File file = ...
...
logger.info( "unable to open file: file-name={0}", file );
Doing this has three key advantages. The first is that the ease of adding relevant data to the message is trivial. This then encourages messages that truly aid monitoring and debugging. The second is that there is a very low performance cost to logging in that if the message is never used (due to the log level) then the only cost to making the call is loading and unloading the stack.
The second advantage becomes significant for debugging messages. For example, the debug logger call
logger.debug("{0} moving towards {1}", a, b );
is a far less costly to ignore than is
logger.debug( a + " moving towards " + b );
And especially so in a tight loop. One can argue that you can always check the debug level before making a debug call. In practice, however, this tends to not to be done and instead the useful debug message is never written.
The third advantage is that it is trivial in the logger implementation to both quote and escape the message's parameters. My implementation passes them through a JSON converter so that I can also see the internals of the parameter's values.
