MoinMoin: MoinMoinChat/Logs/moin-dev/2007-05-27

2007-05-27T14:06:54  <zenhase> moin
2007-05-27T14:58:02  <ThomasWaldmann> dreimark: can you reproduce the SystemInfo problem on test.www.de?
2007-05-27T17:43:03  <zenhase> tomorrow is the great day
2007-05-27T17:50:09  <ThomasWaldmann> when the holy google ghost lets you begin coding
2007-05-27T17:53:34  <zenhase> hmm, i dunno about that ... i don't believe in ghosts ;>
2007-05-27T17:55:38  <zenhase> this evening i will delegate gpn6-tasks to some people ... this will hopefully give me enough time to spend several hours of moin-coding per day starting tomorrow
2007-05-27T17:57:16  <zenhase> and on tuesday i will buy a 2.5" hdd, then i can use the old powerbook of a friend ... this will help working on stuff while moving around
2007-05-27T18:09:23  <ThomasWaldmann> btw, i suggest you keep an eye on performance early, not at the end
2007-05-27T18:15:57  <zenhase> i already have some ideas in my mind
2007-05-27T18:16:21  <zenhase> extensive use of genshis template-caching mechanisms
2007-05-27T18:16:45  <zenhase> i will write them down when i start implementing
2007-05-27T18:18:53  <zenhase> i also tried to figure out a way to measure performance of only the layout code
2007-05-27T18:20:09  <zenhase> perhaps write some decorators for the theme-methods that can be attached easily in one line of code and log the results to the errorlog
2007-05-27T18:35:18  <ThomasWaldmann> btw, i wrote some email to the YAML author about licensing topics
2007-05-27T18:35:36  <ThomasWaldmann> (yaml.de)
2007-05-27T18:37:36  <Fabi> moin
2007-05-27T18:37:56  <ThomasWaldmann> moin Fabi
2007-05-27T18:38:21  <zenhase> hi Fabi
2007-05-27T18:38:55  <zenhase> ThomasWaldmann: what is that?
2007-05-27T18:39:35  <ThomasWaldmann> see yaml.de :)
2007-05-27T18:40:11  <zenhase> ThomasWaldmann: well, i see it but i don't really comprehend
2007-05-27T18:40:28  <zenhase> hmm, at first it looked like something commercial
2007-05-27T18:47:08  * ThomasWaldmann searches the yaml developer version
2007-05-27T18:54:06  <ThomasWaldmann> (the one with english comments and cleaned up directory/file structure...)
2007-05-27T19:03:21  <zenhase> ThomasWaldmann: where did you get that? i don't see anything except this one link to the current german version there
2007-05-27T19:07:04  <ThomasWaldmann> I've attended a talk at webinale (given by Jens Grochtdreis) and he showed that. I just wrote a mail to the YAML author asking for some developer download link / repo ...
2007-05-27T19:08:43  <zenhase> ah ok
2007-05-27T20:20:55  <ThomasWaldmann> hmm, looking at some licensing discussions in their forum, it looks like there will be no "special deal"/license for yaml within GPL software.
2007-05-27T20:22:33  <ThomasWaldmann> so yaml based themes would have to backlink to yaml.de or the site owner will have to get a $$($) license.
2007-05-27T20:32:15  <lanius> hola
2007-05-27T20:33:22  <ThomasWaldmann> wb lanius
2007-05-27T20:34:44  <lanius> sorry for not being active last week, had some stuff to do
2007-05-27T20:37:00  <lanius> i added/changed a few bits to the storage design
2007-05-27T20:38:26  <lanius> what is left is the question what is still unclear and needs to be clarified before coding and where to start coding then
2007-05-27T20:39:30  <ThomasWaldmann> btw, you have to be very clear about whether -1 is current or 0 is current (as is now)
2007-05-27T20:39:57  <lanius> i think it's clear that 0 is current, isn't it
2007-05-27T20:42:24  <ThomasWaldmann> please make a spellchecking pass over the page :)
2007-05-27T20:42:37  <xorAxAx> ThomasWaldmann: which would basically mean that you may not distribute yaml-based themes
2007-05-27T20:42:44  <xorAxAx> without having me sueing somebody :)
2007-05-27T20:44:57  <ThomasWaldmann> hmm?
