OE IRC 5April07

From GO Wiki
Jump to navigation Jump to search
jclark has joined the channel
hjd_mgi: I'm wondering why I'm not hearing anything in the headphonnes, and then I remembered: not skype!
hjd_mgi: We have a fresh foot and a half of snow and it is still  coming down
midori: spring here, which means 65°F outside and almost 80°F in the office ...
tberardi has joined the channel
kchris: quite pleasant here too, though a bit overcast today
Alex_MGI has joined the channel
tberardi: Are we talking about the weather?
kchris: yes 
9:30 AM
Alex_MGI: We're having some weather today -- 16 inches of fresh white snow!
tberardi: Wow!
midori: more than we've had in the past three years ... cumulative
MelissaH has joined the channel
Alex_MGI: It's the biggest snowfall of this "season."
hjd_mgi: http://www.hazecam.net/acadia.html
j-lo has joined the channel
pfey has joined the channel
jrichter has joined the channel
midori: We're just waiting for our fearless leader ...
MelissaH: pretty. I miss snow.
jrichter: I've arrived! Free from fear!
midori: What's the weather like? 
jrichter: Who are you asking? It's sunny and beautiful in Denver.
jrichter: Or so I assume. I work in a near-windowless cell.
9:35 AM
hjd_mgi: This is more like it http://www.monticelloweather.com/weathercam.html
midori: Good one!
midori: Well, shall we start?
jrichter: Yes. A warning: I'm on baby duty right now. He's napping, but he could wake up at any minute. I might have to run out for 5 minutes in the middle to get him.
midori: OK
MelissaH: I know how that goes.
jrichter: Perhaps we should start with the burning question of OBO-Edit 1.1 releasability.
midori: So, the first item is OBO-Edit 1.1: is it ready to release?
midori: snap!
jrichter: (great minds)
midori: Several of us have essentially signed off. Chris had some concerns about the verification plugin, but Jane suggested that just having it off by default would do the trick. Thoughts?
jrichter: The only for-sure outstanding bugs are Chris's bug and Jen's reload bug. Erika's multi-ontology bug is still under consideration.
jclark: seems fine
hjd_mgi: I agree
midori: Erika's problem with loading and saving seems to be addressed by turning off some verification plugin checks.
tberardi: I saw 24 in action yesterday morning during our ontology editing session with Jen and it did all the things I need it to.
tberardi: So, by transitivity, I'm signing off on it too.
jrichter: I think we need Chris & Jen's bugs addressed (especially Chris's). The verification plugin needs to be working well in the official release.
midori: Erika also asked about the headers in the filtered file -- I pointed her to the category-filtering option, but all of the remarks from both GO and ChEBI also end up in the saved file. I do see that the GO remark and the ChEBI remarks all use the same tag, so I've warned Erika that she might have to delete those lines manually, but I promised I'd ask.
9:40 AM
jrichter: I'm surprised it's smart enough to munge the comments together!
jrichter: Can't erika fix that by using the comment editing box when she saves?
jrichter: Or are we talking about a different set of remarks?
midori: Probably -- but even I had forgotten there was a comment editing box!
jrichter: I think that's the fix. She still needs to edit the comments "by hand", but at least she isn't editing the file directly.
jrichter: As far as the other bugs go:
jrichter: Chris's bug has to do with reasoner logic, and I bet it's pretty easy to fix.
jrichter: I don't know what's going on with Jen's bug.
jrichter: Are there additional problems with the verification plugin that aren't on the bug report page?
midori: Sorry, which one is Chris' bug?
jrichter: Or any other module.
jrichter: ?
midori: I'm still having one problem with a custom check in the verification plugin: I don't seem to be able to turn off running a check on load.
jrichter: Chris's bug is not officially reported, but he's having problems with the disjointness check, which is tied into the reasoning system. It's giving him wrong answers.
jrichter: Huh. Is that true of any custom check?
midori: The relevant box is unchecked in the config panel, but the check still runs on load.
midori: Lemme, er, check ...
j-lo: That doesn't seem to be the case for me
midori fires up obo-edit ...
jrichter: While Midori's lighting up, are there any other unreported problems I should know about?
