-x should be C<-x>, reported by Gerben Wierda.

[p5sagit/p5-mst-13.2.git] / pod / perltie.pod

=head1 NAME

perltie - how to hide an object class in a simple variable

=head1 SYNOPSIS

 tie VARIABLE, CLASSNAME, LIST

 $object = tied VARIABLE

 untie VARIABLE

=head1 DESCRIPTION

Prior to release 5.0 of Perl, a programmer could use dbmopen()
to connect an on-disk database in the standard Unix dbm(3x)
format magically to a %HASH in their program.  However, their Perl was either
built with one particular dbm library or another, but not both, and
you couldn't extend this mechanism to other packages or types of variables.

Now you can.

The tie() function binds a variable to a class (package) that will provide
the implementation for access methods for that variable.  Once this magic
has been performed, accessing a tied variable automatically triggers
method calls in the proper class.  The complexity of the class is
hidden behind magic methods calls.  The method names are in ALL CAPS,
which is a convention that Perl uses to indicate that they're called
implicitly rather than explicitly--just like the BEGIN() and END()
functions.

In the tie() call, C<VARIABLE> is the name of the variable to be
enchanted.  C<CLASSNAME> is the name of a class implementing objects of
the correct type.  Any additional arguments in the C<LIST> are passed to
the appropriate constructor method for that class--meaning TIESCALAR(),
TIEARRAY(), TIEHASH(), or TIEHANDLE().  (Typically these are arguments
such as might be passed to the dbminit() function of C.) The object
returned by the "new" method is also returned by the tie() function,
which would be useful if you wanted to access other methods in
C<CLASSNAME>. (You don't actually have to return a reference to a right
"type" (e.g., HASH or C<CLASSNAME>) so long as it's a properly blessed
object.)  You can also retrieve a reference to the underlying object
using the tied() function.

Unlike dbmopen(), the tie() function will not C<use> or C<require> a module
for you--you need to do that explicitly yourself.

=head2 Tying Scalars

A class implementing a tied scalar should define the following methods:
TIESCALAR, FETCH, STORE, and possibly DESTROY.

Let's look at each in turn, using as an example a tie class for
scalars that allows the user to do something like:

    tie $his_speed, 'Nice', getppid();
    tie $my_speed,  'Nice', $$;

And now whenever either of those variables is accessed, its current
system priority is retrieved and returned.  If those variables are set,
then the process's priority is changed!

We'll use Jarkko Hietaniemi <F<jhi@iki.fi>>'s BSD::Resource class (not
included) to access the PRIO_PROCESS, PRIO_MIN, and PRIO_MAX constants
from your system, as well as the getpriority() and setpriority() system
calls.  Here's the preamble of the class.

    package Nice;
    use Carp;
    use BSD::Resource;
    use strict;
    $Nice::DEBUG = 0 unless defined $Nice::DEBUG;

=over

=item TIESCALAR classname, LIST

This is the constructor for the class.  That means it is
expected to return a blessed reference to a new scalar
(probably anonymous) that it's creating.  For example:

    sub TIESCALAR {
        my $class = shift;
        my $pid = shift || $$; # 0 means me

        if ($pid !~ /^\d+$/) {
            carp "Nice::Tie::Scalar got non-numeric pid $pid" if $^W;
            return undef;
        }

        unless (kill 0, $pid) { # EPERM or ERSCH, no doubt
            carp "Nice::Tie::Scalar got bad pid $pid: $!" if $^W;
            return undef;
        }

        return bless \$pid, $class;
    }

This tie class has chosen to return an error rather than raising an
exception if its constructor should fail.  While this is how dbmopen() works,
other classes may well not wish to be so forgiving.  It checks the global
variable C<$^W> to see whether to emit a bit of noise anyway.

=item FETCH this

This method will be triggered every time the tied variable is accessed
(read).  It takes no arguments beyond its self reference, which is the
object representing the scalar we're dealing with.  Because in this case
we're using just a SCALAR ref for the tied scalar object, a simple $$self
allows the method to get at the real value stored there.  In our example
below, that real value is the process ID to which we've tied our variable.

    sub FETCH {
        my $self = shift;
        confess "wrong type" unless ref $self;
        croak "usage error" if @_;
        my $nicety;
        local($!) = 0;
        $nicety = getpriority(PRIO_PROCESS, $$self);
        if ($!) { croak "getpriority failed: $!" }
        return $nicety;
    }

This time we've decided to blow up (raise an exception) if the renice
fails--there's no place for us to return an error otherwise, and it's
probably the right thing to do.