2007-05-27T20:46:21  <ThomasWaldmann> lanius: yeah, i was just reading the diff and thought you did s/0/-1/
2007-05-27T20:46:59  <lanius> ThomasWaldmann: no, i just added -1 as item-wide metadata
2007-05-27T20:51:29  <ThomasWaldmann> if we create TitleIndex with the FS storage, we have to open 2 files per item. I hoped we could reduce that to 1.
2007-05-27T20:51:56  <lanius> Why do we have two open 2 files?
2007-05-27T20:52:28  <ThomasWaldmann> because we need to get the deleted state from item metadata and the acls from current revision
2007-05-27T20:52:51  <lanius> so what do you want to change there
2007-05-27T20:54:08  <ThomasWaldmann> the current fs storage keeps the deleted state at the current revision (by having its data file deleted)
2007-05-27T20:54:36  <lanius> and we have to be compatible with the current design
2007-05-27T20:54:58  <ThomasWaldmann> no
2007-05-27T20:55:25  <ThomasWaldmann> but we have to keep performance in mind :)
2007-05-27T20:55:35  <lanius> sure, the first backend will be a 1.5 compatible backend
2007-05-27T20:56:20  <ThomasWaldmann> well, you can emulate everything you want. 1.5 compatible backend is mainly for migration.
2007-05-27T20:56:33  <lanius> right
2007-05-27T20:56:50  <lanius> but i don't think you can come down to just 1 file access there
2007-05-27T20:57:21  <lanius> and file access does not say much about performance, does it
2007-05-27T20:57:28  <ThomasWaldmann> it could store the deleted state in the current revision metadata
2007-05-27T20:58:28  <ThomasWaldmann> the current 1.5 backend is rather slow and this is because of too much file opening and parsing.
2007-05-27T20:59:02  <lanius> parsing is important, since currently all metadata parsing is done from the same file
2007-05-27T20:59:51  <lanius> so what is left is the question what is still unclear and needs to be clarified before coding and where to start coding then
2007-05-27T20:59:52  <ThomasWaldmann> you have to imagine a page containing 20 links to other pages. the rendering of those links depends on otherpage.exists()
2007-05-27T21:01:47  <ThomasWaldmann> that spellchecking pass was no joke, btw :)
2007-05-27T21:03:47  <lanius> this can be reduced to one file access with filters
2007-05-27T21:04:56  <lanius> the moin spell checker is not very helpfull
2007-05-27T21:06:10  * ThomasWaldmann meant your built-in grey-matter spellchecker :)
2007-05-27T21:07:02  <dreimark> moin
2007-05-27T21:10:09  <ThomasWaldmann> btw, the pagelinks could be stored per item (not per revision), IF we store them to metadata (and not just to a per-item cache)
2007-05-27T21:11:14  <lanius> and what if you wan't the pagelinks from a older revision? but nothing to worry about now
2007-05-27T21:13:02  <ThomasWaldmann> i guess we never want that
2007-05-27T21:15:58  <ThomasWaldmann> hi dreimark
2007-05-27T21:17:09  <dreimark> hi, ThomasWaldmann I try to figure out what causes the pystemmer trouble
2007-05-27T21:17:27  <ThomasWaldmann> lanius: you still need to create your wiki homepage
2007-05-27T21:18:14  <lanius> and what do i put on it
2007-05-27T21:19:25  <ThomasWaldmann> look at the many other users :) and there is a HomepageTemplate
2007-05-27T21:20:22  <dreimark> lanius: and subscribe to your homepage
2007-05-27T21:20:25  <ThomasWaldmann> btw, it is "ACLs"
2007-05-27T21:21:16  <ThomasWaldmann> see http://members.aol.com/apostrophs/ :))
2007-05-27T21:24:50  <ThomasWaldmann> lanius: if you use "user" as param, you won't have access to user module
2007-05-27T21:26:22  <lanius> proposal=
2007-05-27T21:26:23  <lanius> ?
2007-05-27T21:27:45  <ThomasWaldmann> uSeR :)
2007-05-27T21:31:03  <ThomasWaldmann> client ?
2007-05-27T21:31:24  <ThomasWaldmann> userobj ?
