TaggedLogger

Detaches what is logged from how it is logged.

Every time you want to log something, simply write:

logger.debug("verbose debug information") #warn, #info, #error, #fatal also works

and do not worry about what kind of logger you use and how your code accesses it. You may configure these things later, one day writing to STDOUT works for you, another day you'll need something more sophisticated, like several log files each serving different components and for different audience. Here I have described in more details why I created it.

$ gem install tagged_logger

After specifying some logging rules:

TaggedLogger.rules do
  info A, :to => Logger.new("log_for_A_class.log") #1 rule
  error /.*/, :to => Logger.new(ERROR)             #2 rule
  info /.*/, :to => Logger.new(STDOUT)             #3 rule
end

the following will happen:

1. The #logger() method becomes available everywhere, so it is completely safe to have a code like:

    class A
      def foo
         logger.info("Something interesting happened in A#foo") #goes to STDOUT and to 'log_for_A_class.log' file
         logger.debug("I want to see some details.") #goes nowhere
      end
    end
    logger.error("#logger is available everywhere") #goes to STDERR
    class B
       logger.warn("#logger is available everywhere") #goes to STDOUT
    end     
   

2. The A's logger.info() output will show up in two destinations:

   • in STDOUT, as defined by rule #3

   • in 'log_for_A_class.log' file, as defined by rule #1

3. From wherever it gets called from:

    logger.error("ERROR") #will print 'ERROR' in standard error
    logger.info("INFO")  #will print 'INFO' in standard output
    logger.debug("DEBUG") #will not print anything, since there is no 'debug' rule
   

The #logger() returns some object having methods: #debug(), #info(), #warn(), #error() and #fatal(). These methods have generated on first #logger() call and contain only necessary code to satisfy rules conditions. It means, for example, that if no rules defined all these methods do nothing. It is done for performance reasons, I like to log a lot and I do not want calls like #logger.debug() to slow down production code.

The simplest way to have a #logger() available everywhere without specifying any rules is:

TaggedLogger.rules

No rules specified, therefore whenever you call logger.debug() (or alike) you actually paying for just an empty method execution. You may specify rules later, now you may stay focused on code you are writing.

You may define your own formatting:

TaggedLogger.rules do
  format {|level, tag, message| "#{level}-#{tag}: #{msg}"}
end

Each #format() call overrides previous format. If you are wondering what the heck the 'tag' is - the answer is simple. The tag is a class name whose method calls #logger(). This is what allows to specify rules for classes or namespaces and this is what the tagged_logger plugin is named after.

Lets see how you may use it. For example, you want to have separate log files for classes Network and Database:

TaggedLogger.rules do
  debug Network, :to => Logger.new("network.log")
  debug Database, :to => Logger.new("database.log")
end

In case you want to define common log for several classes:

TaggedLogger.rules do
  debug [Ftp, Http, Sockets], :to => Logger.new("network.log")
end 

Or if you want to have all these classes to show up under common tag Network in standard output:

TaggedLogger.rules do
  info /.*/, :to => Logger.new(STDOUT)
  rename [Ftp, Http, Sockets] => :Network
end

You may also use regular expressions in your rules:

TaggedLogger.rules do
  info /Active::/, :to => Logger.new("active.log")
end

There is more general form for rules, it accepts block with three parameters:

TaggedLogger.rules do
  info /Whatever/ do |level, tag, message|
    #do your special logging here
  end      
end

As previously explained the tag is a class name the #logger is being called from (except when you override Rails instrumentation, see below)

$ gem install tagged_logger

In Rails.root/config/application.rb:

   TaggedLogger.config(:replace_existing_logger => true)

Without that original methods ActionController::Base#logger, ActiveRecord::Base#logger and alike will remain untouched, and they have to be patched if we want to redefine their behavior. By default TaggedLogger is safe and it does not patch existing #logger() method.

Rails has two facility for logging - #logger method injected in base classes (ActiveRecord::Base, ActionController::Base, etc.) and instrumentation. Instrumentation in Rails allows to subscribe on event signaled upon block execution, for example:

def sendfile(path, options={})
  ActiveSupport::Notifications.instrument("sendfile.action_controller") do
    #do actual file send
  end
end

The event "sendfile.action_controller" will be signaled after actual work on sending file is done. One could subscribe to that event:

ActiveSupport::Notifications.subscribe("sendfile.action_controller") do |*args|
  event = ActiveSupport::Notifications::Event.new(*args)
  #do something, for example, log an event
end

Logging inside Rails is done by log subscribers - their task to subscribe to instrumentation events, receive and log them. TaggedLogger allows you to override how these subscribers work, for example, you could redirect what is being logged in ActionController to some external hosted log service, like Logbook.me:

# In Rails.root/config/initializers/tagged_logger.rb:
TaggedLogger.rules do
  debug "action_controller.logsubscriber" do |level, tag, msg|
      Logbook.send(level, tag.to_s, :msg => msg.to_s )
  end
end

You may also use active_record.logsubscriber, action_mailer.logsubscriber, action_view.logsubscriber and rack.logsubscriber tags. If you'd like to have a special logging not only for ActionController, but rather for entire logging done in Rails (via instrumentation), you could use a rule with regular expression:

debug /\.logsubscriber$/ do |level, tag, msg|
  #your special logging
end

Rails classes having method #logger() are patched by tagged_logger (only if it is configured with option :replace_existing_logger => true, of cause), so you may define rules for your controllers and models, for example:

class ApplicationController < ActionController::Base
  def welcome
    logger.info "welcome..."
  end
end

# Rails.root/config/initializers/tagged_logger.rb
TaggedLogger.rules do
  debug /.*Controller$/ do |level, tag, msg|
    puts "Here I dump whatever happens in controllers, including ApplicationController"
  end
end

TaggedLogger is released under the MIT license.

The #info(), #debug(), #warn(), #error(), #fatal() rules when having form like :to => logger, the logger has to be an object of standard library Logger class.