XML::Filter::DocSplitter - Multipass processing of documents

    ## See XML::SAX::???? for an easier way to use this filter.

    use XML::SAX::Machines qw( Machine ) ;

    my $m = Machine(
        [ Intake => "XML::Filter::DocSplitter" => qw( Filter ) ],
        [ Filter => "My::Filter"               => qw( Merger ) ],
        [ Merger => "XML::Filter::Merger" => qw( Output ) ],
        [ Output => \*STDOUT ],
    );

    ## Let the distributor coordinate with the merger
    ## XML::SAX::Manifold does this for you.
    $m->Intake->set_aggregator( $m->Merger );

    $m->parse_file( "foo" );

XML::Filter::DocSplitter is a SAX filter that allows you to apply a filter to repeated sections of a document. It splits a document up at a predefined elements in to multiple documents and the filter is run on each document. The result can be left as a stream of separate documents or combined back in to a single document using a filter like XML::SAX::Merger.

By default, the input document is split in all children of the root element. By that reckoning, this document has three sub-documents in it:

    <doc>
        <subdoc> .... </subdoc>
        <subdoc> .... </subdoc>
        <subdoc> .... </subdoc>
    </doc>

When using without an aggregator, all events up to the first record are lost; with an aggregator, they are passed directly in to the aggregator as the first document. All elements between the records (the \n text nodes, in this case) are also passed directly to the merger (these will arrive between the end_document and start_document calls for each of the records), as are all events from the last record until the end of the input document. This means that the first document, as seen by the merger, is incomplete; it's missing it's end_element, which is passed later.

The approach of passing events from the input document right on through to the merger differs from the way XML::Filter::Distributor works.

This class is derived from XML::SAX::Base, see that for details.

new

    my $d = XML::Filter::DocSplitter->new(
        Handler    => $h,
        Aggregator => $a,    ## optional
    );

set_aggregator

    $h->set_aggregator( $a );

Sets the SAX filter that will stitch the resulting subdocuments back together. Set to CWundef to prevent such stitchery. The aggregator should support the CWstart_manifold_document, CWend_manifold_document, and CWset_include_all_roots methods as described in XML::Filter::Merger.

get_aggregator

    my $a = $h->get_aggregator;

Gets the SAX filter that will stitch the resulting subdocuments back together.

set_split_path

    $h->set_split_path( "/a/b/c" );

Sets the pattern to use when splitting the document. Patterns are a tiny little subset of the XPath language:

    Pattern     Description
    =======     ===========
    /*/*        splits the document on children of the root elt (default)
    //record    splits each <record> elt in to a document
    /*/record   splits each <record> child of the root elt
    /a/b/c/d    splits each of the <d> elts in to a document

get_split_path

    my $a = $h->get_split_path;

Can only feed a single aggregator at the moment :). I can fix this with a bit of effort.

    Barrie Slaymaker <barries@slaysys.com>

    Copyright 2000, Barrie Slaymaker, All Rights Reserved.

You may use this module under the terms of the Artistic, GPL, or the BSD licenses.