Question

I’ve been coding in Ruby/Rails for almost 9 months now, having spent years before that in Python.

While I’m really enjoying Rails, there’s one area where I often find myself frustrated: chasing down stubborn bugs. In other languages I can almost always track down difficulties without too much trouble, but when I hit a wall debugging rails, I tend to really hit a wall. I guess what I’m asking is: what strategies do advanced rails users employ to track down more stubborn errors?

At the moment my approach is usually:

1. Examine the stack trace (most simple bugs solved here)

2. Run debugger/pry/console & examine the environment, pace through each step if necessary

3. Google it

4. Post on stack overflow/github issues

5. Procrastinate and/or swear profusely

If any advanced rails-ers would share their strategy for chasing down more stubborn bugs, I’d be really appreciative. In short, what do yo do when trace/debugger don’t offer any clues?

Answer 1

Being with Rails mere 5 years, I do not consider myself an advanced Railser, but nevertheless I’ll happily share my knowledge. 🙂

The main thing when dealing with any (except some very, very trivial ones) is to write a test for this bug.

A few times it happened that I solved the bug at this stage – for example when the bug was related to the automatic class reloading, which is active in development, and turned off in test mode.

Then I usually just place some logger.debug statements with a lot of inspect and caller(0).join("\n\t") in it. Then I very carefully examine the log file.

Because the ‘test.log’ can have a few hundreds megabytes, I always remember to zero it before I run my test. Also I usually run just one test method at the time, because the output would be unreadable otherwise.

I do not use a dedicated debugger. In some old version of ruby the debugger stopped working, I learned to live without it, and never looked back.

---

A few utilities which may be useful:

A function defined in my ~/.bashrc, which lets me to invoke a single test method (or a group of methods):

$ testuj test/unit/user_test.rb -n test_name_validations
$ testuj test/unit/user_test.rb -n /_name_/

function testuj () {
  if [ -n "${BUNDLE_GEMFILE}" ]
  then
    # This is a Rails3 project - it is run by `bundle exec`
    ruby -I"lib:test" "$@"
  else
    # This is a Rails1 project. No bundler.
    ruby -e 'ARGV.each { |f| load f unless f =~ /^-/ ; break if f == "-n" }' "$@"
  fi
}

And this method helps me with logging and checking the timing of some steps:

Object.module_eval do
  def czekpoint(note = nil)
    n = Time.now
    $czekpoint_previous ||= n
    $czekpoint_number ||= 0
    $czekpoint_number += 1
    t = n - $czekpoint_previous
    msg = "CZEKPOINT: %2d  %8.6f  %s %s" % [$czekpoint_number, t, caller.first.to_s.in_yellow, note.to_s.in_red]
    Rails.logger.debug msg # In older Rails it was RAILS_DEFAULT_LOGGER
    STDERR.puts msg
    $czekpoint_previous = n
  end
end