[06:09:25] Hi josephine_l niedzielski :-) [06:09:39] Hi Nicolas_Raoul! [06:10:18] I don't know if niedzielski is around :) [06:11:25] We can start first though maybe? [06:14:05] Sure! [06:14:23] #35 is a big problem, thanks for finding it! [06:14:31] Nicolas_Raoul, are we okay with the latest week's task? Can we move on to the next one (caching)? [06:14:41] Haha, thanks :) [06:15:23] Why not! [06:15:46] Java Maps are misnamed in that they operate on 1 dimension [06:15:59] Nicolas_Raoul, yeah. But dictionary is obsolete I think [06:16:15] for geographical data I am sure some third-party library is available [06:16:34] Nicolas_Raoul, hmm, okay. Can we clarify what exactly we mean by the area->category dictionary? [06:16:53] Nicolas_Raoul, is it to reduce calls to the API? So if we find the area in the cache we just suggest the same categories again? [06:18:55] Yes, for instance if I upload a picture at 12.34,67.89 an API call is made, and then I upload a picture at 12.33,67.88 (I mean, very near to the previous picture) it should pick from the cache. [06:19:25] It can be expected that people will often upload several pictures showing different angles of the same place. [06:20:37] Nicolas_Raoul, ah, true. So I should look for a third party library to be able to map a range of latitude/longitude to a value [06:21:15] And call that instead of the API if the GPS coords are similar [06:21:33] Can we do the storage at the same place as 'recent categories'? [06:21:39] Or is that too long-lasting? [06:21:53] with a certain tolerance, maybe 100 meters or something? [06:22:06] Right [06:22:31] I haven't thought about it much, but same sounds good. New categories don't get created that often. [06:23:31] In your tests, how long did the API call(s) take for each picture, on average? [06:23:39] Yeah, the recent categories are stored in a custom Content Provider I think. Hopefully I can just add a column with the latitude/longitude. Not sure how to handle the range/tolerance though [06:23:43] Hmm [06:24:09] I didn't save the logs. IIRC about 4 sec [06:24:23] But NZ internet is bad also [06:24:46] For instance to load the average website it takes 4 sec also [06:27:22] This looks like a third-party library we could use: http://blog.javascripting.com/2014/10/23/geo-tree-a-library-for-fast-searching-by-latitude-and-longitude/ It stores lat,long points and you can ask it to give you all points contained in a lat1,long1-lat2,long2 rectangle. [06:28:03] Nicolas_Raoul, can we use a JS library with Android? [06:29:44] haha, that came up for a Google search on "java area data structure latitude longitude" but it is not Java indeed! You are right we need a Java library, I am sure there are equivalent Java ones :-) [06:30:39] Nicolas_Raoul, haha, yeah. I will try and look for a similar one in Java [06:30:45] Is 4 seconds for Method D? That method potentially involves several API calls, do several calls also make 4 seconds? [06:31:43] Oh, I haven't implemented the extra calls yet, just the one call. I thought we would do that after all the major tasks if there is time [06:31:49] Should I work on that instead? [06:34:12] Yes, that's the important thing to do now, before caching. You see what needs to be done, right? [06:34:52] Nicolas_Raoul, count the results received, if there is <10 then do another call, etc? [06:35:28] I think we should limit how many times it recurses? [06:37:09] How about we don't limit, and see what happens for various latitude? For instance a pic in Siberia will probably recurses several times before it catches something, but that something is probably quite on-topic, still. [06:38:23] The odd thing is that even though we are using a separate Volley thread, the app still waits for the response [06:38:37] So they don't get to the next screen until the API calls finish [06:38:59] If there were >3 calls they might be waiting a long time [06:45:19] Nicolas_Raoul, besides, the API has a limit of 10,000m [06:49:22] I guess the "long time" will be the same for a small area containing many items? Let's try. niedzielski should know better than me about Volley, and that can be done a bit later so just go ahead with Method D implementation :-) [06:50:02] Okay, sure. :) But I will have to limit it at 10,000 m otherwise the API will return an error [06:50:36] I'll ask niedzielski about the threading issue later. I think we might be able to use a Service to do the calls, but that will take quite a bit of refactoring [06:56:44] Nicolas_Raoul, I'm a bit concerned about the deadlines also. I will be away til Friday so I will only have Sat and Sun left to work on this week, and then the caching task looks quite complicated. I hope I will be able to finish the additional task in time [06:59:54] I will look for an appropriate library, so that no time is wasted using an inappropriate library. [07:01:11] Nicolas_Raoul, okay, thanks. :) [07:01:42] Nicolas_Raoul, do we want the increments for radius to be 10x each recursion? [07:03:14] Sounds good yes [07:05:03] Nicolas_Raoul, okay. I'll work on that when I get back from my travels this week. [07:05:32] I think niedzielski said