AmiGO-chat-10April07

From GO Wiki
Revision as of 12:42, 10 April 2007 by Jl242 (talk | contribs) (New page: j-lo: So the IEA instance of AmiGO - what did we think? eurie: it's pretty sluggish and overwhelming as is. donghui: i tried the servers, depending on the term, the search time varies from...)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

j-lo: So the IEA instance of AmiGO - what did we think? eurie: it's pretty sluggish and overwhelming as is. donghui: i tried the servers, depending on the term, the search time varies from several seconds to over 30 sec, knid of slow rama: I tested it briefly and nothing is broken as far as I can tell hitz: the counts on browse page are off which makes it hard to test terms with ridiculous ## of GAs j-lo: It wasn't that much slower than normal AmiGO for me sjcarbon: i didn't notice a lot of use, so maybe 'slow' people were over lapping sjcarbon: we could schedule a time for everyine to hit it at the same time sjcarbon: just for fun j-lo: how about now? eurie: right now? =) donghui: could be, I tested right after the email was sent donghui: sure sjcarbon: go! gwg: could you send the url again? save me searching my inbox gwg: (sorry!) donghui: that would be great! rama: http://toy.lbl.gov:9002/cgi-bin/amigo/go.cgi gwg: cheers! sjcarbon: right now 6 processes j-lo: okay, everyone test sjcarbon: 12 processes sjcarbon: 9 and holding hitz: pretty painful 8:45 AM hitz: I think the association details pages need to be organized eurie: especially the gene product searches donghui: I tested the term kinase, there was a noticeable delay in getting result cjm: Seth - I haven'tperformance should be easy to benchmark with and without IEAs in an objective way. do we not have some scripts for this....? hitz: searchs are slow in all versions though. hitz: we have a timing branch, no? sjcarbon: i could get some performance scripts ready hitz: also, our new amigo-dev machine arrived yesterday sjcarbon: the timing branch was just for individual use hitz: they would be great to test new loading/old loading gwg: you wouldn't want to be in a situation where your life depended on getting the results from an AmiGO gp search... (something that happens to me all the time, obviously) j-lo: so it sounds like this is currently too slow for our needs sjcarbon: if everybody went and hammered on amigo right now, would it also slow down? gwg: let's give it a go! hitz: I just checked gowebs loads are <1 8:50 AM gwg: i think like Chris says, it would be good to get an objective view of how much slower the full load would be gwg: there are so many variables here that I don't think it's really all that informative to be testing like this hitz: I agree with amelia and chris hitz: it does seem subjectively slower j-lo: okay - so do we have a way to do this? j-lo: measure speed i mean sjcarbon: i can write that j-lo: Okay, great. Action item for Seth rama: the associations page has to be organized. We should have a summary table of associations like the annotations table on the GOC site and let user decide which ones they want to view further gwg: which do you call the associations page? j-lo: yes, I agree - there are too many results with IEAs sjcarbon: i think the issue is getting the summary table without doing as many joins hitz: yes 184 pages of associations = useless. We have filters, but if they are "mandatory" it should be more explicit j-lo: the results page rama: the page you get when you click on the View associations gwg: term associations? gp associations? gwg: term associations hitz: term details page eurie: yeah, i think in terms of user experience, the current results page when you search on gp or view associated annotations is too overwhelming hitz: seth, we probably need to put this in a view sjcarbon: that would be great hitz: unless there are some clever indexes (indices?) we can add sjcarbon: make the API easier too 8:55 AM rama: The 'view all results' doesn't retrieve 45 pages of results by the way rama: I don't know what you guys are talking about (Ben and Seth) hitz: in sql you do joins across multiple tables. We need to do this for the summary table (results) in current schema. sjcarbon: ways of making the searches return faster hitz: a "view" (and please correct!) is just a fake table in the database which is the result of a select statement - in our case joining across tables. rama: I see hitz: You are basically trading memory/disk for CPU time sjcarbon: chris and I have already talked about this hitz: I have to confess, I have never actualy MADE a view. j-lo: so this is changes to the actual database itself? sjcarbon: no sjcarbon: it adds a 'fake' table rama: this table is created on the fly i think