=head1 NAME

  Text::DocumentCollection - a collection of documents




=head2 new

The constructor; arguments must be passed as maps
from keys to values. The key C<file> is mandatory.

  my $c = Text::DocumentCollection->new( file => 'coll.db' );

Documents from the collection are saved as in the  specified file,
which is  currently handled by a C<DB_File> hash.


=head2 Add

Add a document to the collection, tagging it with
a unique key.

  $c->Add( $key, $doc );

C<Add> C<die>s if the key is already present.

To change an existing key, use C<Delete> and then C<Add>.

=head2 Delete

Discard a document from the collection.

=head2 NewFromDB

Loads the collection from the given DB file:

  my $c = Text::DocumentCollection->NewFromDB( file => 'coll.db' );

The file must be either empty or created by a former invocation
of C<new> or C<NewFromDB>, followed by any number of C<Add>
and/or C<Delete>.

Currently, all documents in  the  collection are  revived
(by calling C<NewFromString>). This poses performance problems
for huge collections; a caching mechanism would be an option
in this case.

=head2 IDF

Inverse Term frequency of a given term.

The definition we used is, given a term I<t>, a set of documents
I<DOC> and the binary relationship I<has-term>:

  IDF(t) = log2( #DOC / #{ d in DOC | d has-term t } )

The logarithm is in base 2, since this is related to an
information measurement, and # is the cardinality operator.

=head2 EnumerateV

Enumerates all the document in the collection. Called as:

  my @result = $c->EnumerateV( \&Callback, 'the rock' );

The function C<Callback> will be called on each element
of the collection as:

  my @l = CallBack( $c, $key, $doc, $rock );

where C<$rock> is the second argument to C<Callback>.

Since C<$c> is the first argument, the callback may be
an instance method of C<Text::DocumentCollection>.

The final result is obtained by concatenating all the
partial results (C<@l> in the example above).  If you do
not want a result, simply return the empty list ().

There is no particular order of enumeration, so there
is no particular order in which results are concatenated.

=head1 AUTHORS