from 1 above to designate the total integer count
  5) the number of 4-byte integers         from 4 above
  6) the number of bytes from 1 above to designate the total float   count
  7) the number of double-precision floats from 6 above
  8) node index sets until the end of the file

Normal node index sets consist of two values.  The first is an index
into the three lists (with the number of bytes from 1) as if they were
all linear.  The second is a single-byte integer identifying the node type
(using standard DOM node enumerations).

A few special cases exist in node index sets though.  If the index is
null, it is interpreted as a close-element tag (so no accompanying type
value is read).  On the other end, when the index is non-zero, the type
value is always read.  In the event that the type corresponds to an
attribute or a processing-instruction, the next index is read (without
another accompanying type value) in order to complete the data fields
required by those node types.

bexpand('BinaryInputFilename.xtb')

The bexpand() member function reads a binary file which was
previously written from bcompress().  bexpand() is an XML::Tidy
object constructor like new().

prune()

The prune() member function takes an XPath location to remove (along
with all attributes && child nodes) from the Tidy object.  For
example, to remove all comments:

  $tidy_obj->prune('//comment()');

or to remove the third baz (XPath indexing is 1-based):

  $tidy_obj->prune('/foo/bar/baz[3]');

Pruning your XML tree is a form of tidying too so it snuck in here. =)
It seems L<XML::XPath> objects are dramatically more useful when they
all have access to this class of additional member functions.

write()

The write() member function can take an optional filename parameter
to write out any changes to the Tidy object.  If no parameters
are given, write() overwrites the original XML document file (if
a 'filename' parameter was given to the constructor).

write() will croak() if no filename can be found to write to.

write() can also take a secondary parameter which specifies an XPath
location to be written out as the new root element instead of the
Tidy object's root.  Only the first matching element is written.

toString()

The toString() member function is almost identical to write() except
that it takes no parameters && simply returns the equivalent XML
string as a scalar.  It is a little weird because normally only
XML::XPath::Node objects have a toString member but I figure it makes
sense to extend the same syntax to the parent object as well since
it is a useful option.

createNode Wrappers

The following are just aliases to Node constructors.  They'll work with
just the unique portion of the node type as the member function name.

e() or el() or elem() or createElement()

wrapper for XML::XPath::Node::Element->new()

a() or at() or attr() or createAttribute()

wrapper for XML::XPath::Node::Attribute->new()

c() or cm() or cmnt() or createComment()

wrapper for XML::XPath::Node::Comment->new()

t() or tx() or text() or createTextNode()

wrapper for XML::XPath::Node::Text->new()

p() or pi() or proc() or createProcessingInstruction()

wrapper for XML::XPath::Node::PI->new()

n() or ns() or nspc() or createNamespace()

wrapper for XML::XPath::Node::Namespace->new()

EXPORTED CONSTANTS

XML::Tidy also exports the same node constants as L<XML::XPath::Node>
(which correspond to DOM values).  These include:

  UNKNOWN_NODE
  ELEMENT_NODE
  ATTRIBUTE_NODE
  TEXT_NODE
  CDATA_SECTION_NODE
  ENTITY_REFERENCE_NODE
  ENTITY_NODE
  PROCESSING_INSTRUCTION_NODE
  COMMENT_NODE
  DOCUMENT_NODE
  DOCUMENT_TYPE_NODE
  DOCUMENT_FRAGMENT_NODE
  NOTATION_NODE
  ELEMENT_DECL_NODE
  ATT_DEF_NODE
  XML_DECL_NODE
  ATTLIST_DECL_NODE
  NAMESPACE_NODE

XML::Tidy also exports:

  STANDARD_XML_DECL

which returns a reasonable default XML declaration string.

CHANGES

Revision history for Perl extension XML::Tidy:

- 1.2.54HJnFa  Sun Apr 17 19:49:15:36 2005

* added support for binary ints && floats in bcompress()

* tightened up binary format && added pod

- 1.2.54HDR1G  Sun Apr 17 13:27:01:16 2005

* added bcompress() && bexpand()

* added  compress() &&  expand()

* added toString()

- 1.2.4CKBHxt  Mon Dec 20 11:17:59:55 2004

* added exporting of XML::XPath::Node (DOM) constants

* added node object creation wrappers (like LibXML)

- 1.2.4CCJW4G  Sun Dec 12 19:32:04:16 2004

* added optional 'xpath_loc' => to prune()

- 1.0.4CAJna1  Fri Dec 10 19:49:36:01 2004

* added optional 'filename' => to write()

- 1.0.4CAAf5B  Fri Dec 10 10:41:05:11 2004

* removed 2nd param from tidy() so that 1st param is just indent string

* fixed pod errors

- 1.0.4C9JpoP  Thu Dec  9 19:51:50:25 2004

* added xplc option to write()

* added prune()

- 1.0.4C8K1Ah  Wed Dec  8 20:01:10:43 2004

* inherited from XPath so that those methods can be called directly

* original version (separating Tidy.pm from Merge.pm)

INSTALL

From the command shell, please run:

    `perl -MCPAN -e "install XML::Tidy"`

or uncompress the package && run the standard:

    `perl Makefile.PL; make; make test; make install`

FILES

XML::Tidy requires:

L<Carp>                  to allow errors to croak() from calling sub

L<XML::XPath>            to use XPath statements to query && update XML

L<XML::XPath::XMLParser> to parse XML documents into XPath objects

L<Math::BaseCnv>         to handle base-64 indexing for compress() && expand()

LICENSE

Most source code should be Free!
  Code I have lawful authority over is && shall be!
Copyright: (c) 2004, Pip Stuart.
Copyleft : This software is licensed under the GNU General Public
  License (version 2).  Please consult the Free Software Foundation
  (http://FSF.Org) for important information about your freedom.

AUTHOR

Pip Stuart <Pip@CPAN.Org>

