| --- Log opened Fri Dec 13 00:00:37 2019 |
| --- Day changed Fri Dec 13 2019 |
| 00:00 | | * TheWatcher needs to look up how to configure visual studio's syntax highlighting colours before this setup drives him nuts |
| 00:02 | <&McMartin> | I think there's some way to do that with JSON files or something but I haven't dug into it |
| 00:10 | <@TheWatcher> | Looks like Tools -> Options -> Environment -> Fonts and Colors will do it, just needs me to set 'em all |
| 01:19 | | celticminstrel [celticminst@Nightstar-2kik29.dsl.bell.ca] has joined #code |
| 01:19 | | mode/#code [+o celticminstrel] by ChanServ |
| 01:24 | | celticminstrel [celticminst@Nightstar-2kik29.dsl.bell.ca] has quit [[NS] Quit: KABOOM! It seems that I have exploded. Please wait while I reinstall the universe.] |
| 01:37 | | Degi [Degi@Nightstar-tcpka9.dyn.telefonica.de] has quit [Ping timeout: 121 seconds] |
| 01:38 | | Degi [Degi@Nightstar-ol7381.dyn.telefonica.de] has joined #code |
| 01:38 | | Vornicus [Vorn@ServerAdministrator.Nightstar.Net] has quit [Connection closed] |
| 02:32 | | celticminstrel [celticminst@Nightstar-2kik29.dsl.bell.ca] has joined #code |
| 02:32 | | mode/#code [+o celticminstrel] by ChanServ |
| 02:43 | <&Derakon> | Hrm. Trying to figure out why the gun is consistently aiming well ahead of the airplane. Muzzle velocity = 366, angle is 16 degrees (so flat velocity is 351), target is at (741, 36, 492), gun is at (526, 3.4, 500), flat distance is 215, claimed travel time is .61s. That math checks out...but when I step the game it takes ~16 ticks, each of which is .0167s, for ~.27s, for the projectile to reach the point it's aiming at. |
| 02:44 | <&Derakon> | ...fuck, because my firing code is saying "okay, take the aim vector and multiply it by the launch velocity". |
| 02:44 | <&ToxicFrog> | And the aim vector is not a unit vector? |
| 02:44 | <&Derakon> | Wait, it should be. |
| 02:45 | <&Derakon> | Argh. |
| 02:46 | <&Derakon> | Yeah, normalizing it didn't help. :\ |
| 03:13 | <&Derakon> | Hm, turns out I'm off by exactly half. |
| 03:15 | <&Derakon> | I have a scaling factor that I apply to the muzzle velocity to convert from real-world meters to some made-up in-game scale to help things feel reasonable, which happens to be .5. |
| 03:15 | <&Derakon> | I'm applying it in this case already...but maybe I need to apply it twice? *shrug* |
| 03:25 | <@Alek> | o_O |
| 03:27 | <@Alek> | hope your airplane flies better than Ace Combat 7. >_> |
| 03:27 | <@Alek> | https://i.imgur.com/aJUunlp.mp4 |
| 03:42 | <&Derakon> | Ah...ha |
| 03:43 | <&Derakon> | https://twitter.com/byobattleship/status/1205279359004577792 and its replay have some vi deo. |
| 03:43 | <&Derakon> | video. |
| 04:15 | | VirusJTG [VirusJTG@Nightstar-42s.jso.104.208.IP] has quit [Connection closed] |
| 04:15 | | VirusJTG [VirusJTG@Nightstar-42s.jso.104.208.IP] has joined #code |
| 04:15 | | mode/#code [+ao VirusJTG VirusJTG] by ChanServ |
| 04:22 | <@Alek> | heh |
| 04:39 | | VirusJTG [VirusJTG@Nightstar-42s.jso.104.208.IP] has quit [Connection closed] |
| 04:41 | | VirusJTG [VirusJTG@Nightstar-42s.jso.104.208.IP] has joined #code |
| 04:41 | | mode/#code [+ao VirusJTG VirusJTG] by ChanServ |
| 04:45 | | VirusJTG [VirusJTG@Nightstar-42s.jso.104.208.IP] has quit [Connection closed] |
| 04:46 | | Vorntastic [uid293981@Nightstar-6br85t.irccloud.com] has joined #code |
| 04:46 | | mode/#code [+qo Vorntastic Vorntastic] by ChanServ |
| 04:49 | | VirusJTG [VirusJTG@Nightstar-42s.jso.104.208.IP] has joined #code |
| 04:50 | | mode/#code [+ao VirusJTG VirusJTG] by ChanServ |
| 05:16 | | Derakon is now known as Derakon[AFK] |
| 05:56 | | catalyst [catalyst@Nightstar-v6lb30.cable.virginm.net] has joined #code |
| 06:03 | | Pinkhair [user1@Nightstar-g7hdo5.dyn.optonline.net] has joined #code |
| 06:04 | | Pink [user1@Nightstar-g7hdo5.dyn.optonline.net] has quit [Ping timeout: 121 seconds] |
| 07:56 | <&[R]> | "they still taught programming in cobol until about 2006." |
| 07:56 | <&[R]> | Those poor souls |
| 07:58 | < catalyst> | sign of a successful language |