2007-05-27T21:31:24  <dreimark> ThomasWaldmann: do we have a table which language (word) a shortcut is and whic
2007-05-27T21:32:22  <ThomasWaldmann> i dont understand
2007-05-27T21:32:33  <dreimark> the libstemmer_c used for PyStemmer-1.0.1 does not know ro pages
2007-05-27T21:33:36  <dreimark> you can print the langs which work before the crash and it stops at ro pages
2007-05-27T21:34:17  <dreimark> amd i guess there are some more which are not supported
2007-05-27T21:36:45  <ThomasWaldmann> if you are talking about 1.6devXapianPyStemmerFails, there is a bug in PyStemmer. it shouldn't give that TypeError, but a KeyError (and we catch that KeyError)
2007-05-27T21:38:29  <dreimark> yeah but currently I think that occures because they have not supported romanian pages in this version
2007-05-27T21:39:42  <ThomasWaldmann> sure. nevertheless they have done something wrong or it wouldn't complain about TypeError
2007-05-27T21:39:43  <dreimark> the actual version of libstemmer_c supports it, I'll check if it makes a difference
2007-05-27T21:40:29  <ThomasWaldmann> you can work around it by also catching TypeError
2007-05-27T21:40:57  <ThomasWaldmann> (+ a comment telling that this is a workaround)
2007-05-27T21:42:13  <dreimark> the stemming will not make sense for a lang which isn't supported, do we need to compare the lists of i18n stuff before?
2007-05-27T21:42:47  <ThomasWaldmann> Stemmer(lang) IS the checker whether it is supported
2007-05-27T21:43:18  <dreimark> ok
2007-05-27T21:47:20  <lanius> ThomasWaldmann: i tried my best with spell checking
2007-05-27T21:47:27  <lanius> and my homepage :D
2007-05-27T21:51:37  <dreimark> ThomasWaldmann: what i wanted to say is it is not a problem of the python version
2007-05-27T21:55:19  <ThomasWaldmann> lanius: see latest diff :)
2007-05-27T21:55:43  <ThomasWaldmann> dreimark: ok, add that to the bug report
2007-05-27T21:59:58  <ThomasWaldmann> lanius: did you read http://moinmo.in/MoinMoinTodo/Release_2.0/StepByStep
2007-05-27T22:00:37  <ThomasWaldmann> (this is older than the stuff we discussed, but maybe has the one or other hint still useful)
2007-05-27T22:00:39  <lanius> yes i did some time ago
2007-05-27T22:03:25  <ThomasWaldmann> hmm, iirc we talked about that twikidraw stuff, but we didnt document it, right?
2007-05-27T22:04:25  <lanius> we said that it can be implemented on top of the design with two items
2007-05-27T22:06:23  <ThomasWaldmann> hmm. well, we'll see later. :)
2007-05-27T22:07:14  <lanius> yeah
2007-05-27T22:08:38  <ThomasWaldmann> you'll create some storage/ directory for your stuff?
2007-05-27T22:09:44  <lanius> yes
2007-05-27T22:10:05  <lanius> i don't need to touch existing code for at least half of the time
2007-05-27T22:11:06  <ThomasWaldmann> you should try to integrate early
2007-05-27T22:12:04  <lanius> i think integration should be the second step
2007-05-27T22:13:23  <ThomasWaldmann> it will begin after you have something to integrate, of course
2007-05-27T22:13:32  <dreimark> bbl
2007-05-27T22:13:38  <lanius> ^^
2007-05-27T22:13:51  <ThomasWaldmann> but if you delay it to the end, you might get trouble
2007-05-27T22:14:32  <lanius> we have to see how fast everything will be, maybe we have to delay integration, maybe not
2007-05-27T22:15:13  <ThomasWaldmann> (and the only REAL proof that your stuff works, will be seeing it running)
2007-05-27T22:28:48  <ThomasWaldmann> lanius: you could maybe start with user storage after doing the basic stuff
2007-05-27T22:30:46  <lanius> yes
2007-05-27T22:31:21  <ThomasWaldmann> feel free to ignore *.trail
2007-05-27T22:32:16  <ThomasWaldmann> 1.6 stores the trail into the session object
2007-05-27T22:41:57  <ThomasWaldmann> lanius: same for *.bookmark
2007-05-27T22:42:14  <ThomasWaldmann> (stored into user profile for 1.6)
2007-05-27T22:44:19  <ThomasWaldmann> so you just have some easy single files per user in there :)
2007-05-27T22:51:29  <ThomasWaldmann> lanius: still there? :D
2007-05-27T22:51:34  <lanius> yes, wondering what to say ;)
2007-05-27T22:52:26  <ThomasWaldmann> btw, we could use YAML for the user profiles later
2007-05-27T22:52:51  <lanius> haven't looked into YAML yet, but that is all for later
2007-05-27T22:52:55  <lanius> so where do i start soon
2007-05-27T22:53:56  <ThomasWaldmann> implement a minimal external interface, implement a 1.6 user storage backend, integrate and get it working
2007-05-27T22:54:47  <lanius> ok
2007-05-27T22:55:30  <ThomasWaldmann> and write some tests :)
2007-05-27T22:55:39  <lanius> that's the first thing ;)
2007-05-27T22:57:37  <ThomasWaldmann> could be storage/__init__.py,user16.py
2007-05-27T22:58:19  <lanius> yeah and some test data from the current underla
2007-05-27T22:58:19  <lanius> y
2007-05-27T22:58:40  <ThomasWaldmann> there are no users in the underlay :)
2007-05-27T22:59:33  <lanius> but you need pages as well for testing the backend, for metadata, revisions, data stuff
2007-05-27T23:00:12  <ThomasWaldmann> yes, that is after the users are working, isn't it?