sjcarbon: no it is not j-lo: so this would be for AmiGO 2.0 really? sjcarbon: not in mysql anyways sjcarbon: i think it could be used for both sjcarbon: i'll try making views this week gwg: why don't we just create views for the various go db objects now? I think it would make life much easier hitz: it's not really a schema change, but they do have to be created in the database; presumably at loading hitz: I think this should be an action item for SW group to figure out who and when sjcarbon: sounds good gwg: ok hitz: chris? comments? sjcarbon: i think he's mulitasking with a phone call right now too 9:00 AM j-lo: so the action item is for the SW group to look into creating views for the database? j-lo: and report back j-lo: ? rama: Is it the plan to worry about the format/details of a table on the webpage later? sjcarbon: i'll email the list and see if i can get it added to the agenda j-lo: Rama, yes I think so j-lo: Seth - great j-lo: Okay, anything further on IEAs, or shall we move on? rama: okay rama: Download option? rama: Will that also be solved with the DB 'views'? sjcarbon: no, but it may be speeded up j-lo: which download option Rama? rama: Having a button to download 45 pages of annotations to your desktop as a excel file (maybe) 9:05 AM sjcarbon: I don't think Excel is on the list, but I'm working on adding new filters and fixing some of the older ones sjcarbon: right now I'm adding GO association rama: i gave excel as an example. tab delimited will work too j-lo: we already have this in both the AmiGO 1.0 and 2.0 plans I think j-lo: shall we move on to the boolean searching? hitz: excel is just comma-delimited (.csv) hitz: and of course, excel can read tab-delim if user has 1/2 a clue rama: I agree Ben sjcarbon: GO assoc is tab-delimited hitz: I don't think Go assoc is a useful format for anything other than GOA. hitz: BUT maybe it's the easiest way to handle NOTs/etc. eurie: But we've already defined the GO assoc file so why not use it instead of generating a new file format hitz: there are extra columns which will make little sense to naive juser 9:10 AM rama: I agree with eurie. It is easy to dela with the GO-associations file. cjm has quit the server saying: Read error: EOF from client hitz: good points. question is: is loading a GOA file into excel useful? hitz: and basicaly we are using amigo as a very overwritten "grep" eurie: sure someone might download the whole GOA file but a download option allows someone to only download aspergillus annotations without having to download all of GOA. eurie: espeically those users who don't know how to grep eurie: this is the audience we need to be meeting cjm has joined the channel eurie: this is the audience folks in the outreach group have been trying to get involved in submitting annotations rama: yes, thanks eurie hitz: ok I'm convinced. and I am resisting comenting on the download thing eurie: why go through the effort of getting groups to submit annotations if they have to know "grep" to see them? j-lo: absolutely - we need this feature 9:15 AM j-lo: Right, lets stay on target people - we have 15 minutes to discuss boolean searchinh j-lo: g j-lo: so, how have people found the boolean searching j-lo: ? rama: I am assuming download option is going to be looked into? gwg: @rama: yup j-lo: (rama - downloading is a big feature - we will definitely do it) rama: okay. thanks j-lo: Seth - does the boolean search allow wildcards? sjcarbon: not in the middle of words sjcarbon: that wouldn't be hard to add though rama: cool j-lo: that was the only thing i found lacking sjcarbon: i was mostly concerned with the grammar, sjcarbon: but I'll add that and repost it 9:20 AM j-lo: I found the grammar pretty intuitive sjcarbon: great! j-lo: any other comments on that? sjcarbon: i assume '*' is an unused character and therefore safe for me to use as a wild card? rama: % is also used gwg: presumably you wrote the parser from scratch for the boolean search? sjcarbon: yes sjcarbon: simple and reusable--already in a module gwg: cool j-lo: so Seth, you're going to send round another email about the boolean grammar did you say? 9:25 AM sjcarbon: after i add the wildcard eurie: sounds good. i haven't had a chance to test it sjcarbon: no problem--be aware that it is slow for testing purposes though j-lo: great, we can discuss next time then j-lo: anything else anyone wants to bring up? hitz: not it j-lo: okay, I'll post some notes about this meeting on the wiki j-lo: and see you all next time sjcarbon: see ya! cjm: bye! rama: thanks. bye cjm has quit the server saying: Read error: EOF from client eurie: bye! gwg: cyas donghui has left the channel eurie has left the channel pascale has quit the server