Say I have a loop in my code that calls the rails debugger a few times
def show
animals = [\'dog\', \'cat\', \'owl\', \'tiger\']
for animal in animals
I came up with another answer to this today that I like even better:
debugger unless @no_debug
Use that on every line that has a debugger stop. When you want to stop stopping just set @no_debug to something.
Just put conditions on the debugger statement so that it stops only when you want it to, e.g.:
debugger if animal == 'tiger'
or if, say, you want to examine the code only on loop 384:
animals.each_with_index do |animal, i|
debugger if i == 384
# do something
end
or put in a variable that will let you continue ad hoc:
continue_debugger = false
animals.each do |animal|
debugger unless continue_debugger
# in the debugger type `p continue_debugger = true` then `c` when done
end
Putting this here as an alternative since this question showed up first in my own searches. Let's say you have a piece of code that isn't working under a specific circumstance, but works otherwise, and you have a whole slew of tests that exercise this but one specific test that fails. It's a PITA to have to continually type continue
into the debug console until you get to the test you really want to debug, so I started using this convention:
In your code:
def some_common_function
debugger if defined? ::ASDF
# do something
end
Then in your test:
should "do this thing that it isn't doing under a specific circumstance" do
# setup the specific situation
::ASDF = true
# your tests
end
put your debugger
statement somewhere before the iteration, then set a breakpoint inside the iteration which you can then clear later.
Example:
def index
debugger
@things = Thing.all
@things.each do |thing|
# ... something you want to check out in the debugger
thing.some_calculation
end
end
When you enter the debugger, set a breakpoint inside:
b app/controllers/things_controller.rb:42
(Where 42 is the line number you want to break at, like thing.some_calculation
above. Note that it has to be an executable line of code -- comments, blank lines won't work). The debugger will show a breakpoint number and location:
Breakpoint 1 at .../app/controllers/things_controller.rb:42
Now, every time you continue, you will stop at the breakpoint. When you're done and want to complete the request, delete the breakpoint:
delete 1
continue once more, and you will complete the request!
It looks like in the source of ruby-debug, the call to debugger
will always stop execution whenever it is hit. So one solution is to do as was suggested by Mori in his 'ad-hoc' solution, to make a conditional around the call to debugger that you can tweak inside the debugger session itself, such that you avoid calling debugger
. This is probably the neatest solution, and what I would do unless you have some strong nagging purity issues with the code involved.
If you really want to only do this without some external conditional and inside the debugger session itself, it is possible. What you have to do is set a breakpoint in the code itself, then you can delete that breakpoint in the debugger when it is triggered:
require 'rubygems'
require 'ruby-debug'
Debugger.start
Debugger.add_breakpoint(__FILE__, __LINE__ + 2)
while true do
puts "Hi"
puts "mom"
end
Debugger.stop
This produces this sort of interaction:
Breakpoint 1 at debug_test.rb:10
debug_test.rb:10
puts "Hi"
(rdb:1) c
Hi
mom
Breakpoint 1 at debug_test.rb:10
debug_test.rb:10
puts "Hi"
(rdb:1) c
Hi
mom
Breakpoint 1 at debug_test.rb:10
debug_test.rb:10
puts "Hi"
(rdb:1) info b
Num Enb What
1 y at ./debug_test.rb:10
breakpoint already hit 3 times
(rdb:1) del 1
(rdb:1) c
Hi
mom
Hi
mom
Hi
mom
...and so on.
In this way, you are setting the breakpoint in code, then deleting it when you are done. Note that any time the line Debugger.add_breakpoint
is called, it will re-set the breakpoint, so that's why it is outside of the loop and pointing 2 lines down. This technique can easily be extracted to require
-ing a script that sets the breakpoint only when loading your server - Heck, you could write a whole framework class around controlling the Debugger module however you want. Of course, if you went this far, I would just create a singleton class that helps you implement Mori's ad-hoc solution and does or does-not call the debugger statement.
I have another answer to this one: set a @debug
on the class that you want to debug. That way you can do:
if (@debug && (the_condition)) then debugger end
or
debugger unless !@debug
then when you are done with the debugger just @debug = false
and c
.
However, I'm not really happy with having debugger 'hard stops' in live code. These are the kind of things that can be accidentally checked in and forgotten about until something breaks. The @debug
would certainly fall under that as well. To that end I think my ideal solution would use Matt's idea and a script that sets up a breakpoint inside the object when the object is created. That way you'd have the debugging that you want but you wouldn't have any code in source control that is specifically for development. I'll update this answer if I find such a solution.