ruote log_2010-10-14

2010-10-14 07:34:27 utc tosch_le__ hi all!
2010-10-14 07:34:43 utc tosch_le__ just contemplating about http://github.com/kennethkalmer/ruote-kit/issues/#issue/5
2010-10-14 07:35:04 utc tosch_le__ i'm unsure if it should be addressed in ruote-kit directly
2010-10-14 07:35:31 utc jmettraux tosch_le__: hello
2010-10-14 07:36:02 utc tosch_le__ there are too many ways to run rk; changing the exception handling in one way or another is bound to be a problem for someoone.
2010-10-14 07:36:29 utc jmettraux ok
2010-10-14 07:37:16 utc jmettraux I have something I have to do with rk, it seems related
2010-10-14 07:37:46 utc tosch_le__ tell me about it :-)
2010-10-14 07:38:00 utc jmettraux /_ruote is vanilla, but I have a customer who wants to deploy to something like /xyz/_ruote
2010-10-14 07:38:13 utc jmettraux argh, it's not related
2010-10-14 07:38:19 utc tosch_le__ rails 3 is great for that case
2010-10-14 07:39:01 utc tosch_le__ the hardcoded routes to /_ruote in rk are some kind of problem, though
2010-10-14 07:39:21 utc jmettraux indeed
2010-10-14 07:39:21 utc jmettraux (mea culpa)
2010-10-14 07:39:51 utc jmettraux back to the original topic, is it possible to shrink that to one match call ?
2010-10-14 07:40:33 utc tosch_le__ no, i tried
2010-10-14 07:41:22 utc jmettraux ok
2010-10-14 07:42:29 utc jmettraux I'd put a link to the issue in your solution's comment
2010-10-14 07:42:48 utc jmettraux I like the fact you commented the issue in the code
2010-10-14 07:43:00 utc jmettraux looks good
2010-10-14 07:43:13 utc tosch_le__ oups. should have put the link in the commit message, of course.
2010-10-14 07:43:56 utc jmettraux great
2010-10-14 07:44:11 utc jmettraux would you also like to work on the /xyz/_ruote/ thing ?
2010-10-14 07:44:55 utc tosch_le__ i tried that one some time ago, we discussed it, too. there's no possible solution within sinatra/rk.
2010-10-14 07:45:11 utc jmettraux orly ?
2010-10-14 07:45:37 utc tosch_le__ pardon?
2010-10-14 07:45:41 utc jmettraux oh really ?
2010-10-14 07:45:50 utc tosch_le__ yes, indeed ;-)
2010-10-14 07:46:16 utc tosch_le__ i suppose there's some rack middleware in the stack needed which runs before rk and rewrites the path
2010-10-14 07:46:26 utc jmettraux ok
2010-10-14 07:46:41 utc tosch_le__ or a customized RuoteKit::Application which does that
2010-10-14 07:47:02 utc tosch_le__ together with rails3's routing, that should be a powerful solution
2010-10-14 07:47:06 utc jmettraux the problem seems localized to our link generation could
2010-10-14 07:47:21 utc jmettraux s/could/code/g
2010-10-14 07:47:25 utc tosch_le__ hello rebo!
2010-10-14 07:47:32 utc jmettraux rebo: hello
2010-10-14 07:47:35 utc rebo hi guys
2010-10-14 07:47:43 utc tosch_le__ we're just discussing about the issue you submitted yesterday
2010-10-14 07:48:03 utc rebo ah ok, do you think its more of a rails / sinatra issue?
2010-10-14 07:48:07 utc rebo thank a ruotekit issue?
2010-10-14 07:48:12 utc rebo than*
2010-10-14 07:48:12 utc tosch_le__ i don't think that should be addressed within ruote-kit itself
2010-10-14 07:48:26 utc rebo yeah i thought that might be the case
2010-10-14 07:48:58 utc tosch_le__ the problem is that routing and exception handling are duplicated: both sinatra (used by rk) and rails provide that
2010-10-14 07:49:19 utc rebo though from a useability point of view its a shame that it blocks the rails exception reporting page
2010-10-14 07:49:26 utc rebo Yeah
2010-10-14 07:49:37 utc tosch_le__ you used the wrong configuration option ;-)
2010-10-14 07:49:51 utc rebo Ah, what should i be using?
