OE Webex 9Dec08

From GO Wiki
Jump to navigation Jump to search

12/09/2008 09:28:00 AM from Karen: Hi
12/09/2008 09:30:00 AM from midori: hi folks
12/09/2008 09:30:07 AM from Jen Deegan: Hi
12/09/2008 09:30:32 AM from Amina Abdulla: Hi Everyone
12/09/2008 09:30:58 AM from Amina Abdulla: We have a short list of agenda items for today
12/09/2008 09:31:55 AM from Amina Abdulla: I have to confess I haven't done much testing
12/09/2008 09:32:20 AM from Amina Abdulla: I'm working on getting another release out in a day or two
12/09/2008 09:32:25 AM from Jen Deegan: I have been digging in the tree viewer
12/09/2008 09:32:32 AM from midori: Neither have I; paperwork has eaten my day today.
12/09/2008 09:32:58 AM from Jen Deegan: I solved one bug, and then after a week it came back.
12/09/2008 09:33:11 AM from Karen: My GO time has gone to a content issue, but it will give me something real to edit for testing soon ;)
12/09/2008 09:33:12 AM from Amina Abdulla: Hows that going Jen - any more insights on the treeviewer bug
12/09/2008 09:33:31 AM from Jen Deegan: Yes I have got a lot more insight.
12/09/2008 09:33:39 AM from Amina Abdulla: thats great Karen
12/09/2008 09:33:54 AM from Harold: hi everyone
12/09/2008 09:33:55 AM from Jen Deegan: I would like to understand much better how oboedit is meant to deal with transitivity though.
12/09/2008 09:34:11 AM from midori: I do at least have more editing-in-OE2 time logged, so not totally irrelevant to testing.
12/09/2008 09:34:41 AM from David OS: Jen: What don't you understand about OE2 transitivity?
12/09/2008 09:35:34 AM from Jen Deegan: hang on I just have to check something. Carry on talking for a bit.
12/09/2008 09:35:45 AM from Jen Deegan: even the question is hard to write.
12/09/2008 09:36:01 AM from Amina Abdulla: ok
12/09/2008 09:36:12 AM from Jen Deegan: carry on for a bit, I'll be back.
12/09/2008 09:37:05 AM from Amina Abdulla: I sent out an email for the dictionary fix: OE2 now has a seperate user-defined and standard dictionary
12/09/2008 09:37:18 AM from Jen Deegan: great!
12/09/2008 09:37:22 AM from Karen: cool :)
12/09/2008 09:37:25 AM from David OS: Jen - I know what you mean. Reasoning is hard enough to write about when you have time and space to do it. Can be tough in a chat forum.
12/09/2008 09:37:44 AM from Amina Abdulla: users can backup the user.dic to a seperate location from the Configuration Manager
12/09/2008 09:37:56 AM from David OS: Excellent
12/09/2008 09:38:12 AM from Jen Deegan: I have my question now when you've got a minute.
12/09/2008 09:38:16 AM from Amina Abdulla: and reseting the config files does not affect the dictionary files
12/09/2008 09:38:36 AM from Jen Deegan: excellent
12/09/2008 09:38:37 AM from midori: dictionary stuff sounds excellent
12/09/2008 09:38:39 AM from Amina Abdulla: The standard dictionary files can be updated thru the Configuration Manager too
12/09/2008 09:39:02 AM from Amina Abdulla: this is when the editors add new terms to the standard GO dic and submit to svn
12/09/2008 09:39:39 AM from Jen Deegan: do you mean that they can submit directly to svn?
12/09/2008 09:39:46 AM from Amina Abdulla: there are actually 4 standard GO dictionaries
12/09/2008 09:40:01 AM from Amina Abdulla: No users would have to send it to me
12/09/2008 09:40:07 AM from Jen Deegan: ah, good to know.
12/09/2008 09:40:44 AM from Amina Abdulla: or you could add it yourself Jen
12/09/2008 09:40:58 AM from Jen Deegan: Yes I can also do that if people send them.