9:45 AM
MelissaH: everythings been well in anatomy land.
jrichter: (An aside: I know it's taking forever to get the finishing touches done to get this out the door, but that seems to be the typical software release cycle. 20% of problems take 80% of the work)
midori: that's not software-specific!
jrichter resists making a "I'm glad the symptoms have cleared up" joke in reference to Melissa's last comment.
MelissaH: heeheee!
j-lo: I had some questions about disjointness, but no bugs
jrichter: Anyway, whether or not Midori's bug is real, I want to get Chris & Jen's (& Midori's, if applicable) bugs fixed before this goes out the door. These are tiny fixes that should be mostly self-contained, so I think I should release fixes as new betas, ONLY the interested people need test them and sign off, and once I have the signoffs, I'll drop an official release.
jrichter: Sound okay?
jrichter: Jane - What's your question about disjointedness?
midori: OK ... I just successfully turned off custom check #2; custom check #1 is still running!
jrichter: ?
jrichter hears baby squeaks, he rushes off to the rescue.
midori: So it's apparently not happening with just any custom check
jrichter: Be back in a few minutes.
j-lo: two qs: 1 if A and B are disjoint, is it sufficient to mark A disjoint_from B, or do I also need to do the inverse?
j-lo: second q
jrichter: While I'm out, Midori, can you describe the custom check that's having the problem.
9:50 AM
jrichter: And when I get back I'll address Jane's questions too.
jrichter steps out
midori: The problem one is simply "NOT [self] [is is_a  complete]"
j-lo: is there any way I can limit to only is_a disjointness? (i.e. I only want to see cases where terms have is_a children in common, not is_a or part_of
MelissaH: This is similiar I think, to an earlier request I had to filter for terms with more than one is_a parent. John said 1.3 format release.
j-lo: right - thought that might be the case
Anon271 has joined the channel
9:55 AM
jrichter: I'm back.
jrichter: Midori - can you post a bug report for this problem?
midori: ok
jrichter: Jane 1) I need to look that up. You *should* only need to mark a disjoint_from b, but I don't know if the reasoner is smart enough to do the reverse direction yet.
jrichter: Jane 2) disjointedness only ever considers is_a
jrichter: No other relationship is taken into account.
j-lo: huh - I think there's a bug then
jclark has quit the server saying: Ping timeout
j-lo: yesterday I marked some terms as disjoint, and then  ran the verification check
Anon271: I haven't quit really.
j-lo: and it pulled out term that had an is_a to term A and part_of to term B
j-lo: (terms)
j-lo: is that the same as Chris's bug though?
10:00 AM
jrichter: I'm not sure. I think I'd like you to send me the ontology file you're working with.
jrichter: But that's certainly bad behavior.
j-lo: okay - I'll send it all over...
jrichter: Jane - now that I think about it, you only need to mark disjoint in one direction. The reasoner isn't smart enough to figure out the symmetry, but the verification check will still pick up any problems (assuming it's working properly)
jrichter: Are people okay with the release scheme I proposed a page or two ago?
j-lo: yep
hjd_mgi: yes
kchris: I have a question about the verification plugin
jrichter: Karen - please proceed
kchris: I'm fine with making it off by default, and agree with Chris that it's not quite ready for general  use
kchris: but could we add some of the checks that GO wants into the general list, checking for missing namespaces seems generally useful, and if it can be turned on or off, then it would be helpful to GO if it's  part of the standard package
kchris: and no one who doesn't want this check is stuck with it
midori: That actually touches on agenda item 2, which is how to distribute any set of checks that all members of a given ontology development group should use.
Anon271: Midori suggested we might make checks load and saveable so we can distribute good new ones.
Anon271: I thought that was a good idea
10:05 AM
jrichter: Okay, let me get into some of these...