=item STORE this, value

This method will be triggered every time the tied variable is set
(assigned).  Beyond its self reference, it also expects one (and only one)
argument--the new value the user is trying to assign.

    sub STORE {
        my $self = shift;
        confess "wrong type" unless ref $self;
        my $new_nicety = shift;
        croak "usage error" if @_;

        if ($new_nicety < PRIO_MIN) {
            carp sprintf
              "WARNING: priority %d less than minimum system priority %d",
                  $new_nicety, PRIO_MIN if $^W;
            $new_nicety = PRIO_MIN;
        }

        if ($new_nicety > PRIO_MAX) {
            carp sprintf
              "WARNING: priority %d greater than maximum system priority %d",
                  $new_nicety, PRIO_MAX if $^W;
            $new_nicety = PRIO_MAX;
        }

        unless (defined setpriority(PRIO_PROCESS, $$self, $new_nicety)) {
            confess "setpriority failed: $!";
        }
        return $new_nicety;
    }

=item DESTROY this

This method will be triggered when the tied variable needs to be destructed.
As with other object classes, such a method is seldom necessary, because Perl
deallocates its moribund object's memory for you automatically--this isn't
C++, you know.  We'll use a DESTROY method here for debugging purposes only.

    sub DESTROY {
        my $self = shift;
        confess "wrong type" unless ref $self;
        carp "[ Nice::DESTROY pid $$self ]" if $Nice::DEBUG;
    }

=back

That's about all there is to it.  Actually, it's more than all there
is to it, because we've done a few nice things here for the sake
of completeness, robustness, and general aesthetics.  Simpler
TIESCALAR classes are certainly possible.

=head2 Tying Arrays

A class implementing a tied ordinary array should define the following
methods: TIEARRAY, FETCH, STORE, FETCHSIZE, STORESIZE and perhaps DESTROY. 

FETCHSIZE and STORESIZE are used to provide C<$#array> and
equivalent C<scalar(@array)> access.
    
The methods POP, PUSH, SHIFT, UNSHIFT, SPLICE are required if the perl
operator with the corresponding (but lowercase) name is to operate on the
tied array. The B<Tie::Array> class can be used as a base class to implement
these in terms of the basic five methods above.  

In addition EXTEND will be called when perl would have pre-extended 
allocation in a real array.

This means that tied arrays are now I<complete>. The example below needs
upgrading to illustrate this. (The documentation in B<Tie::Array> is more
complete.)

For this discussion, we'll implement an array whose indices are fixed at
its creation.  If you try to access anything beyond those bounds, you'll
take an exception.  For example:

    require Bounded_Array;
    tie @ary, 'Bounded_Array', 2;
    $| = 1;
    for $i (0 .. 10) {
        print "setting index $i: ";
        $ary[$i] = 10 * $i;
        $ary[$i] = 10 * $i;
        print "value of elt $i now $ary[$i]\n";
    }

The preamble code for the class is as follows:

    package Bounded_Array;
    use Carp;
    use strict;

=over

=item TIEARRAY classname, LIST

This is the constructor for the class.  That means it is expected to
return a blessed reference through which the new array (probably an
anonymous ARRAY ref) will be accessed.

In our example, just to show you that you don't I<really> have to return an
ARRAY reference, we'll choose a HASH reference to represent our object.
A HASH works out well as a generic record type: the C<{BOUND}> field will
store the maximum bound allowed, and the C<{ARRAY}> field will hold the
true ARRAY ref.  If someone outside the class tries to dereference the
object returned (doubtless thinking it an ARRAY ref), they'll blow up.
This just goes to show you that you should respect an object's privacy.

    sub TIEARRAY {
        my $class = shift;
        my $bound = shift;
        confess "usage: tie(\@ary, 'Bounded_Array', max_subscript)"
            if @_ || $bound =~ /\D/;
        return bless {
            BOUND => $bound,
            ARRAY => [],
        }, $class;
    }

=item FETCH this, index

This method will be triggered every time an individual element the tied array
is accessed (read).  It takes one argument beyond its self reference: the
index whose value we're trying to fetch.

    sub FETCH {
      my($self,$idx) = @_;
      if ($idx > $self->{BOUND}) {
        confess "Array OOB: $idx > $self->{BOUND}";
      }
      return $self->{ARRAY}[$idx];
    }

As you may have noticed, the name of the FETCH method (et al.) is the same
for all accesses, even though the constructors differ in names (TIESCALAR
vs TIEARRAY).  While in theory you could have the same class servicing
several tied types, in practice this becomes cumbersome, and it's easiest
to keep them at simply one tie type per class.