2010-10-14 07:50:02 utc tosch_le__ use :raise_errors instead of :show_exceptions
2010-10-14 07:50:08 utc rebo ok
2010-10-14 07:50:15 utc tosch_le__ are you using rails 3 or 2?
2010-10-14 07:50:18 utc rebo 3
2010-10-14 07:50:34 utc tosch_le__ have a look at http://github.com/tosch/ruote-on-rails/commit/758144abcd3c6675e82a198f001e5105e0d7d5e3, then
2010-10-14 07:50:41 utc tosch_le__ far better solution
2010-10-14 07:51:09 utc tosch_le__ (sinatra's routing will only be called for rk's routes)
2010-10-14 07:51:31 utc rebo ah interesting
2010-10-14 07:51:32 utc rebo yes
2010-10-14 07:51:45 utc rebo i was using the middleware option
2010-10-14 07:52:02 utc rebo thanks that looks like it solves it
2010-10-14 07:53:05 utc tosch_le__ it won't do anything for exeptions within the /_ruote namespace, if you like to have rails' exception pages there, you'll have to use RuoteKit::Application.set(:raise_errors, true)
2010-10-14 07:53:32 utc rebo its ok
2010-10-14 07:53:42 utc tosch_le__ but i'd recommend using rails' routing to call rk anyway, routing will be slightly faster
2010-10-14 07:53:57 utc rebo the /_ruote stuff is fine. Just used to seeing the rails exceptions on my rails app code
2010-10-14 07:57:59 utc tosch_le_ ACTION cries: the internet went away
2010-10-14 07:58:11 utc tosch_le_ but happily, it returned.
2010-10-14 07:58:17 utc tosch_le_ btw: in rails 2, rails' exception handling is used anyway
2010-10-14 07:59:31 utc rebo must be due to the changes in rack
2010-10-14 08:00:05 utc tosch_le_ i suppose so, yes.
2010-10-14 08:01:17 utc tosch_le_ so that's the ticket closed :-)
2010-10-14 08:01:28 utc rebo ok cool, thanks again
2010-10-14 08:01:39 utc jmettraux yes, thanks
2010-10-14 09:12:58 utc rebo in ruote participants can reply, but is there a way to rollback to a previous workitem if the reply was accidentally issued ?
2010-10-14 09:22:52 utc rebo or is the standard way to build it into the process definition via a cursor jump / rewind?
2010-10-14 09:23:34 utc tosch_le_ it depends.
2010-10-14 09:23:38 utc tosch_le_ have a look at http://ruote.rubyforge.org/process_administration.html#re_applying
2010-10-14 09:26:16 utc rebo ah looks like i can use something like workitem.command = 'back' ?
2010-10-14 09:26:32 utc rebo within a cursor
2010-10-14 09:27:47 utc tosch_le_ yes, you may do that
2010-10-14 09:28:31 utc tosch_le_ http://ruote.rubyforge.org/exp/cursor.html
2010-10-14 09:28:44 utc rebo yep reading it now
2010-10-14 09:28:53 utc rebo its surprising how powerful ruote is actually
2010-10-14 09:29:06 utc rebo can't believe i've not heard of it before
2010-10-14 09:29:55 utc tosch_le_ this powerfulness comes at a certain price: it's sometimes hard to implement/understand
2010-10-14 09:30:11 utc rebo well yes, thats what i found
2010-10-14 09:30:28 utc rebo very hard to understand it at first
2010-10-14 09:30:35 utc rebo you definitely need some more newbie guides
2010-10-14 09:31:02 utc tosch_le_ they're not easy to write if you're no newbie anymore
2010-10-14 09:31:13 utc rebo yeah haha
2010-10-14 09:31:15 utc tosch_le_ some some help would definitely be very welcome
2010-10-14 09:31:47 utc rebo when i've got this app up and running i can put something together
2010-10-14 09:31:55 utc rebo from a new users point of view
2010-10-14 09:32:26 utc rebo was gonna use a state_machine but it really isnt suited for this business logic
2010-10-14 09:32:57 utc tosch_le_ looking forward to read your report!