he will be away til Jan 9, so I'll try to email him about the Volley issue if needed [07:30:01] Nicolas_Raoul, shall we call it a day? [07:31:02] OK, contact me anytime! [07:32:22] Nicolas_Raoul, thanks! Have a nice day :) [07:32:47] Good evening! :-) [09:12:21] hola [09:18:59] hi phuedx ! [09:22:15] hey joakino [09:22:17] how're you? [09:23:06] fine! last work day before SF :p and you? [09:27:32] i'm watching joelspolsky's talk, very interesting [09:28:35] i've got two days before sf [09:28:53] joakino: what're you working on? [12:52:18] heading out for lunch [14:01:16] back [16:58:07] hi folks [17:06:10] hey joakino [17:07:52] hi bmansurov ! [17:12:25] \o [17:55:13] hi jdlrobson [17:55:30] im playing on the slides. wwweeeeee [17:56:27] phuedx: there? [17:57:43] yup [17:57:52] just helping the kids with the pre-bed ritual [17:59:47] dr0ptp4kt: ^ i might be a few minutes late [18:00:16] phuedx: k [18:03:01] dr0ptp4kt: h-o issues [18:04:17] dr0ptp4kt: or has nobody joined yet? i'm the only one in the room [19:17:45] dr0ptp4kt: so what's the next steps? [19:18:27] dr0ptp4kt: i've moved my slide to the end but gwicke i'm not sure how you want to incorporate it [19:18:43] dr0ptp4kt: i need to work out what to do about demos but showing similar demos twice seems problematic [19:37:23] jdlrobson: lemme watch the videos, and i'll convey my game day thinking on monday :) . i agree too many videos doesn't add value [19:37:33] dr0ptp4kt: but presentation is at 10am right? [19:37:41] dr0ptp4kt: ummmmm [19:37:46] jdlrobson: ummmmmmm [19:38:04] and dev summit starts at 9 [19:38:29] jdlrobson: yeah, presentation is at 10 pacific. welcome & keynote 9-9:50 [19:38:33] but i think the slides around the mission statement and video should be right before or inside the performance section [19:38:53] jdlrobson: I wonder if the pointer slide would best fit in the wrap-up after the discussion [19:38:58] it feels awkward in current form - that's only thing i'm worried about my stretch of the presentation :) [19:39:11] jdlrobson: what is "pointer slide"? [19:39:22] pointers to other relevant sessions [19:39:29] gwicke: i already moved it to end [19:39:33] er, gwicke :) [19:39:34] do what you want with it [19:39:40] sounds good [19:39:41] but i think we should plug sessions [19:39:44] to continue conversaton [19:39:46] +1 [19:39:50] i'll do the mission statement slide [19:39:58] the video, move that to the perf section [19:40:06] (or videos) [19:41:31] dr0ptp4kt: it would be good to talk more generally then about the statement [19:41:46] ill move the video into my section [19:42:03] jdlrobson: cool. yes, i'll keep it pretty general. dramatic pauses and all that [19:43:02] dr0ptp4kt: it would be nice if the intro & summary both stressed the same points [19:43:19] gwicke: roger. preview, review. [19:43:25] basically framing the discussion in the middle part [19:43:40] now that i've heard you guys present, it's going to be easier for me, too :) [19:43:41] *nod* [19:43:43] halting problem [20:46:13] mdholloway: did you see my gerrit comment? That's the only thing left. The rest looks good to me. [20:46:33] bearND: great! will look now [20:47:32] great [20:50:33] bearND|brb: for the function name, how about rmSpanTags? [20:50:46] bearND|brb: maybe with a short comment [20:52:12] I would replace the RM part, too, since it's too similar to the other RM methods but it's doing something different [20:58:57] bearND|brb: maybe i'll use inlineSpanText as you suggested, with a comment [21:38:08] jdlrobson: I think that reading-web-staging.eqiad.wmflabs instance is sick. I can't ssh in either and we are both project admins there [21:38:20] * bd808 will see if Yuvi can provide some help [21:39:40] jdlrobson: ah ha! the instance name is -- reading-web-staging-2.reading-web-staging.eqiad.wmflabs [22:02:07] jdlrobson: able to make the phlogiston meeting? [22:51:46] bearND: \o/ wanna merge https://gerrit.wikimedia.org/r/#/c/257695/ as well? [22:52:50] mdholloway: we haven't deployed the service patch yet [22:53:04] bearND: good point, getting ahead of myself with excitement. [22:56:17] i'm excited about it, too. We need to check with Marko and gwicke to have them add the new definition route to RB. We'll probably also need to discuss the endpoint for this. I've stated a discussion about this on https://phabricator.wikimedia.org/T119235 but have not received a reply. [22:56:26] started [23:00:12] bd808: sorry i missed that message. [23:00:14] thanks for sussing it out [23:00:22] jdlrobson: np [23:01:31] dr0ptp4kt, jdlrobson: doing something like this once or twice a quarter doesn't seem that bad -- https://github.com/wikimedia/phlogiston/pull/1/files [23:02:02] high level goals shouldn't come and go willy nilly anyhow [23:02:12] yeh it's worth a go. i just doubt it's going to catch "distractions" accurately [23:03:52] yeah. that is hard to do when collaboration is what is pulling against the other work [23:04:38] when we did tracking in Oct/Nov for reading infrastrcture we tracked hours worked instead of tasks to get realistic numbers [23:04:53] but that was a different kind of pain in the butt