=item STORE this, index, value

This method will be triggered every time an element in the tied array is set
(written).  It takes two arguments beyond its self reference: the index at
which we're trying to store something and the value we're trying to put
there.  For example:

    sub STORE {
      my($self, $idx, $value) = @_;
      print "[STORE $value at $idx]\n" if _debug;
      if ($idx > $self->{BOUND} ) {
        confess "Array OOB: $idx > $self->{BOUND}";
      }
      return $self->{ARRAY}[$idx] = $value;
    }

=item DESTROY this

This method will be triggered when the tied variable needs to be destructed.
As with the scalar tie class, this is almost never needed in a
language that does its own garbage collection, so this time we'll
just leave it out.

=back

The code we presented at the top of the tied array class accesses many
elements of the array, far more than we've set the bounds to.  Therefore,
it will blow up once they try to access beyond the 2nd element of @ary, as
the following output demonstrates:

    setting index 0: value of elt 0 now 0
    setting index 1: value of elt 1 now 10
    setting index 2: value of elt 2 now 20
    setting index 3: Array OOB: 3 > 2 at Bounded_Array.pm line 39
            Bounded_Array::FETCH called at testba line 12

=head2 Tying Hashes

As the first Perl data type to be tied (see dbmopen()), hashes have the
most complete and useful tie() implementation.  A class implementing a
tied hash should define the following methods: TIEHASH is the constructor.
FETCH and STORE access the key and value pairs.  EXISTS reports whether a
key is present in the hash, and DELETE deletes one.  CLEAR empties the
hash by deleting all the key and value pairs.  FIRSTKEY and NEXTKEY
implement the keys() and each() functions to iterate over all the keys.
And DESTROY is called when the tied variable is garbage collected.

If this seems like a lot, then feel free to inherit from merely the
standard Tie::Hash module for most of your methods, redefining only the
interesting ones.  See L<Tie::Hash> for details.

Remember that Perl distinguishes between a key not existing in the hash,
and the key existing in the hash but having a corresponding value of
C<undef>.  The two possibilities can be tested with the C<exists()> and
C<defined()> functions.

Here's an example of a somewhat interesting tied hash class:  it gives you
a hash representing a particular user's dot files.  You index into the hash
with the name of the file (minus the dot) and you get back that dot file's
contents.  For example:

    use DotFiles;
    tie %dot, 'DotFiles';
    if ( $dot{profile} =~ /MANPATH/ ||
         $dot{login}   =~ /MANPATH/ ||
         $dot{cshrc}   =~ /MANPATH/    )
    {
        print "you seem to set your MANPATH\n";
    }

Or here's another sample of using our tied class:

    tie %him, 'DotFiles', 'daemon';
    foreach $f ( keys %him ) {
        printf "daemon dot file %s is size %d\n",
            $f, length $him{$f};
    }

In our tied hash DotFiles example, we use a regular
hash for the object containing several important
fields, of which only the C<{LIST}> field will be what the
user thinks of as the real hash.

=over 5

=item USER

whose dot files this object represents

=item HOME

where those dot files live

=item CLOBBER

whether we should try to change or remove those dot files

=item LIST

the hash of dot file names and content mappings

=back

Here's the start of F<Dotfiles.pm>:

    package DotFiles;
    use Carp;
    sub whowasi { (caller(1))[3] . '()' }
    my $DEBUG = 0;
    sub debug { $DEBUG = @_ ? shift : 1 }

For our example, we want to be able to emit debugging info to help in tracing
during development.  We keep also one convenience function around
internally to help print out warnings; whowasi() returns the function name
that calls it.

Here are the methods for the DotFiles tied hash.

=over

=item TIEHASH classname, LIST

This is the constructor for the class.  That means it is expected to
return a blessed reference through which the new object (probably but not
necessarily an anonymous hash) will be accessed.