2010-10-14 09:34:10 utc rebo i think the main difficulty is that in of itself ruote isn't the whole solution
2010-10-14 09:34:17 utc rebo not that it should be
2010-10-14 09:34:40 utc tosch_le_ neither is aasm
2010-10-14 09:34:58 utc tosch_le_ both are tools
2010-10-14 09:35:08 utc rebo yeah, but for some reason a statemachine is conceptually easier for people to understand
2010-10-14 09:35:33 utc rebo even though it is nothing like a business process in real life
2010-10-14 09:36:10 utc rebo maybe because state machines can be tied to actual things and a workflow is more abstract
2010-10-14 09:37:13 utc rebo a person is in state "X" as apposed to a person is referenced by a workflow process that is in stage "Y"
2010-10-14 09:39:42 utc tosch_le_ jmettraux is better on those things than me: http://jmettraux.wordpress.com/2009/07/03/state-machine-workflow-engine/
2010-10-14 09:45:45 utc tosch_le_ also worth a read: http://www.opensourcery.co.za/2009/07/06/driving-business-processes-in-ruby/
2010-10-14 14:58:35 utc jmettraux codebeaker: welcome to #ruote
2010-10-14 14:59:10 utc codebeaker hi all, I am a typical clueless state machine user… normally I would use a state machine for an appication - but I started reading about workflow engines Vs. state machines, and since the ruote example is about editing and moderation, I think I might be in the right place
2010-10-14 15:00:10 utc jmettraux well, if you only have one resource that has 3 states, maybe a workflow is overkill
2010-10-14 15:00:16 utc codebeaker I have an idea of `recommendations` to improve my site's data - a user can for credit submit a recommendation (think of it like `here's how the data should look`) - the submission then has to be reviewed with someone with moderator or admin privileges, at which point it can be either rejected, or accepted (… and the uses gets credit, etc)
2010-10-14 15:00:40 utc jmettraux will those "workflows" change ?
2010-10-14 15:01:02 utc codebeaker I don't know, and ruote doesn't look like a massive barrier
2010-10-14 15:01:10 utc jmettraux ok
2010-10-14 15:01:14 utc codebeaker there are multiple resources that will have similar behaviour
2010-10-14 15:03:10 utc codebeaker what's not clear from the example is whether or not the workflow can diverge… it's [submitted]->{rejected, accepted}… simple enough workflow - but I would like to learn something about doing it right :)
2010-10-14 15:03:37 utc jmettraux yes, the workflow can diverge
2010-10-14 15:04:15 utc jmettraux which example ?
2010-10-14 15:04:26 utc codebeaker ruote.rubyforge, the first one
2010-10-14 15:04:33 utc jmettraux ok, looking
2010-10-14 15:05:12 utc codebeaker it almost seems to my naieve understanding that the workflow is submission->review and in parallel there is a state to the object - of "awatiting review" then "rejected" or "accepted"
2010-10-14 15:05:31 utc jmettraux it's not naive, it's what is happening
2010-10-14 15:05:48 utc jmettraux a workflow can change the state of 1+ object
2010-10-14 15:06:28 utc codebeaker ah… that's interesting
2010-10-14 15:06:42 utc codebeaker so my workflow could encompass applying the recommendation to an object
2010-10-14 15:06:57 utc codebeaker and notifying the user, and updating their reputation
2010-10-14 15:07:03 utc jmettraux yes
2010-10-14 15:07:05 utc codebeaker which is suddenly four systems involved
2010-10-14 15:07:32 utc jmettraux you can (you already did) do it without a workflow engine
2010-10-14 15:07:55 utc jmettraux but the workflow engine forces you to come up with a recipe on how to reach a result
2010-10-14 15:08:07 utc codebeaker well, right now I have nothing - a bunch of activerecord objects, and what I fear will be a hell of complex interactions :-D
2010-10-14 15:08:57 utc tosch_le_ you really should answer the question if your workflow(s) will change
2010-10-14 15:08:59 utc codebeaker I didn't tackle how to implement "applying" a recommendation to an object, as I feel that "applying" or "rejecting" a recommendation should be transactional, and reliable
2010-10-14 15:09:38 utc codebeaker tosch_le_: right now, I don't know - the paradigm of user submitted content, being reviewed and applied is unlikely to change, but the mechanics of who can submit what, or whether an object is open for recommendations might change
2010-10-14 15:09:54 utc codebeaker but… that suggests that the things which could change aren't part of the workflow
2010-10-14 15:09:54 utc jmettraux an applied recommendation is a record in your db ?