12/09/2008 09:42:00 AM from Amina Abdulla: Great - and while adding a new word to dic through the Text Editor there will be an option to add to user.dic or standard.dic
12/09/2008 09:42:49 AM from Harold: great
12/09/2008 09:44:11 AM from David OS: So, Jen. What was your question?
12/09/2008 09:44:16 AM from Amina Abdulla: Midori re: http://sourceforge.net/tracker/index.php?func=detail&aid=2355661&group_id=36855&atid=418257
12/09/2008 09:44:39 AM from Jen Deegan: Okay, it's kind of complex.
12/09/2008 09:44:53 AM from midori: oh yes, that bug. more annoying than crippling
12/09/2008 09:44:53 AM from Jen Deegan: I think that the tree viewer is getting into a loop when it handles disjoints.
12/09/2008 09:44:56 AM from Amina Abdulla: You only see this the first time you start a new dbxref right
12/09/2008 09:45:16 AM from Jen Deegan: because each term that is disjoint to another term, that other term is also disjoint to the first term
12/09/2008 09:45:31 AM from Jen Deegan: so OE goes round and round and round and eventually flips out.
12/09/2008 09:45:38 AM from Jen Deegan: but there's this test in the code.
12/09/2008 09:45:41 AM from Jen Deegan: like this:
12/09/2008 09:45:51 AM from Jen Deegan: if graph is circular, do X
12/09/2008 09:45:54 AM from midori: I think so, but it's intermittent anyway (sometimes it doesn't happen at all when I add a dbxref), so it' shard to be sure.
12/09/2008 09:46:34 AM from Jen Deegan: or if OE is configured to show non-transitives and the path contains non-transitives do X
12/09/2008 09:46:36 AM from Amina Abdulla: I've looked into it and see it in all sections where dbxrefs and xrefs are created (I can talk about it after Je n's discussion)
12/09/2008 09:47:04 AM from midori: ok, one topic at a time is good
12/09/2008 09:47:14 AM from Jen Deegan: and the disjoint relationship in OE claims that it is non non-transitive
12/09/2008 09:47:27 AM from Jen Deegan: or in other words disjoint is transitive.
12/09/2008 09:47:40 AM from Nomi Harris: The ontology tree editor had a similar problem with disjoints
12/09/2008 09:47:53 AM from Nomi Harris: I added code to make sure it wouldn't go down that circular path.
12/09/2008 09:48:01 AM from Nomi Harris: You could look at that and see if you could do something similar in the Tree Viewer.
12/09/2008 09:48:02 AM from Amina Abdulla: Jen would you like to demo this bug?
12/09/2008 09:48:04 AM from Jen Deegan: but is disjoint transtive? I thought not.
12/09/2008 09:48:15 AM from Nomi Harris: It's not exactly that it's transitive;
12/09/2008 09:48:18 AM from midori: disjoint is symmetric
12/09/2008 09:48:21 AM from Jen Deegan: It's not terribly exciting actually.
12/09/2008 09:48:25 AM from Nomi Harris: the issue is that there's usually another disjoint that makes it circular.
12/09/2008 09:48:33 AM from Jen Deegan: yes
12/09/2008 09:48:33 AM from Nomi Harris: cat is disjoint from dog is disjoint from cat...
12/09/2008 09:48:50 AM from Jen Deegan: so the check on whether it is circular to should by triggered when I select a disjoint term?
12/09/2008 09:48:58 AM from Jen Deegan: exactly
12/09/2008 09:49:09 AM from Jen Deegan: cats and dogs all over the shop before you know it.
12/09/2008 09:49:12 AM from Nomi Harris: I can look back at the discussion from a few months ago when I fixed the disjoint issue in the OTE
12/09/2008 09:49:20 AM from David OS: So the OTE deals with this by suppressing display of children from a term hanging from a disjoint_from parent
12/09/2008 09:49:23 AM from Jen Deegan: that would be great if you didn't mind.
12/09/2008 09:49:30 AM from Jen Deegan: does it?
12/09/2008 09:49:35 AM from Jen Deegan: that's good to know.
12/09/2008 09:49:42 AM from Nomi Harris: David, right, and I also made it show those relationships in gray so it was visually clear that they weren't to be expanded.