Here's the constructor:

    sub TIEHASH {
        my $self = shift;
        my $user = shift || $>;
        my $dotdir = shift || '';
        croak "usage: @{[&whowasi]} [USER [DOTDIR]]" if @_;
        $user = getpwuid($user) if $user =~ /^\d+$/;
        my $dir = (getpwnam($user))[7]
                || croak "@{[&whowasi]}: no user $user";
        $dir .= "/$dotdir" if $dotdir;

        my $node = {
            USER    => $user,
            HOME    => $dir,
            LIST    => {},
            CLOBBER => 0,
        };

        opendir(DIR, $dir)
                || croak "@{[&whowasi]}: can't opendir $dir: $!";
        foreach $dot ( grep /^\./ && -f "$dir/$_", readdir(DIR)) {
            $dot =~ s/^\.//;
            $node->{LIST}{$dot} = undef;
        }
        closedir DIR;
        return bless $node, $self;
    }

It's probably worth mentioning that if you're going to filetest the
return values out of a readdir, you'd better prepend the directory
in question.  Otherwise, because we didn't chdir() there, it would
have been testing the wrong file.

=item FETCH this, key

This method will be triggered every time an element in the tied hash is
accessed (read).  It takes one argument beyond its self reference: the key
whose value we're trying to fetch.

Here's the fetch for our DotFiles example.

    sub FETCH {
        carp &whowasi if $DEBUG;
        my $self = shift;
        my $dot = shift;
        my $dir = $self->{HOME};
        my $file = "$dir/.$dot";

        unless (exists $self->{LIST}->{$dot} || -f $file) {
            carp "@{[&whowasi]}: no $dot file" if $DEBUG;
            return undef;
        }

        if (defined $self->{LIST}->{$dot}) {
            return $self->{LIST}->{$dot};
        } else {
            return $self->{LIST}->{$dot} = `cat $dir/.$dot`;
        }
    }

It was easy to write by having it call the Unix cat(1) command, but it
would probably be more portable to open the file manually (and somewhat
more efficient).  Of course, because dot files are a Unixy concept, we're
not that concerned.

=item STORE this, key, value

This method will be triggered every time an element in the tied hash is set
(written).  It takes two arguments beyond its self reference: the index at
which we're trying to store something, and the value we're trying to put
there.

Here in our DotFiles example, we'll be careful not to let
them try to overwrite the file unless they've called the clobber()
method on the original object reference returned by tie().

    sub STORE {
        carp &whowasi if $DEBUG;
        my $self = shift;
        my $dot = shift;
        my $value = shift;
        my $file = $self->{HOME} . "/.$dot";
        my $user = $self->{USER};

        croak "@{[&whowasi]}: $file not clobberable"
            unless $self->{CLOBBER};

        open(F, "> $file") || croak "can't open $file: $!";
        print F $value;
        close(F);
    }

If they wanted to clobber something, they might say:

    $ob = tie %daemon_dots, 'daemon';
    $ob->clobber(1);
    $daemon_dots{signature} = "A true daemon\n";

Another way to lay hands on a reference to the underlying object is to
use the tied() function, so they might alternately have set clobber
using:

    tie %daemon_dots, 'daemon';
    tied(%daemon_dots)->clobber(1);

The clobber method is simply:

    sub clobber {
        my $self = shift;
        $self->{CLOBBER} = @_ ? shift : 1;
    }

=item DELETE this, key

This method is triggered when we remove an element from the hash,
typically by using the delete() function.  Again, we'll
be careful to check whether they really want to clobber files.

    sub DELETE   {
        carp &whowasi if $DEBUG;

        my $self = shift;
        my $dot = shift;
        my $file = $self->{HOME} . "/.$dot";
        croak "@{[&whowasi]}: won't remove file $file"
            unless $self->{CLOBBER};
        delete $self->{LIST}->{$dot};
        my $success = unlink($file);
        carp "@{[&whowasi]}: can't unlink $file: $!" unless $success;
        $success;
    }

The value returned by DELETE becomes the return value of the call
to delete().  If you want to emulate the normal behavior of delete(),
you should return whatever FETCH would have returned for this key.
In this example, we have chosen instead to return a value which tells
the caller whether the file was successfully deleted.

=item CLEAR this

This method is triggered when the whole hash is to be cleared, usually by
assigning the empty list to it.

