ruote log_2010-09-07

2010-09-07 01:20:55 utc jmettraux chad fowler live :
2010-09-07 01:25:24 utc fixr jmettraux: hi!! good afternoon John-san
2010-09-07 01:25:30 utc fixr watching it
2010-09-07 01:26:00 utc fixr thanks :)
2010-09-07 01:28:19 utc jmettraux good evening fixr-san
2010-09-07 06:28:33 utc tosch_le hello jmettraux!
2010-09-07 06:28:51 utc jmettraux tosch_le: hello !
2010-09-07 06:29:28 utc tosch_le a question on rk's pagination: shall we handle values for skip which are out of bounds (below 0 and larger than the number of resources)?
2010-09-07 06:30:15 utc jmettraux it should fail gracefully
2010-09-07 06:30:38 utc jmettraux dropping to 0 or to max
2010-09-07 06:31:19 utc tosch_le i wondered wether a redirect to the same action but with a sane param would be appropriate
2010-09-07 06:31:47 utc jmettraux we wouldn't be providing the wrong links in the first place
2010-09-07 06:32:24 utc jmettraux but a redirect sounds nice
2010-09-07 06:32:28 utc tosch_le right. this is just a problem if the user forces an insane parameter
2010-09-07 06:33:15 utc tosch_le at the moment, when giving a skip out of bounds, it yields to interesting results.
2010-09-07 06:33:38 utc jmettraux interesting in which sense ?
2010-09-07 06:34:09 utc tosch_le negative values do work
2010-09-07 06:34:31 utc tosch_le with skip > count, you get all resources
2010-09-07 06:34:44 utc jmettraux aaah
2010-09-07 06:37:34 utc tosch_le negative values are a peace of cake to solve, but the skip > count case is more difficult
2010-09-07 06:37:59 utc tosch_le not exactely difficult, but more work
2010-09-07 06:38:13 utc jmettraux I could fix that at the storage level
2010-09-07 06:38:23 utc tosch_le that would be great
2010-09-07 06:39:28 utc jmettraux could you please add an issue for that at ?
2010-09-07 06:39:53 utc tosch_le yes, of course i can
2010-09-07 06:45:23 utc tosch_le oh, just seeing that i could solve it within rk, as there is a count done before the paginate method is called
2010-09-07 06:45:46 utc jmettraux aaah, OK
2010-09-07 06:47:44 utc jmettraux much appreciated
2010-09-07 07:18:56 utc tosch_le pushed, but without redirection, just using a sane value for now
2010-09-07 07:21:55 utc jmettraux sounds good
2010-09-07 07:24:22 utc jmettraux merged and pushed
2010-09-07 07:25:14 utc jmettraux excellent
2010-09-07 07:25:23 utc tosch_le thanks
2010-09-07 07:26:10 utc jmettraux thanks to you
2010-09-07 07:26:35 utc jmettraux I have to write a page about storage implementations, I really hope Nathan will contribute a MongoDB storage
2010-09-07 07:27:04 utc tosch_le i've seen the message on the ml
2010-09-07 07:27:29 utc tosch_le nice to see there are so much happy ruote users :-)
2010-09-07 07:28:08 utc jmettraux :-)
2010-09-07 07:37:17 utc tosch_le_ what is missing: a documentation for the rest api rk provides. some painstaking effort needed…
2010-09-07 07:38:19 utc jmettraux I was thinking about some slides
2010-09-07 07:38:39 utc jmettraux wanted to release 2.1.18 ruote, rk and all the circus
2010-09-07 07:38:46 utc jmettraux and then emit the slides about rk
2010-09-07 07:39:10 utc tosch_le_ :-)
2010-09-07 07:39:13 utc tosch_le_ sounds good
2010-09-07 07:40:05 utc tosch_le_ gnah. network problems.
2010-09-07 07:41:05 utc tosch_le_ could do with a list of rest-api endpoints and their params for rk-client, though. but i'll simply strive through the code of the resources
2010-09-07 07:41:54 utc jmettraux you're welcome if you want to write something about it
2010-09-07 07:42:21 utc tosch_le_ why did i expect that answer? ;-)
2010-09-07 07:42:32 utc jmettraux you suggested ;-)
2010-09-07 07:42:41 utc jmettraux there is only 1 endpoint : _ruote/
2010-09-07 07:42:50 utc jmettraux well
2010-09-07 07:42:54 utc jmettraux _ruote/processes
2010-09-07 07:43:00 utc jmettraux the rest is "linked"
2010-09-07 07:43:39 utc tosch_le_ i'd see _ruote/workitems as endpoint, too
2010-09-07 07:44:01 utc jmettraux right
2010-09-07 07:44:15 utc jmettraux any will do, they all link to each other
2010-09-07 07:44:37 utc jmettraux please don't tell it's RESTFUL because of endpoints
2010-09-07 07:44:45 utc jmettraux it's restful because of links
2010-09-07 07:45:04 utc tosch_le_ i don't care about restful and stuff
2010-09-07 07:45:24 utc jmettraux I was seriously laughed at initially by REST djihadists, with my old RESTful interface
2010-09-07 07:45:26 utc jmettraux OK
2010-09-07 07:46:06 utc tosch_le_ i just want to have rk-client handle all of rk's features in a neat way
2010-09-07 07:46:20 utc jmettraux aah
2010-09-07 07:46:32 utc jmettraux look at the "rels"
2010-09-07 07:47:01 utc jmettraux if you infer endpoints you become tightly coupled
2010-09-07 07:47:06 utc tosch_le_ they will help, i'm sure
2010-09-07 07:47:37 utc tosch_le_ true.