2007-05-27T23:01:13  <lanius> 'm currently thinking how to support 1.6 style userrs and attachements and pages
2007-05-27T23:02:18  <ThomasWaldmann> user = user16(user_dir) maybe
2007-05-27T23:02:45  <ThomasWaldmann> for attachments, you need to simulate sub-items
2007-05-27T23:03:11  <lanius> or if it is better to just support pages in 1.6 style and the rest only new style
2007-05-27T23:03:15  <xorAxAx> ThomasWaldmann: yaml-based themes may not be distributed without infridging our copyright
2007-05-27T23:03:40  <ThomasWaldmann> brrr
2007-05-27T23:04:36  * ThomasWaldmann thought of the fridge :)
2007-05-27T23:04:51  <lanius> lol
2007-05-27T23:05:08  <xorAxAx> ah, its infringing
2007-05-27T23:06:01  <ThomasWaldmann> xorAxAx: so how does papaya do it?
2007-05-27T23:06:32  <xorAxAx> well, the idea would be that the authors of papaya agree to this gpl exception
2007-05-27T23:06:38  <xorAxAx> then everything would be ok
2007-05-27T23:07:37  <ThomasWaldmann> you could even distribute a yaml based theme without yaml and put in the yaml css files later
2007-05-27T23:08:28  <xorAxAx> hmm
2007-05-27T23:08:49  <xorAxAx> if that doesnt require you to set a link before you use their files, that might be ok, yes
2007-05-27T23:10:06  <xorAxAx> but you could already set the link in the theme that is stripped from css files already
2007-05-27T23:10:55  <ThomasWaldmann> well, let's wait for the author's answer, maybe he has other creative ideas :)
2007-05-27T23:13:01  <ThomasWaldmann> if it can't be integrated as one of the default theme's ressources, I am not interested anyway
2007-05-27T23:23:23  <ThomasWaldmann> lanius: ItemCollection needs the backend object as __init__ param
2007-05-27T23:25:53  <lanius> ThomasWaldmann: i would say it uses the backend defined in the config
2007-05-27T23:27:28  <ThomasWaldmann> if you have multiple, it needs to know which one
2007-05-27T23:27:48  <ThomasWaldmann> e.g. users, pages, caches
2007-05-27T23:27:58  <lanius> that's an argument
2007-05-27T23:28:25  <ThomasWaldmann> and for 1.6 compat storage, they are really different
2007-05-27T23:29:22  <lanius> yeah, that's good
2007-05-27T23:38:39  <ThomasWaldmann> the backend could have some attribute telling its capability to do revisioning
2007-05-27T23:39:17  <lanius> i don't think that is needed
2007-05-27T23:39:44  <lanius> since there is no difference for the backend if it does or not
2007-05-27T23:39:57  <ThomasWaldmann> but for the higher level code :)
2007-05-27T23:40:16  <lanius> but that is done in the higher level stuff then
2007-05-27T23:42:47  <ThomasWaldmann> hmm, can work if it just asks for creation of "next possible" revision (will be just overwriting the same rev if no revisioning is supported)
2007-05-27T23:48:20  <lanius> yes