MelissaH: (from a non-GO persepective) I have to admit I never use the verification plug-in- I get over a thousand non-critical warnings and no critical warnings. Also, I have always assumed is_a by definition implied disjointness. 
MelissaH: (sorry to interrupt- just wanted to comment I think it ok to have it turned off by default)
jrichter: First - the verification plugin needs to be ready for prime-time before we release OBO-Edit 1.1. A lot of basic checks that were built into the text editor component are now part of the verification plugin. A lot of responsibilities have been offloaded to it, and it HAS to work.
jrichter: Second - the "missing" namespace check isn't as simple as it seems. For GO, it's not a missing namespace that's the problem - it's a bad namespace. Those terms belong to the "gene_ontology" namespace, and they shouldn't.
jrichter: That means that a check would either be ontology-specific, or we'd need to add a "terms belong to default namespace" check. If we do the latter, it'll have to be off by default, because most ontologies only have one namespace - the default - and all the terms belong to it.
midori: I've configured the relevant check, so I don't think we should spend precious chat time on its parameters. Instead, I want to concentrate on making it convenient for everyone to use the check (and inconvenient not to).
jrichter: Right!
jrichter: There's a way to do this now, but maybe it's too inconvenient...
midori: It seems to me that GO isn't the only ontology that would want to run a specific set of checks.
jrichter: In your .oboedit folder there's a file called "verify.xml". It contains all your verification settings, including any custom checks you
jrichter: 've defined.
jrichter: Anyone can copy your verify.xml file into their .oboedit directory and take on all your verification settings.
10:10 AM
midori: I don't mind having different verification plugin configs -- for GO, CL, anatomies, whatever -- come into existence, as long as the configs are sharable and documented.
Anon271: we could probably also hand edit the file then?
midori has left the channel saying: ... and I'm happy to expose my verify.xml to (some of) the world; in fact, it's now attached to my new bug report, SF 1695127
midori has joined the channel
midori: aaah! I closed the chat window by mistake! Can someone save a transcript??
Anon271: I think it would be good if we don't have to obliterate our own list of checks to add someone else's new one but if I can just edit a check in that would be fine
kchris: I can, I've got from the beginning, including entertaining pre-chat chat 
midori: phew! thanks!
midori: anyway, 
midori: when my hand slipped I was just saying that my verify.xml is now attached to my new bug report, SF 1695127
Anon271: I think it would be good if we don't have to obliterate our own list of checks to add someone else's new one but if I can just edit a check in that would be fine 
jrichter has quit the server saying: Ping timeout
10:15 AM
hjd_mgi: getting back to the name space errors, is it clear that they could NOT be made IN Oboedit itself (because you can't make a child of gene_ontology directly unless you use the drop down to specify that name space?
Anon271: Isn't it right that if you load a term with no namespace then it adds the default one?
midori: yes
kchris: Do people have any thoughts about the likely cause of the namespace errors?
Anon271: the last one was made when I did a format conversion in oboedit
Anon271: I have no idea of how it managed to do that.
kchris: sorry, who lurks behind 'Anon271'? 
Anon271: I think it was to do with it being an old file maybe
Anon271: jclark
Anon271: I don't know how that happened either
kchris: in "it was to do with it being an old file maybe" what does "it" refer to?
Anon271: the loss of the namespace must have been to do with the file that I converted having been made some time ago
Anon271: I think
Anon271: that's the only unusual thing that was going on
Anon271: it's not something I'm planning to do again
Anon271: so I don't think we need to worry too much about it
kchris: but we have had this problem 3 times in the last 6 weeks or so, and it breaks scripts
10:20 AM
Anon271: the other two time were a hand edit then a bad merge
Anon271: the verification check will solve this
midori: I think the fact that we'll be able to catch the default-namespace errors by using the custom check is the most important thing -- that's what will enable us to fix them before they go live and foul things up.