12/09/2008 09:50:20 AM from Jen Deegan: but if I just click on a disjoint term then it should just display that term and the relationship to the root right?
12/09/2008 09:50:27 AM from Jen Deegan: not the other disjoint terms?
12/09/2008 09:50:37 AM from Nomi Harris: Right
12/09/2008 09:50:38 AM from Chris Mungall: disjoint is symmetric, non-transitive. symmetric relations should not be shown in the tree viewer. that should fix it.
12/09/2008 09:50:41 AM from Jen Deegan: and what if the program is configured to show non-transtives?
12/09/2008 09:50:53 AM from Jen Deegan: oh
12/09/2008 09:51:17 AM from Jen Deegan: thank you.
12/09/2008 09:51:21 AM from Chris Mungall: The OTE is still having problems with disjoint_from. Half of the branches in SO are invisible.
12/09/2008 09:51:39 AM from Chris Mungall: ...last time I check
12/09/2008 09:51:42 AM from Chris Mungall: ed
12/09/2008 09:51:44 AM from David OS: The problem with the fix for the OTE is that it prevents terms under disjoint roots from being displayed.
12/09/2008 09:51:44 AM from Jen Deegan: that's enough for me to be going on with. Thanks. :-)
12/09/2008 09:52:18 AM from Chris Mungall: ah, that's a problem David
12/09/2008 09:52:25 AM from Nomi Harris: There is a problem with SO but I wasn't sure it was because of the way disjoint_from is displayed.
12/09/2008 09:52:29 AM from Chris Mungall: we'll have to fix that
12/09/2008 09:52:44 AM from David OS: https://sourceforge.net/tracker/index.php?func=detail&aid=2412063&group_id=36855&atid=418257
12/09/2008 09:53:05 AM from David OS: Don't know whether this is the same problem as for SO
12/09/2008 09:53:31 AM from Nomi Harris: I have always thought that disjoint_from relationships should not be included in the ontology. They create a lot of problems.
12/09/2008 09:54:35 AM from David OS: They are very useful in the ontology - esp for error checking. The problem is more with treating them as relationships I think
12/09/2008 09:55:22 AM from Nomi Harris: Yes
12/09/2008 09:55:30 AM from David OS: Perhaps it would be better to display them in a tab of the text editor rather then in trees/graphs?
12/09/2008 09:55:49 AM from Nomi Harris: That's what I meant--they should be separate from the structure of the ontology itself.
12/09/2008 09:57:10 AM from Jen Deegan: if I just don't display them in the tree viewer it will help a lot I think. Now I have a better idea of how to do it.
12/09/2008 09:57:18 AM from David OS: They might be easier to manage in that way - especially if we had auto-propagation between partners in a disjoint_from relationship (as in Protege4)
12/09/2008 09:57:42 AM from midori: What does auto-propagation do?
12/09/2008 09:58:10 AM from Jen Deegan: By the way, I'm going on a two day protege ontology editing meeting next week so if anyone has specific things they want me to find out this would be a great time to tell me.
12/09/2008 09:59:40 AM from David OS: By auto-propagation I mean - You should only need to add disjoint to one of the partners to see it displayed for the other. This is how P4 works. It's a pain to work with disjoints without this functionality.
12/09/2008 10:00:02 AM from Jen Deegan: that's a very interesting thought.
12/09/2008 10:00:05 AM from midori: OK, makes sense (and sounds useful)
12/09/2008 10:01:16 AM from David OS: I'm afraid I have to go. Just one quick thing for those interested in XPs. I've written up some stuff on the GO wiki about using them. http://wiki.geneontology.org/index.php/Working_with_XPs Hope some of you find this useful background to XO meeting this Thurs.
12/09/2008 10:01:42 AM from Amina Abdulla: thanks David
12/09/2008 10:02:05 AM from David OS: See you. Gotta rush.
12/09/2008 10:02:25 AM from Amina Abdulla: bye
12/09/2008 10:02:38 AM from midori: thanks & bye
12/09/2008 10:02:57 AM from Jen Deegan: If I make it so there is only one disjoint relationship going between each pair of terms in the file, my tree viewer bug goes away.
