Re: [mh] Saving persistent statefor Objectsdefined inLibraryModules?

[Jim D. <ji...@du...>]

Gregg,

Well, this actually works well.

I needed to use "#" as the higher-level quote delimeter.  I think this is okay.

Whevever some object intantiates other objects whose reference is stored
in the $self->{ } hash of the object, this method can be used to safely restore
the objects within the object.

If there was a means for each object to maintain a list of all instantiates
objects, then we could abstract this whole thing in Generic_Item.
Since it's not used much, I'm not sure it's worth the trouble.

For my needs, I find this quite acceptable and not too much of a hack.

# Trim leading and trailing spaces
sub _trim {
    my $string = shift;
    $string =~ s/^\s+//;
    $string =~ s/\s+$//;
    return $string;
}

# We need to save the persistent information for the Process_Item p_sphinx object between restart
# and startup such that we can safely kill any running pocketsphinx processes.  We'll save the
# persistent information in restore_data and eval to play it out and restore the object
# information.
sub restore_string {
    &main::print_log ("PocketSphinx_Listener:: restore_string called") if $main::Debug{pocketsphinx};
    my ($self) = @_;
    my $restore_string = '';
    my $restore_data = &_trim( $self->{p_sphinx}->restore_string( ) );
    $restore_data =~ s/\->/\$self\->{p_sphinx}\->/g;
    $restore_string .= $self->{object_name} . "->{restore_data} = q#$restore_data#;\n";
    $restore_string .= $self->{object_name} . "->restore();\n" if $self->{p_sphinx}->pid( );
    return $restore_string;
}

# restore will be called on startup or restart such that we can play out the information
# contained in p_shpinx_state, which represents the persistent data for the Process_Item
# object.
sub restore {
    my ($self) = @_;
    my $restore_data = $self->{restore_data};
    &main::print_log ("PocketSphinx_Listener:: restore_string: $restore_data") if $main::Debug{pocketsphinx};
    eval $restore_data;
    &main::print_log ("PocketSphinx_Listener:: restore: Error in Persistent data restore: $@\n") if $@;
    $self->{p_sphinx}->stop( );
}

An example of what appears in mh_temp.saved_states is:

if (defined $pocketsphinx_listener) {$pocketsphinx_listener->{restore_data} = q#@{$self->{p_sphinx}->{cmds}} = split($;, 
q~pocketsphinx  -host localhost -port 3235 -log_file /home/misterhouse/data/pocketsphinx/pocketsphinx -sent_file 
/home/misterhouse/data/pocketsphinx/current.sent -lm_file /home/misterhouse/data/pocketsphinx/current.lm -dict_file 
/home/misterhouse/data/pocketsphinx/current.dic -hmm_file /usr/local/share/pocketsphinx/model/hmm/wsj1 -program 
/usr/local/bin/pocketsphinx_continuous -device microphone -sample 16000~);$self->{p_sphinx}->{cmd_index} = q~1~; 
$self->{p_sphinx}->{output} = q~/home/misterhouse/data/pocketsphinx/pocketsphinx_microphone.stdout~; 
$self->{p_sphinx}->{errlog} = q~/home/misterhouse/data/pocketsphinx/pocketsphinx_microphone.stderr~; 
$self->{p_sphinx}->{killsig} = q~TERM~; $self->{p_sphinx}->{started} = q~1205420613~; $self->{p_sphinx}->{pid} = 
q~32504~; $self->{p_sphinx}->restore_active();#; $pocketsphinx_listener->restore(); } else {'var not used'}

Thanks for all you help and suggestions.

Jim

"Gregg Liming" <gr...@li...> wrote in message news:47D...@li......
> Hi Jim,
>
> Jim Duda wrote:
>
>> Is it safe to do this in the contructor?:
>>
>> $self->{object_name} = '$self';
>
> If by "safe"--you mean--throw an error, then yes--it won't immediately
> throw an error.  But, if you mean "ok for use within mh", then no--it
> will cause unexpected problems because '$self' will not resolve to what
> you hope.  See discussion below.
>
>> I've seen {object_name} referenced many times in various places, but
>> never seen where/how it gets loaded.
>
> It's in "the big loop".  See bin/mh and specifically, read_user_code.
> Within read_user_code, all of the user code is scanned and parsed.  As
> part of that parsing, object names (literal strings) are pulled out.
> Then, a call is made to store_object_data (you likely recall that one)
> except that the first argument is a string (the object name) vice an
> object.  And, store_object_data sets $obj->{object_name}.  There's a bit
> of indirection.  Of course, this won't work unless you're in the context
> of the user code parser since it's the only thing that knows the actual
> object names.  That's why separate calls made within constructors to
> store_object_names pass the object vice the object name; other code
> understands that occasionally object refs get returned from the lookup
> tables vice object names.
>
> I hope the above helps (not sure if I would go as far as "makes sense").
>  It's possible that I've missed something useful to you; so, my
> suggestion is to take a look.  It's also possible that my understanding
> of the user code scanning and the equivalent of "auto wire up" is
> slightly flawed.
>
> Gregg
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2008.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> ________________________________________________________
> To unsubscribe from this list, go to: http://sourceforge.net/mail/?group_id=1365
>
>