/ 
XML-Fast-0.04
River stage two • 8 direct dependents • 17 total dependents
/ XML::Fast

NAME

XML::Fast - Simple and very fast XML to hash conversion

SYNOPSIS

use XML::Fast;

my $hash = xml2hash $xml;
my $hash2 = xml2hash $xml, attr => '.', text => '~';

DESCRIPTION

This module implements simple, state machine based, XML parser written in C.

It could parse and recover some kind of broken XML's. If you need XML validator, use XML::LibXML

RATIONALE

Another similar module is XML::Bare. I've used it for some time, but it have some failures:

  • If your XML have node with name 'value', you'll got a segfault

  • If your XML have node with TextNode, then CDATANode, then again TextNode, you'll got broken value

  • It doesn't support charsets

  • It doesn't support any kind of entities.

So, after count of tries to fix XML::Bare I've decided to write parser from scratch.

It is about 40% faster than XML::Bare and about 120% faster, than XML::LibXML

I got this results using the following test on 35kb xml doc:

cmpthese timethese -10, {
    libxml  => sub { XML::LibXML->new->parse_string($doc) },
    xmlfast => sub { XML::Fast::xml2hash($doc) },
    xmlbare => sub { XML::Bare->new(text => $doc)->parse },
};

          Rate  libxml xmlbare xmlfast
libxml  1107/s      --    -38%    -56%
xmlbare 1782/s     61%      --    -28%
xmlfast 2490/s    125%     40%      --

Of course, the results could be defferent for different xml files. With non-utf encodings and with many entities it could be slower. This test was taken for a sample RSS feed in utf-8 mode with a small count of xml entities.

Here is some features and principles:

  • It uses minimal count of memory allocations.

  • All XML is parsed in 1 scan.

  • All values are copied from source XML only once (to destination keys/values)

  • If some types of nodes (for ex comments) are ignored, there are no memory allocations/copy for them.

EXPORT

xml2hash $xml, [ %options ]

OPTIONS

order [ = 0 ]

Not implemented yet. Strictly keep the output order. When enabled, structures become more complex, but xml could be completely reverted.

attr [ = '-' ]

Attribute prefix

<node attr="test" />  =>  { node => { -attr => "test" } }
text [ = '#text' ]

Key name for storing text

When undef, text nodes will be ignored

<node>text<sub /></node>  =>  { node => { sub => '', '#text' => "test" } }
join [ = '' ]

Join separator for text nodes, splitted by subnodes

Ignored when order in effect

# default:
xml2hash( '<item>Test1<sub />Test2</item>' )
: { item => { sub => '', '~' => 'Test1Test2' } };

xml2hash( '<item>Test1<sub />Test2</item>', join => '+' )
: { item => { sub => '', '~' => 'Test1+Test2' } };
trim [ = 1 ]

Trim leading and trailing whitespace from text nodes

cdata [ = undef ]

When defined, CDATA sections will be stored under this key

# cdata = undef
<node><![CDATA[ test ]]></node>  =>  { node => 'test' }

# cdata = '#'
<node><![CDATA[ test ]]></node>  =>  { node => { '#' => 'test' } }
comm [ = undef ]

When defined, comments sections will be stored under this key

When undef, comments will be ignored

# comm = undef
<node><!-- comm --><sub/></node>  =>  { node => { sub => '' } }

# comm = '/'
<node><!-- comm --><sub/></node>  =>  { node => { sub => '', '/' => 'comm' } }
array => 1

Force all nodes to be kept as arrays.

# no array
<node><sub/></node>  =>  { node => { sub => '' } }

# array = 1
<node><sub/></node>  =>  { node => [ { sub => [ '' ] } ] }
array => [ 'node', 'names']

Force nodes with names to be stored as arrays

# no array
<node><sub/></node>  =>  { node => { sub => '' } }

# array => ['sub']
<node><sub/></node>  =>  { node => { sub => [ '' ] } }

SEE ALSO

  • XML::Bare

    Another fast parser, but have problems

  • XML::LibXML

    The most powerful XML parser for perl. If you don't need to parse gigabytes of XML ;)

  • XML::Hash::LX

    XML parser, that uses XML::LibXML for parsing and then constructs hash structure, identical to one, generated by this module. (At least, it should ;)). But of course it is much more slower, than XML::Fast

TODO

  • Ordered mode (as implemented in XML::Hash::LX)

  • Create hash2xml, identical to one in XML::Hash::LX

  • Partial content event-based parsing (I need this for reading XML streams)

Patches, propositions and bug reports are welcome ;)

AUTHOR

Mons Anderson, <mons@cpan.org>

COPYRIGHT AND LICENSE

Copyright (C) 2010 Mons Anderson

This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.