In our example, that would remove all the user's dot files!  It's such a
dangerous thing that they'll have to set CLOBBER to something higher than
1 to make it happen.

    sub CLEAR    {
        carp &whowasi if $DEBUG;
        my $self = shift;
        croak "@{[&whowasi]}: won't remove all dot files for $self->{USER}"
            unless $self->{CLOBBER} > 1;
        my $dot;
        foreach $dot ( keys %{$self->{LIST}}) {
            $self->DELETE($dot);
        }
    }

=item EXISTS this, key

This method is triggered when the user uses the exists() function
on a particular hash.  In our example, we'll look at the C<{LIST}>
hash element for this:

    sub EXISTS   {
        carp &whowasi if $DEBUG;
        my $self = shift;
        my $dot = shift;
        return exists $self->{LIST}->{$dot};
    }

=item FIRSTKEY this

This method will be triggered when the user is going
to iterate through the hash, such as via a keys() or each()
call.

    sub FIRSTKEY {
        carp &whowasi if $DEBUG;
        my $self = shift;
        my $a = keys %{$self->{LIST}};          # reset each() iterator
        each %{$self->{LIST}}
    }

=item NEXTKEY this, lastkey

This method gets triggered during a keys() or each() iteration.  It has a
second argument which is the last key that had been accessed.  This is
useful if you're carrying about ordering or calling the iterator from more
than one sequence, or not really storing things in a hash anywhere.

For our example, we're using a real hash so we'll do just the simple
thing, but we'll have to go through the LIST field indirectly.

    sub NEXTKEY  {
        carp &whowasi if $DEBUG;
        my $self = shift;
        return each %{ $self->{LIST} }
    }

=item DESTROY this

This method is triggered when a tied hash is about to go out of
scope.  You don't really need it unless you're trying to add debugging
or have auxiliary state to clean up.  Here's a very simple function:

    sub DESTROY  {
        carp &whowasi if $DEBUG;
    }

=back

Note that functions such as keys() and values() may return huge lists
when used on large objects, like DBM files.  You may prefer to use the
each() function to iterate over such.  Example:

    # print out history file offsets
    use NDBM_File;
    tie(%HIST, 'NDBM_File', '/usr/lib/news/history', 1, 0);
    while (($key,$val) = each %HIST) {
        print $key, ' = ', unpack('L',$val), "\n";
    }
    untie(%HIST);

=head2 Tying FileHandles

This is partially implemented now.

A class implementing a tied filehandle should define the following
methods: TIEHANDLE, at least one of PRINT, PRINTF, WRITE, READLINE, GETC,
READ, and possibly CLOSE and DESTROY.

It is especially useful when perl is embedded in some other program,
where output to STDOUT and STDERR may have to be redirected in some
special way. See nvi and the Apache module for examples.

In our example we're going to create a shouting handle.

    package Shout;

=over

=item TIEHANDLE classname, LIST

This is the constructor for the class.  That means it is expected to
return a blessed reference of some sort. The reference can be used to
hold some internal information.

    sub TIEHANDLE { print "<shout>\n"; my $i; bless \$i, shift }

=item WRITE this, LIST

This method will be called when the handle is written to via the
C<syswrite> function.

    sub WRITE {
        $r = shift;
        my($buf,$len,$offset) = @_;
        print "WRITE called, \$buf=$buf, \$len=$len, \$offset=$offset";
    }

=item PRINT this, LIST

This method will be triggered every time the tied handle is printed to
with the C<print()> function.
Beyond its self reference it also expects the list that was passed to
the print function.

    sub PRINT { $r = shift; $$r++; print join($,,map(uc($_),@_)),$\ }

=item PRINTF this, LIST

This method will be triggered every time the tied handle is printed to
with the C<printf()> function.
Beyond its self reference it also expects the format and list that was
passed to the printf function.

    sub PRINTF {
        shift;
        my $fmt = shift;
        print sprintf($fmt, @_)."\n";
    }

=item READ this, LIST

This method will be called when the handle is read from via the C<read>
or C<sysread> functions.

    sub READ {
        $r = shift;
        my($buf,$len,$offset) = @_;
        print "READ called, \$buf=$buf, \$len=$len, \$offset=$offset";
    }

=item READLINE this

This method will be called when the handle is read from via <HANDLE>.
The method should return undef when there is no more data.

    sub READLINE { $r = shift; "PRINT called $$r times\n"; }

=item GETC this

This method will be called when the C<getc> function is called.

    sub GETC { print "Don't GETC, Get Perl"; return "a"; }

=item CLOSE this

This method will be called when the handle is closed via the C<close>
function.

    sub CLOSE { print "CLOSE called.\n" }

=item DESTROY this

As with the other types of ties, this method will be called when the
tied handle is about to be destroyed. This is useful for debugging and
possibly cleaning up.

    sub DESTROY { print "</shout>\n" }

=back

Here's how to use our little example:

    tie(*FOO,'Shout');
    print FOO "hello\n";
    $a = 4; $b = 6;
    print FOO $a, " plus ", $b, " equals ", $a + $b, "\n";
    print <FOO>;

=head2 The C<untie> Gotcha

If you intend making use of the object returned from either tie() or
tied(), and if the tie's target class defines a destructor, there is a
subtle gotcha you I<must> guard against.

