dgaramond2 has asked for the wisdom of the Perl Monks concerning the following question:

I've used Log::Log4perl for quite a while. It turns out that most of the time the :easy interface is all I'm using. Also, "multiple categories" is something I've wanted and haven't exactly found. Maybe it's time to try something else.

Here's how I prefer to use the logging module: Pick a level using from one of the log statements (debug()/info()/warn()/error()/fatal()), give it zero or more "tags" (or "categories", or "keywords"), followed by the message. That's it. I don't want to have to explicitly create in every package a logger object with a certain category, or do any other unnecessary setup.

Later, in a log file or the configuration section, I can configure which levels, which set of tags, and which package/subroutine pattern, should go to which output (stdout/stderr, file, etc).

I want maximum level of flexibility in filtering log messages in output. Sometimes I want to stay at level warn() globally, but turn on debugging for only one or two subroutines which I'm interested in. Or, just shows debug() messages in a package which include the tag "dump", to let them dump data structures. Or maybe, in an SMTP server code, sometimes I want to temporarily turn on debugging for just a client IP/IP range, in Sundays, and then log all the DATA being sent by the client _if_ said mail body is larger than 20k and contain certain regex pattern. All this without modifying the source code, of course.

Any recommendations?

Replies are listed 'Best First'.
Re: Choosing logging module Log::Log4perl with subcategories
by imp (Priest) on May 25, 2007 at 03:55 UTC
    Using the easy interface is a good way to start with Log::Log4perl, but to get much out of it you need to put a little more work into customizing it.

    The approach I like is to use the class name as the generic logger, and to create subcategories for the certain types of logs in that class. A recent module I wrote used HTML::Parser for event based parsing, and I wanted to select which events were logged. Here is a very stripped down snippet that shows the logging definitions: 

    package SomeParser;
use strict;
use warnings;
use Log::Log4perl;
my $logger     = Log::Log4perl->get_logger('parser');
my $logexport  = Log::Log4perl->get_logger('parser.export');
my $logtext    = Log::Log4perl->get_logger('parser.events.text');
my $logstart   = Log::Log4perl->get_logger('parser.events.start');
my $logend     = Log::Log4perl->get_logger('parser.events.end');

# stuff
sub tag_opened {
    my ($self, $parser, $tag, $attr, $text) = @_;
    $logstart->is_debug && $logstart->debug(sprintf "<%s>",  $tag);
}

sub tag_closed {
    my ($self, $parser, $tag) = @_;
    $logend->is_debug && $logend->debug(sprintf "</%s>",  $tag);
}

1;

    [download]
    I follow similar approaches throughout the code so I can selectively enable categories such as 'parser.events'.

    Another area that I find useful is to have a base class for my test cases (or a library to load), and have the following snippet in it: 

    if ($ENV{TEST_VERBOSE}) {
    Log::Log4perl->easy_init({ level => $DEBUG, layout => '#%5p %F(%L)
+ - %m%n'});
}
else {
    Log::Log4perl->easy_init({ level => $WARN, layout => '#%5p %F(%L) 
+- %m%n'});
}

    [download]
    This will set the log level to debug when the tests are run in verbose mode (e.g. prove -v t/*.t), otherwise it will be set to warn. In both cases it will have a # before the log message to fit in with the TAP output.
[reply]
[d/l]
[select]
Re: Choosing logging module
by blazar (Canon) on May 25, 2007 at 10:50 UTC

    I don't know if this may help you, but Log::Log4perl itself is based on Log::Dispatch. Since your requirements are relatively simple, you may want to build your own logging system on the latter.

[reply]

Back to Seekers of Perl Wisdom