2010-09-07 07:49:43 utc tosch_le_ but some thinking will be necessary, i suppose. scenario: user has an expression id and wants to update the expression at rk. however, he got rid of the rels. what should rk-client do in that case?
2010-09-07 07:50:01 utc jmettraux tell the user he's an idiot
2010-09-07 07:50:05 utc jmettraux ;-)
2010-09-07 07:50:15 utc tosch_le_ that a bit harsh.
2010-09-07 07:50:38 utc tosch_le_ he doesn't bother about the rels, he just needs the id to identify the resource in question
2010-09-07 07:50:50 utc tosch_le_ so he only saves the id, not the rels.
2010-09-07 07:51:01 utc jmettraux the id should be URL
2010-09-07 07:51:16 utc jmettraux Uniform Resource Locator
2010-09-07 07:51:34 utc jmettraux URI
2010-09-07 07:51:39 utc jmettraux can't remember that one
2010-09-07 07:51:41 utc tosch_le_ that'd help
2010-09-07 07:51:44 utc jmettraux Identifier
2010-09-07 07:52:16 utc jmettraux bookmarking is OK
2010-09-07 07:52:44 utc jmettraux but your "user" is acting through the ruote-kit-client
2010-09-07 07:53:05 utc tosch_le_ nice: "common practices when writing clients for ruote-kit"
2010-09-07 07:53:09 utc jmettraux the rk-client is smart enough to derive an expression resource URI from a string fei
2010-09-07 07:53:51 utc tosch_le_ that's what i said. i build a link to an endpoint ;-)
2010-09-07 07:54:28 utc tosch_le_ s/said/meant
2010-09-07 07:54:48 utc jmettraux yeah, the rk client extracts the wfid, hits the endpoint, follows the #processes link
2010-09-07 07:54:56 utc jmettraux ...
2010-09-07 07:56:02 utc tosch_le_ anyway, i still unsure how to spec rk-client. it's completely independent of rk atm, and that has its virtues, but i'm afraid of divergences between rk and rk-client's specs
2010-09-07 07:56:43 utc jmettraux understood
2010-09-07 07:57:12 utc jmettraux client or best practices, that is the question
2010-09-07 07:58:17 utc jmettraux sorry, was hamletting
2010-09-07 07:58:27 utc tosch_le_ :-D
2010-09-07 07:59:12 utc tosch_le_ the first question is: do we need a rest client?
2010-09-07 07:59:35 utc jmettraux not sure
2010-09-07 07:59:44 utc jmettraux it could be an "example"
2010-09-07 08:00:03 utc jmettraux people tend to hit the storage directly engine(storage)
2010-09-07 08:00:21 utc jmettraux so engine is the client
2010-09-07 08:00:45 utc tosch_le_ that's the reason for me being unsure about the answer, too.
2010-09-07 08:01:24 utc jmettraux maybe people would build a client, but from Python or something else
2010-09-07 08:01:27 utc jmettraux or javascript
2010-09-07 08:01:37 utc tosch_le_ a client makes sense when using a language other than ruby, especially when using remote
2010-09-07 08:02:01 utc tosch_le_ you were faster :-)
2010-09-07 08:02:13 utc jmettraux ;-)
2010-09-07 08:05:11 utc tosch_le_ ok, i suppose it will be best for now to mark rk-client as deprecated for now, just don't have the time to update it. i'll make some thoughts about a general test suite for rk clients, though
2010-09-07 08:06:03 utc tosch_le_ that way, rk integrators using other languages will have an easy way to test their implementations
2010-09-07 08:06:36 utc tosch_le_ let's see how far i get with that
2010-09-07 08:06:50 utc jmettraux OK
2010-09-07 08:07:13 utc jmettraux but, how would you go about testing a RK Python client ?
2010-09-07 08:07:23 utc jmettraux mock RK ?
2010-09-07 08:08:38 utc tosch_le_ i thought of a kind of interactive api documentation
2010-09-07 08:09:15 utc jmettraux OK
2010-09-07 08:09:39 utc jmettraux ruote kit HTML interface should be that interactive api documentation
2010-09-07 08:09:42 utc tosch_le_ but that'll need more thinking in a quiet minute
2010-09-07 08:09:50 utc tosch_le_ true.
2010-09-07 08:09:54 utc jmettraux hence the GET and POST it displays
2010-09-07 08:10:02 utc jmettraux of course it needs improvement
2010-09-07 08:19:40 utc jmettraux feel free to do what you feel passion for
2010-09-07 08:20:01 utc tosch_le_ :-)
2010-09-07 08:23:30 utc jmettraux have to escape, ttyl !
2010-09-07 08:23:36 utc tosch_le_ bye