As setup, consider this (admittedly rather contrived) example of a
tie; all it does is use a file to keep a log of the values assigned to
a scalar.

    package Remember;

    use strict;
    use IO::File;

    sub TIESCALAR {
        my $class = shift;
        my $filename = shift;
        my $handle = new IO::File "> $filename"
                         or die "Cannot open $filename: $!\n";

        print $handle "The Start\n";
        bless {FH => $handle, Value => 0}, $class;
    }

    sub FETCH {
        my $self = shift;
        return $self->{Value};
    }

    sub STORE {
        my $self = shift;
        my $value = shift;
        my $handle = $self->{FH};
        print $handle "$value\n";
        $self->{Value} = $value;
    }

    sub DESTROY {
        my $self = shift;
        my $handle = $self->{FH};
        print $handle "The End\n";
        close $handle;
    }

    1;

Here is an example that makes use of this tie:

    use strict;
    use Remember;

    my $fred;
    tie $fred, 'Remember', 'myfile.txt';
    $fred = 1;
    $fred = 4;
    $fred = 5;
    untie $fred;
    system "cat myfile.txt";

This is the output when it is executed:

    The Start
    1
    4
    5
    The End

So far so good.  Those of you who have been paying attention will have
spotted that the tied object hasn't been used so far.  So lets add an
extra method to the Remember class to allow comments to be included in
the file -- say, something like this:

    sub comment {
        my $self = shift;
        my $text = shift;
        my $handle = $self->{FH};
        print $handle $text, "\n";
    }

And here is the previous example modified to use the C<comment> method
(which requires the tied object):

    use strict;
    use Remember;

    my ($fred, $x);
    $x = tie $fred, 'Remember', 'myfile.txt';
    $fred = 1;
    $fred = 4;
    comment $x "changing...";
    $fred = 5;
    untie $fred;
    system "cat myfile.txt";

When this code is executed there is no output.  Here's why:

When a variable is tied, it is associated with the object which is the
return value of the TIESCALAR, TIEARRAY, or TIEHASH function.  This
object normally has only one reference, namely, the implicit reference
from the tied variable.  When untie() is called, that reference is
destroyed.  Then, as in the first example above, the object's
destructor (DESTROY) is called, which is normal for objects that have
no more valid references; and thus the file is closed.

In the second example, however, we have stored another reference to
the tied object in C<$x>.  That means that when untie() gets called
there will still be a valid reference to the object in existence, so
the destructor is not called at that time, and thus the file is not
closed.  The reason there is no output is because the file buffers
have not been flushed to disk.

Now that you know what the problem is, what can you do to avoid it?
Well, the good old C<-w> flag will spot any instances where you call
untie() and there are still valid references to the tied object.  If
the second script above is run with the C<-w> flag, Perl prints this
warning message:

    untie attempted while 1 inner references still exist

To get the script to work properly and silence the warning make sure
there are no valid references to the tied object I<before> untie() is
called:

    undef $x;
    untie $fred;

=head1 SEE ALSO

See L<DB_File> or L<Config> for some interesting tie() implementations.

=head1 BUGS

Tied arrays are I<incomplete>.  They are also distinctly lacking something
for the C<$#ARRAY> access (which is hard, as it's an lvalue), as well as
the other obvious array functions, like push(), pop(), shift(), unshift(),
and splice().

You cannot easily tie a multilevel data structure (such as a hash of
hashes) to a dbm file.  The first problem is that all but GDBM and
Berkeley DB have size limitations, but beyond that, you also have problems
with how references are to be represented on disk.  One experimental
module that does attempt to address this need partially is the MLDBM
module.  Check your nearest CPAN site as described in L<perlmodlib> for
source code to MLDBM.

=head1 AUTHOR

Tom Christiansen

TIEHANDLE by Sven Verdoolaege <F<skimo@dns.ufsia.ac.be>> and Doug MacEachern <F<dougm@osf.org>>