12/09/2008 10:03:19 AM from Amina Abdulla: yeah that should work
12/09/2008 10:04:00 AM from Jen Deegan: so should we just be changing the policy on having disjoint relationships in both directions? Or should I not be displaying symmetrical relationships? That sounds like an important policy decisions.
12/09/2008 10:04:25 AM from Amina Abdulla: I like Davids suggestion to display disjoints in a seperate tab
12/09/2008 10:05:25 AM from midori: me too
12/09/2008 10:05:41 AM from Amina Abdulla: Jen should we send this out to the working group?
12/09/2008 10:05:52 AM from Jen Deegan: I think we're mostly here anyway
12/09/2008 10:06:10 AM from Jen Deegan: I'm wondering what Midori and Chris think.
12/09/2008 10:06:18 AM from midori: Harold & Tanya aren't here ...
12/09/2008 10:06:44 AM from Jen Deegan: If I don't have disjoints in both directions between two terms then the disjoint only shows in the parent editor of one of the terms.
12/09/2008 10:06:46 AM from Jen Deegan: true
12/09/2008 10:07:08 AM from Jen Deegan: it would be nice to hear from Chris at least, and then we could send it to the list.
12/09/2008 10:07:14 AM from Jen Deegan: since he's here.
12/09/2008 10:07:36 AM from midori: Ideally, it wouldn't matter to OE whether we explicitly state a disjoint in both directions or not, because with a symmetrical relationship, the two (one dicection or both) are logically equivalent.
12/09/2008 10:08:04 AM from Jen Deegan: so in fact the relationship should show in the parent editor of both
12/09/2008 10:08:15 AM from Jen Deegan: because it's symmetrical, and it should only need to be made once?
12/09/2008 10:08:45 AM from midori: because making it once or making it twice are saying exactly the same thing
12/09/2008 10:08:49 AM from Jen Deegan: so in fact that would indicate a change to the parent editor, rather than a change to the tree viewer.
12/09/2008 10:09:17 AM from Jen Deegan: currently making it once and twice are very different things to the tree viewer, but maybe it needs some help with that.
12/09/2008 10:09:25 AM from Chris Mungall: 1. Editors should only need to assert symmetric relations in one direction
12/09/2008 10:09:43 AM from Jen Deegan: if it's not going to display symmetrical relationships, then it shouldn't matter how many there are. They still shouldn't be there.
12/09/2008 10:10:09 AM from Chris Mungall: 2. The various components need to be smarter in general about showing symmetric relations
12/09/2008 10:10:21 AM from Chris Mungall: 3. a separate tab for disjoints is a good idea
12/09/2008 10:11:15 AM from midori: The thing that worries me about not showing symmetric relations is that some ontologies will need to use them (not just disjoint)
12/09/2008 10:11:18 AM from Chris Mungall: but remember for the OE2 release we only have to be as good as OE1. OE1 was not great for disjoint_from - but SO worked of course
12/09/2008 10:11:56 AM from midori: agreed - we can improve in later releases
12/09/2008 10:12:04 AM from Chris Mungall: We should definitely show them. Perhaps not in the OTE, but always in the parent editor, and possibly optionally in the various graphical displays
12/09/2008 10:12:43 AM from Chris Mungall: It's actually rare that other ontologies will want to add relations that are symmetric at the type level. even adjacent_to is not symmetric at the type level. for example
12/09/2008 10:12:59 AM from Chris Mungall: (all) nucleus adjacent_to (some) cytoplasm
12/09/2008 10:13:12 AM from midori: OK; I just don't want to take away the option lightly.
12/09/2008 10:13:13 AM from Chris Mungall: but not the inverse. e.g. proks
12/09/2008 10:13:38 AM from Chris Mungall: Yep, I agree, I'm against hiding things all together.
12/09/2008 10:13:58 AM from Nomi Harris: What's proks?
12/09/2008 10:14:07 AM from midori: prokaryotes
12/09/2008 10:14:17 AM from Nomi Harris: Oh, right, I figured that out just as I hit return...
12/09/2008 10:14:21 AM from midori: bacteria & archaea
12/09/2008 10:15:32 AM from Amina Abdulla: Midori thanks for replying to Victoria - that was an easy fix.
