| 00:00.17 | *** join/#storm _mup_ (~mup@li37-90.members.linode.com) |
| 00:28.10 | *** join/#storm niemeyer (~niemeyer@201-35-138-159.pltce701.dsl.brasiltelecom.net.br) |
| 00:28.10 | *** join/#storm marienz (marienz@freenode/staff/marienz) |
| 00:28.10 | *** mode/#storm [+o niemeyer] by pratchett.freenode.net |
| 02:33.39 | *** join/#storm lifeless_ (~robertc@122.58.129.196) |
| 02:40.04 | *** join/#storm lifeless (~robertc@122.58.129.196) |
| 02:48.08 | *** join/#storm lifeless (~robertc@122.58.129.196) |
| 03:01.08 | *** join/#storm lifeless (~robertc@122.58.129.196) |
| 04:39.01 | *** join/#storm niemeyer (~niemeyer@201-35-138-159.pltce701.dsl.brasiltelecom.net.br) |
| 04:39.01 | *** mode/#storm [+o niemeyer] by ChanServ |
| 05:37.59 | *** join/#storm stub (~stub@canonical/launchpad/stub) |
| 07:20.33 | *** join/#storm bigkevmcd_ (~kevin@87-194-156-112.bethere.co.uk) |
| 09:11.36 | *** join/#storm allenap (~gavin@tobermory.gromper.net) |
| 09:11.36 | *** join/#storm allenap (~gavin@canonical/launchpad/allenap) |
| 11:24.27 | *** join/#storm niemeyer (~niemeyer@201-35-138-159.pltce701.dsl.brasiltelecom.net.br) |
| 11:24.28 | *** mode/#storm [+o niemeyer] by ChanServ |
| 16:53.20 | *** join/#storm mgedmin_ (~mg@Maemo/community/contributor/mgedmin) |
| 19:12.18 | *** join/#storm niemeyer_ (~niemeyer@201-35-72-178.pltce701.dsl.brasiltelecom.net.br) |
| 19:12.18 | *** mode/#storm [+o niemeyer_] by ChanServ |
| 20:07.03 | magmatt | marienz: I went with making my own sqlite aggregate function, fwiw. I'm contemplating other, better fixes, but it works for now (and likely for years to come) |
| 20:07.21 | marienz | well, you're trying to add a datatype to sqlite, really. It's not meant for that. |
| 20:07.49 | magmatt | yes, maybe submitting a patch to sqlite is the best solution |
| 20:08.05 | marienz | hrm |
| 20:08.19 | marienz | I don't know the guts of sqlite, so I can't tell if that'd be something you could realistically do |
| 20:10.37 | magmatt | hmm.. a user-defined collation might be better |
| 20:12.18 | marienz | that's a good point, actually |
| 20:12.23 | marienz | I'd for some reason not thought of that |
| 20:12.33 | marienz | that might be sufficient for your needs |
| 20:28.06 | magmatt | darn, the collation is only used on strings. Since Storm stores the decimals as a buffer, the collation isn't invoked |
| 20:46.48 | magmatt | anyone know what code makes Decimals inserted into sqlite turn into buffers -- I'd like for them to unicode instead |
| 20:51.59 | marienz | has another look |
| 20:54.49 | magmatt | when DecimalVariable.parse_get is called with to_db=True, it's definitely unicode, right? |
| 20:58.53 | magmatt | aaaaahhhh... I'm using .17 and looking at .18 source |
| 20:59.23 | magmatt | DecimalVariable.parse_get->str in .17 and ->unicode in .18 |
| 20:59.39 | magmatt | marienz: I apologize -- let's see where using .18 gets me |
| 20:59.47 | magmatt | or even .19 |
| 21:00.39 | marienz | I'm not at all sure if unicode will cause sqlite to do the desirable thing here |
| 21:01.02 | magmatt | marienz: I'll try it |
| 21:02.05 | magmatt | using storm 0.18, the collation is invoked when using max() where it is not in 0.17 |
| 21:02.22 | marienz | IIUC sqlite columns with a type affinity of "numeric" (which yours ends up having) do an awkward thing where it converts to real if it thinks it can do so losslessly |
| 21:02.29 | marienz | that'll give somewhat confusing results when storing decimals |
| 21:02.56 | magmatt | marienz: storing them as text with a custom collation seems to work |
| 21:03.11 | magmatt | this is with 0.18: http://paste.pocoo.org/show/555764/ |
| 21:03.44 | magmatt | and this is with 0.17: http://paste.pocoo.org/show/555765/ |
| 21:11.25 | marienz | weird, I don't see any obviously related changes in storm's changelog |
| 21:51.02 | *** join/#storm kr4d10 (~kr4d10@2001:49f0:a01d::203) |