Anon271: quite
kchris: having it in the verification is enough for me, I just meant that we needed to make sure we were catching them before they got committed
midori: That's why I'm trying to keep the discussion on the topic of distributing sets of checks.
Anon271: good plan
midori: John's gone missing ...
midori: In case he doesn't resurface, shall we (meaning GO developers; apologies to Mel) agree to deploy the verification plugin check for default namespaces?
Anon271: yes
midori: ... and is there anything else to discuss in John's absenceL
Anon271: we were going to make a list of other checks but we can do that another time
midori: er, that was supposed to be a question ...
kchris: at some point, maybe we could talk about what and where is appropriate documentation for GO changes
midori: The most important one is the namespace check; second most important is checking for is_a-completeness.
10:25 AM
midori: and at some point we'll have to come back to disjointness
midori: none of the rest really break scripts, so they aren't as urgent.
midori: documentation for GO changes isn't an OBO-Edit issue, so let's take that up in a different forum.
kchris: yes, but what forum, we don't have a list for GO editors?
MelissaH: Do you all check for is_a completeness by looking for orphans?
MelissaH: its ok, I'll sign off if you guys want to continue. but I have one more related question first.
midori: In the recent 1.1 betas, the search pulldown has an option "is isa complete" -- I just use that with NOT.
MelissaH: I see, then you can go straight to the term and fix it while its in its native environs. makes more sense.
midori: and an easier filter to set upp ! 
MelissaH: But you can also see all the orphans just by changing the root algorithm.
MelissaH: (thats how I was looking for them before, but we are very far from is_a complete).
midori: true ... but  that wouldn't fit neatly into a verification plugin check ...
midori: whatever works, I guess!
j-lo: yes, that's how I used to look at them too Melissa
MelissaH: ok, next question. Our DBA wants to know why we have to use catagories/slims (for PATO) instead of just putting the different sub-hierarchies in different namespaces. Are these really two ways of accomplishing the same thing?
10:30 AM
midori: From a GO-centric perspective, the answer is that using categories lets us put any term into more than one of them -- I think you wouldn't be able to do that with namespaces.
MelissaH: Also, if you add a child of something in a given namespace, does it automatically assign that namespace to all descendents?
j-lo: no, I think they're different
j-lo: and no again
j-lo: namespaces are for being able to align separate ontologies
j-lo: and shouldn't change
j-lo: categories are much more flexible
MelissaH: so does it make sense for two sub-ontologies in pato to be in different namespaces or in different slims?
j-lo: (I think!)
MelissaH: they are disjoint
j-lo: hmmm - good question
midori: I don't think I know PATO well enough to have a good answer.
j-lo: that's probably a judgement call
midori: But certainly you'll want to think about whether you'll ever have subsets that aren't disjoint ... for those my guess would be categories better.
MelissaH: ok, just thought I'd pick your brains a bit before I suggest any changes... carry on!
hjd_mgi: \I need to leave for another meeting; will check log of the chat.
hjd_mgi: bye
hjd_mgi has quit the server saying: Quit: ChatZilla 0.9.78 [Firefox]
midori: I think we're about ready to wrap up here anyway ... certainly I could also do to get going.
midori: Karen has volunteered for transcript duty, since I'm a klutz.
kchris: before you leave Midori,
10:35 AM
kchris: would it make sense to have a list for all people editing GO in order to talk about GO practice for editing. I agree with your comment in the email you posted that such a list wouldn't be a good place for content discussions, but right now the OBO-Edit WG is the closest place to a forum where people can discuss this type of issue, and it's not quite the right forum
midori: let me think about it ... could you send an email to the go-editors@ebi.ac.uk list as a start? I've got to go now or I'll miss my train.
MelissaH: Just so you guys know, I think its fine- I am happy to beg out when its not relevant. Usually its more relevant than you might think. and I learn about ontology editing issues that I might not otherwise be privy to.
kchris: OK
midori: bye!
kchris: there are also people editing who are not in the OEWG though...
midori has quit the server saying: Quit: Goodbye