2010-10-14 15:09:59 utc codebeaker yes
2010-10-14 15:10:08 utc codebeaker well… no - sorry read that too quickly
2010-10-14 15:10:28 utc codebeaker a "Recommendation" is– owned by a user, related to the object it suggests to improve
2010-10-14 15:10:38 utc codebeaker right now there's no notion of acepted, rejected or applied
2010-10-14 15:11:31 utc codebeaker but they exist in the database
2010-10-14 15:12:44 utc jmettraux codebeaker: in the first page example, the state of the "document" is only touched at the "publish" stage
2010-10-14 15:13:10 utc jmettraux the rest of the flow is a workitem passing among human participants
2010-10-14 15:14:02 utc codebeaker I wonder if I'm having a bad day or something :-D that example is hurting my brain…
2010-10-14 15:14:24 utc tosch_le_ nah, you have to get used to the terminology
2010-10-14 15:14:31 utc jmettraux what is hard to get ?
2010-10-14 15:14:56 utc jmettraux the second part reads like a bunch of callbacks
2010-10-14 15:15:06 utc jmettraux a participants "reacts" on incoming workitems
2010-10-14 15:15:44 utc jmettraux the name of the participants appear in the process definition and they are registered in the second part
2010-10-14 15:16:07 utc jmettraux the first part (the process definition) orchestrates the work flow among participants
2010-10-14 15:16:42 utc codebeaker ok lemme look again
2010-10-14 15:18:44 utc codebeaker ah, ok - I see it now
2010-10-14 15:19:40 utc codebeaker what's the significance of "process instances" - that doesn't appear to make sense in the context… where do those three args passed go ?
2010-10-14 15:20:16 utc jmettraux you have a third part
2010-10-14 15:20:23 utc jmettraux "launching some process instances"
2010-10-14 15:20:31 utc jmettraux it means running the process
2010-10-14 15:20:37 utc jmettraux 2 times in that example
2010-10-14 15:20:44 utc jmettraux each time for a different document
2010-10-14 15:20:53 utc jmettraux so the result is two process instances
2010-10-14 15:21:15 utc tosch_le_ in your case you would launch a process instance each time a user submits a recommendation
2010-10-14 15:21:19 utc jmettraux like in your operating system you can have two running instances of a program
2010-10-14 15:21:29 utc jmettraux +1 for tosch_le
2010-10-14 15:21:55 utc codebeaker ah, ok - all clear
2010-10-14 15:22:07 utc codebeaker and that has to run in a separate process, along side the application ?
2010-10-14 15:23:10 utc jmettraux not necessarily[
2010-10-14 15:23:36 utc jmettraux if you can spare a thread for it, it will happily live inside of your app
2010-10-14 15:23:44 utc codebeaker ah, intersting
2010-10-14 15:24:07 utc codebeaker jmettraux - as the author, what's the typical usecase for this?
2010-10-14 15:24:12 utc codebeaker … am I overcooking this egg this way ?
2010-10-14 15:24:26 utc jmettraux it's a workflow engine
2010-10-14 15:24:53 utc jmettraux if the process never changes, a classical app is OK
2010-10-14 15:25:18 utc jmettraux if you want the process [definition] to be first class and modifiable, a workflow engine might be valuable
2010-10-14 15:25:56 utc codebeaker "never" can also mean "is well tested, and simple enoguh to change once a year if it needs it" - right ?
