Hook into Exception Chain

Hi there,

I’m using Log4r in my rails projects. On log.error an email is sent using the EmailOutputter.

I know changed the EmailOutputter to include a global var in the subject (MDC) since the subject is normally static.

I want to set this var to the exception message, so it is sent as the subject.

Can anyone tell me how to do this in rails?

Basically:

raise e or some other cause
-> Log4r::MDC.put(‘subject’, e.message)
continue with the exception

I found rescue_from but it just gives it to the handler and then stops.

Thanks a lot,

Christoph

Hi Christoph,

All you need to do is re-raise the exception after you’re done using it in your rescue_from, so something along the lines of:

$ cat ./app/controllers/foo_controller.rb

rescue_from Exception do |e|

do something with e before re-raising it …

Rails.logger.debug(“TEST: before re-raising … e=#{e.inspect}”)

raise e

end

def testfoo

x = ‘bar’ if 1/0

end

$ curl -sLi http://foo.localhost/testfoo

$ cat ./log/development.log

[c55f5] TEST: before re-raising … e=#<ZeroDivisionError: divided by 0>

[c55f5] Completed 500 Internal Server Error in 17ms

[c55f5]

ZeroDivisionError (divided by 0):

app/controllers/foo_controller.rb:89:in `/’

app/controllers/foo_controller.rb:89:in `testfoo’

Jeff

Hi Jeff, thanks for your reply!

I found the article here:

http://www.simonecarletti.com/blog/2009/11/re-raise-a-ruby-exception-in-a-rails-rescue_from-statement/

And he mentions that re-raising does not work since rails won’t catch it anymore…

has this been changed since?

I generally thought there might be a better way, as usually re-raising exceptions is considered to be bad :slight_smile:

Thanks a lot,

Christoph

Hi Christoph,

My (simple) example was just trying to give you an idea of how to do what you were asking using rescue_from.

And note that the log does show that the re-raising worked as expected, where we did something with the caught exception first (ie write a debug TEST line to the log) before then re-raising it and letting the rails env stack handle it (ie log shows the stacktrace including the orig line where the bug is in the code, …).

As for “a better way”, it always depends on what you’re ultimately trying to accomplish and the tradeoffs for getting there.

But if you really did want to use rescue_from for any/all exceptions/errors that could occur, then at a minimum you’re going to want to make sure that what you do with that caught exception (ie your Log4r::MDC… call) doesn’t itself result in any exceptions/errors. One way would be to wrap that work in a begin … rescue Exception => e2 … end … before then re-raising the originally caught exception.

Jeff

Thanks a lot Jeff,

Yeah I saw it works, that’s why I wondered about the linked post…

Anyway, I’m just trying to basically get emails that contain the exception/error message in the subject so it’s easier to sort

if there are many error mails.

Of course the rescue_from method only targets one specific case - the one where I get a rails exception.

But I think I cannot do it by monkey patching log4r as rails already sends a complete (and formatted) string to

log.error making it hard to extract the relevant message.

I’ll see how I can use it, but I think I can get there somehow :slight_smile:

Thank a lot for your suggestion!

Christoph

I’ve used the above approach now

However, this only works for controller actions.

I’ve got some cronjobs in the project (lib folder) as well as a sinatra app that is mounted within the rails app.

Is there any way I could catch the exceptions in these parts as well?

Thanks,

Christoph

For a cronjob type thing it’s up to you to wrap your script with something that will log the error. My cronjobs normally end up looking like

TaskWrapper(‘some job’) do

#work here

end

and elsewhere

def TaskWrapper

begin

yield

rescue Exception => ex

#do something with the exception

raise

end

end

I use rescue Exception because for once I think it’s appropriate that things like SyntaxError are caught so that I can be notified about them

For the sinatra (or even the rails case) you could write a rack middleware that would look something like

class ErrorNotifier

def initialize(app)

@app = app

end

def call(env)

begin

@app.call(env)

rescue Exception => ex

#do something with the exception

raise

end

end

end

You could also render an error message if you want to override that.

You might also be interested in the exception handling stuff described at http://blog.plataformatec.com.br/2012/01/my-five-favorite-hidden-features-in-rails-3-2/

Fred

Thanks a lot! That’s useful