| 00:00.15 | *** join/#storm wallflower (n=wallflow@ip-205-246-113-216.pool.grokthis.net) |
| 00:01.45 | jamesh | niemeyer: thanks for the email re: prejoins |
| 00:01.59 | niemeyer | jamesh: Hey! Good morning! |
| 00:02.14 | niemeyer | jamesh: No problem.. was quite fun to bother kiko to understand how it works :-) |
| 00:03.19 | jamesh | niemeyer: I've been running into some annoying SQLObject compat issues related to Storm delaying INSERTs |
| 00:03.27 | niemeyer | jamesh: (even though his opinion probably varies) |
| 00:03.48 | niemeyer | jamesh: Do you know what the actual issue is? |
| 00:03.56 | niemeyer | jamesh: I mean, how the expectations differ |
| 00:04.13 | jamesh | niemeyer: code that expects foo.id to not be a database ID rather than None |
| 00:04.33 | niemeyer | jamesh: Oh, that's easy to fix |
| 00:04.50 | niemeyer | jamesh: default=AutoReload on the primary key |
| 00:05.12 | jamesh | cool. I'll try that. |
| 00:05.36 | jamesh | hopefully that's all that's required. |
| 00:06.03 | niemeyer | jamesh: We have a few cases like this.. even if the code is aware of how it works, it's more elegant than the explicit flush |
| 00:06.37 | jamesh | niemeyer: I'll give your prejoins code a try today and will report on how it works |
| 00:07.14 | jamesh | I still need to get fog to do a new psycopg2 release ... |
| 00:08.13 | niemeyer | jamesh: I wouldn't be surprised if I got some detail wrong, since the semantics look a bit non obvious to me in a few cases, but I think it's pretty resilient in terms of table usage (I've improved even clauseTables support), and I believe it should be easy to move from this to something similar |
| 00:08.14 | jamesh | niemeyer: I've got some pre-release packages of psycopg2 in my PPA if you want to give it a go, btw. |
| 00:08.44 | niemeyer | jamesh: Oh, you're actually handling the releases as well? |
| 00:08.57 | niemeyer | jamesh: That's very awesome man! |
| 00:09.11 | jamesh | niemeyer: nope. Those packages are built from subversion snapshots |
| 00:09.36 | niemeyer | jamesh: Ahh.. :) |
| 00:09.58 | niemeyer | jamesh: I was getting excited about broad coolness coming |
| 00:10.31 | niemeyer | As in "jamesh takes over psycopg2 maintenance" |
| 00:10.48 | jamesh | niemeyer: I should try valgrinding the Storm test suite with new psycopg2 |
| 00:11.04 | jamesh | should give it a good workout |
| 00:11.21 | jamesh | (hopefully better than psycopg2's test suite) |
| 00:15.10 | jamesh | niemeyer: the default=AutoReload thing works great |
| 00:15.42 | niemeyer | jamesh: It could give a good base indeed.. at least in the most common operations |
| 00:15.45 | niemeyer | jamesh: Sweet! |
| 01:07.03 | jamesh | niemeyer: is there any public API to flush an individual object? |
| 01:07.46 | niemeyer | jamesh: Not yet.. :( |
| 01:08.00 | niemeyer | jamesh: The flush function should probably be extended to take a parameter |
| 01:08.07 | niemeyer | (optionally) |
| 01:08.33 | niemeyer | jamesh: There is _flush_one, but that's not what we want as it considers that dependencies have been handled |
| 01:09.52 | jamesh | niemeyer: I've run into a few cases where Storm gets the update order wrong |
| 01:10.16 | jamesh | when there are unique indexes involved |
| 01:10.33 | jamesh | iirc we previously did manual flushes of individual objects to handle this |
| 01:10.51 | niemeyer | jamesh: That usually happens when the objects are not assigned to each other as references |
| 01:10.57 | niemeyer | jamesh: IOW, Storm doesn't know about the relationship |
| 01:11.25 | jamesh | niemeyer: imagine a unique index on a column 'a' |
| 01:11.26 | niemeyer | jamesh: If the assignment happens by id, Storm doesn't track them as well.. this case we could possibly handle at some point |
| 01:11.34 | niemeyer | jamesh: Ok |
| 01:11.39 | jamesh | I've got two rows: row1.a = 1, row2.a = 2 |