12/09/2008 10:15:55 AM from midori: always nice when it's an easy one :)
12/09/2008 10:16:28 AM from Jen Deegan: Is everybody agreeing with Chris's three points then? Should we ask the list before we call that a made decision?
12/09/2008 10:16:51 AM from Jen Deegan: Just before I fix any bugs on that basis.
12/09/2008 10:17:57 AM from midori: I'm happy with those points. It would be polite to ask the list, but make sure there's a firm, and not too distant, deadline for replies, because I suspect there won't be many strong feelings.
12/09/2008 10:18:22 AM from Jen Deegan: Shall I do that then since it's me that needs the answer?
12/09/2008 10:18:39 AM from Jen Deegan: and just to the working group?
12/09/2008 10:18:46 AM from midori: would be great if yo u would; thanks
12/09/2008 10:18:52 AM from Jen Deegan: okay, will do.
12/09/2008 10:18:53 AM from Amina Abdulla: yes go ahead Jen and thanks
12/09/2008 10:19:08 AM from Jen Deegan: great. Thanks, this was fantastically useful discussion.
12/09/2008 10:21:00 AM from Chris Mungall: sorry for the lag. I was on a cell ontology call.. taking down OE requirements!
12/09/2008 10:21:21 AM from midori: :D
12/09/2008 10:21:27 AM from Amina Abdulla: so the dbxref bug I was talking about earlier propogates in all sections of the TextEditor - what its actually doing is asuming editing is complete and draws the label for the xref - fixe d in the next beta
12/09/2008 10:21:41 AM from Chris Mungall: The dendritic cell folks are using the intersection editor component quite a lot (the one that gives the big long list of xps withoyt giving you anyway to filter it...)
12/09/2008 10:21:45 AM from midori: fixed is good; thanks!
12/09/2008 10:22:55 AM from Jen Deegan: have they seen the bug tracker?
12/09/2008 10:22:57 AM from Amina Abdulla: yes I'd like to know use cases for Intersection Editor
12/09/2008 10:23:50 AM from Chris Mungall: the main use case for the IE is that it is currently v hard to get a summary or list of all xps any other way
12/09/2008 10:24:13 AM from Amina Abdulla: ok makes sense
12/09/2008 10:24:16 AM from Chris Mungall: in fact this is true for a lot of other things in OE. e.g. just seeing a list of terms with their text definitions
12/09/2008 10:24:32 AM from Chris Mungall: but this is getting into OE2.1 terriotory
12/09/2008 10:25:38 AM from Amina Abdulla: so the cell folks would like to filter xps listed in the IE?
12/09/2008 10:26:55 AM from Chris Mungall: They didn't say this, I was just observing that they were using this. But I think that the moment you have to scroll down more than a couple of pages to find what you want, you want some kind of filtering capability (this goes for any component in any pie
12/09/2008 10:27:36 AM from Amina Abdulla: Right
12/09/2008 10:27:57 AM from Amina Abdulla: I'll put it in the feature tracker
12/09/2008 10:28:01 AM from Chris Mungall: ce of software
12/09/2008 10:28:18 AM from Chris Mungall: (not pies)
12/09/2008 10:28:51 AM from midori: I kind of liked the mental image of pie ...
12/09/2008 10:29:10 AM from Amina Abdulla: :)
12/09/2008 10:29:28 AM from Amina Abdulla: Any more items for today?
12/09/2008 10:29:34 AM from midori: (dinnertime fast approaching here)
12/09/2008 10:29:38 AM from Jen Deegan: I'm done
12/09/2008 10:29:44 AM from midori: me too
12/09/2008 10:29:59 AM from Amina Abdulla: Alright Thanks everyone
12/09/2008 10:30:09 AM from Nomi Harris: Bye
12/09/2008 10:30:09 AM from Chris Mungall: bye
12/09/2008 10:30:12 AM from Jen Deegan: Thanks for all the help on symmetry :-)
12/09/2008 10:30:12 AM from Amina Abdulla: Bye for now
12/09/2008 10:30:14 AM from Jen Deegan: bye!
12/09/2008 10:30:30 AM from midori: thanks all; bye