2010-10-14 15:26:10 utc jmettraux yes
2010-10-14 15:26:16 utc codebeaker ok… thanks jmettraux
2010-10-14 15:26:19 utc codebeaker and tosch_le_ :)
2010-10-14 15:26:25 utc jmettraux it also means
2010-10-14 15:26:26 utc tosch_le_ it was a pleasure
2010-10-14 15:26:32 utc codebeaker I wonder if it's too complex… but I sure would love to get some experience using it
2010-10-14 15:26:39 utc tosch_le_ s/was/is ;-)
2010-10-14 15:26:43 utc jmettraux that you can have lots of different process instances running together in your app
2010-10-14 15:27:12 utc jmettraux take an afternoon to play with it
2010-10-14 15:27:28 utc tosch_le_ pro state machine: probably less overhead
2010-10-14 15:27:31 utc jmettraux or simply come back in a year
2010-10-14 15:27:33 utc codebeaker sure, the process running on the side worries me a bit… it's another external reliability –
2010-10-14 15:27:48 utc tosch_le_ pro workflow engine: process definition re-usable for different objects
2010-10-14 15:27:53 utc codebeaker I have to monitor that thread, or make sure that process is always running… although if it dies, the states are on disk - right ?
2010-10-14 15:28:09 utc tosch_le_ yes, there's persistence available
2010-10-14 15:28:12 utc codebeaker tosch_le_: ok that might be a seller, because i have right now already 5 kinds of recommendations
2010-10-14 15:28:25 utc codebeaker and the workflow is always the same
2010-10-14 15:28:45 utc tosch_le_ and ruote itself is said to be reliable
2010-10-14 15:28:50 utc jmettraux lol
2010-10-14 15:28:54 utc jmettraux strives to
2010-10-14 15:29:24 utc codebeaker ok, so after a dip i confidence there, multiple objects having the same workflow is a good feature…
2010-10-14 15:29:38 utc tosch_le_ beware, ymmv
2010-10-14 15:29:51 utc codebeaker sure – but now it's worth a coupe of hours
2010-10-14 15:30:23 utc codebeaker … I guess it's normal, completely normal to pick up these interations differently - I had the UI in mind that a moderator would see anything he is able to moderate, online - typical webform…
2010-10-14 15:31:11 utc jmettraux maybe playing with ruote-kit might help
2010-10-14 15:31:15 utc tosch_le_ that ui may be built easily with ruote – or a state machine
2010-10-14 15:31:23 utc jmettraux it comes with a "worklist"
2010-10-14 15:31:35 utc codebeaker ah, awesome - that sounds exactly like what I had in mind
2010-10-14 15:31:50 utc tosch_le_ if you're using rails, have a look at http://github.com/tosch/ruote-on-rails (there's a branch for rails 3)
2010-10-14 15:32:22 utc tosch_le_ it shows how to integrate ruote-kit (and therefore ruote) easily into a rails app
2010-10-14 15:32:50 utc codebeaker tosch_le_: das wird genau perfekt
2010-10-14 15:33:44 utc codebeaker so, thank you both - I think it's time to get outta the office now :)
2010-10-14 15:33:48 utc tosch_le_ have fun! i have to leave now. and it would be great if you could drop in after trying and report back
2010-10-14 15:34:03 utc jmettraux escaping as well, gute Nacht !
2010-10-14 15:34:05 utc codebeaker sure tosch_le_ : it's Padrino->Rails 3 port-my-app-day
2010-10-14 15:34:16 utc codebeaker gute Nacht jungs, bis morgen hoffentlich
2010-10-14 15:34:19 utc tosch_le_ gives us a chance to improve docs and apps
2010-10-14 15:34:25 utc tosch_le_ ciao!
2010-10-14 21:57:05 utc rebo is wi.fields['params']['form_not_ok'] = "true" the right way to modify the payload of a workitem before replying ?
2010-10-14 22:09:45 utc rebo ah, need it without the params