| 01:11.49 | niemeyer | jamesh: Following |
| 01:11.56 | jamesh | and I execute the following commands: row1.a = None; row2.a = 1 |
| 01:12.16 | niemeyer | jamesh: Ohhh, and they're assigned to a sequence? |
| 01:12.26 | jamesh | niemeyer: no sequences |
| 01:12.29 | niemeyer | Hmm |
| 01:12.45 | niemeyer | jamesh: AHhhhhh |
| 01:12.49 | niemeyer | jamesh: Sorry, got it now |
| 01:12.59 | jamesh | niemeyer: the column is set to be unique (or unique given some other constraint) in the DB |
| 01:13.09 | niemeyer | jamesh: Yeah, that sounds a bit tricky indeed |
| 01:13.21 | niemeyer | jamesh: The manual flush would be our best bet for now |
| 01:13.32 | niemeyer | jamesh: Any reason why you don't want to flush everything? |
| 01:14.07 | jamesh | niemeyer: I'll probably do that. Was just wondering if it'd be worth doing a partial flush |
| 01:15.21 | niemeyer | jamesh: Probably not worth the trouble in this case.. Storm will flush changes in the next opportunity in any case |
| 02:56.38 | *** part/#storm marianom (n=marianom@ubuntu/member/marianom) |
| 03:55.40 | *** join/#storm jkakar (n=jkakar@S0106001b11b2d97a.vc.shawcable.net) [NETSPLIT VICTIM] |
| 04:48.37 | *** join/#storm bigdog (n=scmikes@8.8.197.72) |
| 06:38.30 | *** join/#storm jukart (n=jukart@lsfw01.lovelysystems.com) |
| 06:39.53 | *** join/#storm jukart (n=jukart@lsfw01.lovelysystems.com) |
| 07:10.18 | *** join/#storm dobee (n=dobeee@lsfw01.lovelysystems.com) |
| 07:26.39 | *** join/#storm goschtl (n=goschtl@p5B0BCC47.dip.t-dialin.net) |
| 08:28.11 | *** join/#storm mcella (n=michele@ip-171-130.sn3.eutelia.it) |
| 12:30.02 | *** join/#storm Brend (n=brendonh@61-225-160-214.dynamic.hinet.net) |
| 12:30.06 | Brend | Hi! |
| 12:45.42 | *** join/#storm niemeyer (n=niemeyer@200-103-245-169.ctame705.dsl.brasiltelecom.net.br) |
| 13:53.30 | *** join/#storm marianom (n=marianom@ubuntu/member/marianom) |
| 14:46.21 | *** join/#storm bigdog (n=scmikes@72-197-8-8-arpa.cust.cinci.current.net) |
| 16:14.00 | jamesh | niemeyer: sent a reply about the sqlobject-prejoins branch |
| 16:14.11 | jamesh | with a few fixes |
| 16:14.38 | jamesh | and details of another problem I've encountered :( |
| 16:15.35 | jamesh | I need to sleep, so I might catch you tomorrow morning |
| 16:23.03 | niemeyer | jamesh: Cool, I'll try to be online at the time you wake up then |
| 16:23.07 | niemeyer | jamesh: Have a good night! |
| 17:39.31 | *** join/#storm jukart (n=jukart@85-124-221-45.static.xdsl-line.inode.at) |
| 18:52.59 | *** join/#storm dobee (n=dobeee@85-124-200-100.static.xdsl-line.inode.at) |
| 19:10.47 | marianom | hi everyone. is there a way to make "create_database" return an error instead of creating the database? |
| 19:11.08 | marianom | finishing the sentence :) : when the database does not exist |
| 19:11.15 | marianom | ? |
| 19:29.13 | Brend | Wait, when it _doesn't_ exist? |
| 19:43.47 | radix | marianom: no, create_database doesn't actually connect to the database |
| 19:48.37 | *** join/#storm marianom (n=marianom@ubuntu/member/marianom) |
| 20:31.29 | *** join/#storm dobee (n=dobeee@85-124-200-100.static.xdsl-line.inode.at) |
| 20:33.20 | niemeyer | <radix> marianom: no, create_database doesn't actually connect to the database |
| 20:33.32 | niemeyer | marianom: It just creates the database object |
| 20:34.34 | marianom | niemeyer: thanks! (by the way, rrule saved me lots of hours of work, thank you for that too!!) |
| 20:34.59 | niemeyer | marianom: Nice :-) |
| 20:35.16 | niemeyer | marianom: I'm glad it was useful somehow |
| 20:35.43 | marianom | it is one step beyond of everything I needed to do: amazing |
| 20:36.24 | marianom | I'm sure storm will do the same :) |
| 20:57.57 | *** join/#storm Brend (n=brendonh@61-225-165-76.dynamic.hinet.net) |
| 21:33.11 | *** join/#storm jkakar (n=jkakar@S0106000625761564.vc.shawcable.net) |