How to protect Redis calls with eval ?

leoboy has asked for the wisdom of the Perl Monks concerning the following question:

Hello,

I'm developping a tiny app with Redis (some kind of noSQL database) as storage back-end. Redis.pm (http://search.cpan.org/~melo/Redis-1.904/lib/Redis.pm) is fairly cool, however, I would like to protect Redis methods with eval{}, to be able to handle situations on which the Redis server disappears, and so on.

My first idea was to create a new class, let's say SRedis, that inherits from Redis :

package SRedis ;
our @ISA = qw(Redis) ;
[download]

As I'm a lazy boy, I just want to handle all original Redis methods with an AUTOLOAD method on my SRedis package. Code looks like this :


## Wonderful catch-all
## Input  : Nothing
## Output : Asked data 
sub AUTOLOAD {
    # Input variables
    my ($instance_ref) = shift(@_) ;

    # Internal vars
    $tries = 0 ;

    # Determine initially called sub
    my $command = $AUTOLOAD ;
    
    # Format it to call corresponding method on Redis package
    $command =~ s/.*:// ;
    $command = 'SUPER::' . $command ;
    
    TRY:while($tries < $maxtries) {
        # Go on
        $errorcode = eval {@results = $instance_ref->$command(@_)} ;
        
        # Did it went well ?
        if(!defined($errorcode)) {
            # No - I should log something here
            
            # Update count
            $tries++ ;
            
            # Wait a little bit ?
            sleep(1) ;
            
            # Go back to work
            next TRY;
            }
            
        # Otherwise that's cool
        last TRY;            
        }
        
    # Return data
    return (@results) ;
    }
[download]

My problem is that I don't know what kind of data the original command may return. Could be a scalar, an array or a hash. In this piece of code, I assume that the command returns an array.

How can I handle any kind of return in the eval, and make AUTOLOAD return it correctly to the initial caller ?

Thank you in advance for your advices.

Comment on How to protect Redis calls with eval ? Select or Download Code

Replies are listed 'Best First'.
Re: How to protect Redis calls with eval ? by Tux (Canon) on Apr 27, 2011 at 15:39 UTC
Wouldn't it be way more easy to ask the author of Redis to include/implement that feature? That way all people can enjoy the extra features. I had few issues with Redis and they all got fixed relatively soon after I contacted him, so he has proven to be rather responsive. Of course offering hime a patch that implements your suggestion to the original module would probably be even more welcome :) Enjoy, Have FUN! H.Merijn	[reply]
Re^2: How to protect Redis calls with eval ? by leoboy (Initiate) on Apr 28, 2011 at 19:37 UTC
Yeah, you're right, I'll submit a patch to the author to share my work. Actually I've written some extra code that tries to disconnect / reconnect in case of problem, and that works pretty well. Maybe soon in the next version of Redis.pm ;)	[reply]
Re: How to protect Redis calls with eval ? by Anonymous Monk on Apr 27, 2011 at 11:26 UTC
wantarray, Want	[reply]
Re^2: How to protect Redis calls with eval ? by leoboy (Initiate) on Apr 27, 2011 at 14:27 UTC
Worked fine with wantarray(). Thanks for the hint. Here is the final code : ## Wonderful catch-all sub AUTOLOAD { # Input variables my ($instance_ref) = shift(@_) ; # Internal vars $tries = 0 ; my $wantarray = wantarray() ; # Determine initially called sub my $command = $AUTOLOAD ; # Format it to call corresponding method on Redis package $command =~ s/.*:// ; $command = 'SUPER::' . $command ; TRY:while($tries < $maxtries) { # Array $errorcode = eval {@result_array = $instance_ref->$command(@_) +} if(defined($wantarray) && $wantarray) ; # Scalar $errorcode = eval {$result_scalar = $instance_ref->$command(@_ +)} if(defined($wantarray) && !$wantarray) ; # Scalar $errorcode = eval {$instance_ref->$command(@_)} if(!defined($wantarray)) ; # Did it went well ? if(!defined($errorcode) && $@ ne '') { # No - log something ? warn("Could not send command $command to Redis server, try + $tries on $maxtries, reason : $@") ; # Update count $tries++ ; # Wait a little bit sleep(1) ; # Go back to work next TRY; } # Otherwise that's cool last TRY; } # Return data return @result_array if(defined($wantarray) && $wantarray) ; return $result_scalar if(defined($wantarray) && !$wantarray) ; return if(!defined($wantarray)) ; } [download] Thanks for your help.	[reply] [d/l]
Re: How to protect Redis calls with eval ? by shmem (Chancellor) on Apr 27, 2011 at 15:10 UTC
That could work. However, if you say `$errorcode = eval {@results = $instance_ref->$command(@_)} ;` [download] you are evaluating `@results` in scalar context and assigning the result of that evaluation to `$errorcode` - which is not what your want. Just say e.g. `@results = eval { $instance_ref->$command(@_)} ;` [download] and check $@ for errors, which gets set if your eval'ed code dies.	[reply] [d/l] [select]