| 08:01 | <&McMartin> | Yeah, and it didn't die either |
| 08:01 | <&McMartin> | Though it has evolved with the technology it is paired with over the decades |
| 08:02 | <&McMartin> | And it's gotten literal decades of optimization work at this point |
| 08:02 | <&McMartin> | It probably is genuinely the fastest batch data processing language out there right now and I bet it's got better safety guarantees than C too. |
| 08:04 | <@TheWatcher> | For some inexplicable reason, I suddenly have a certain rhyming couplet in my head |
| 08:04 | <&McMartin> | Do tell |
| 08:04 | <@TheWatcher> | The one attributed to a certain mad Arab. |
| 08:05 | <&McMartin> | Ah. |
| 08:05 | <&McMartin> | # kill -9 $death_pid |
| 08:05 | <@TheWatcher> | Quite. |
| 10:06 | <&McMartin> | The obvious result is endless gibbering about forbidden secrets |
| 10:06 | <&McMartin> | To wit: https://www.youtube.com/watch?v=zwZISypgA9M |
| 11:55 | <&Reiver> | People give old languages too much shit |
| 11:55 | <&Reiver> | If it is in active use and development, it's almost like it might have something going for it |
| 11:56 | <&Reiver> | ISTR that a whole bunch of transaction systems effectively run a naieve insertion-sort on their data in, yeah, pretty much COBOL |
| 11:56 | <&Reiver> | This seems like a terrible idea until one realises that /most of the data is already sorted/ |
| 11:57 | <&Reiver> | By dint of having been previously sorted, and representing decades of data. |
| 11:57 | <&Reiver> | And Insertion Sorts run at effectively linear time, if the list is pre-sorted, so... why not? |
| 11:57 | <&Reiver> | Clever algorathms would just slow it all down. <g> |
| 12:11 | <@sshine> | McMartin, I think any value assigned to the variable $death_pid deserves to be fed to a command called `kill'. |
| 12:12 | <@sshine> | McMartin, are you saying that there's a thing called modern cobol? |
| 12:22 | < catalyst> | to be fair a very common way of doing sort is to quicksort until it becomes easy enough to do the last bit with something more naive but faster if your data's not degenerate |
| 12:32 | <@TheWatcher> | Reiver: in my case it's not so much old languages - it's just, well, I find COBOL to be awful >.> |
| 12:32 | <&Reiver> | TheWatcher: Aye, well, can't argue with that~ |
| 12:32 | <&Reiver> | catalyst: ha, yes, I've heard that one - when the theoretical speed is lost out due to algorathmic speed, yes? |
| 12:33 | <~Vorntastic> | So |
| 12:34 | <~Vorntastic> | There's a new multiplication algorithm that someone's been working on |
| 12:34 | <&Reiver> | uh oh |
| 12:34 | <~Vorntastic> | It has theoretically optimal performance |
| 12:35 | <&Reiver> | ... what |
| 12:35 | <~Vorntastic> | But its k is so high that the numbers that it works faster on than existing algorithms are larger than can be theoretically be stored in the entire universe |
| 12:42 | < catalyst> | I bet there's a theoretical computer scientist out there who published that who's thinking 'AHA I got them now! There's no way this can ever be practically useful in the real world!' |
| 12:43 | <@TheWatcher> | I wonder if you could get it to use ↑ |
| 12:43 | <@sshine> | Vorntastic, what's the timeline here? I feel like I've heard this before within the last 10 years. |
| 12:44 | <@sshine> | ohhh |
| 12:44 | <@sshine> | https://www.quantamagazine.org/mathematicians-discover-the-perfect-way-to-multiply-20190411/ |
| 12:44 | <@sshine> | 2019 |
| 12:50 | <~Vorntastic> | Yeah this is the third improvement in like 15 years, this one down to the barrier for this general technique, but it doesn't help in any practical way |
| 12:50 | <~Vorntastic> | Yet |
| 13:35 | | Kindamoody[zZz] is now known as Kindamoody |
| 13:56 | | celticminstrel is now known as celmin|away |
| 15:06 | | Vornicus [Vorn@ServerAdministrator.Nightstar.Net] has joined #code |
| 15:06 | | mode/#code [+qo Vornicus Vornicus] by ChanServ |
| 16:43 | | Vornicus [Vorn@ServerAdministrator.Nightstar.Net] has quit [Connection closed] |
| 16:56 | | Vorntastic [uid293981@Nightstar-6br85t.irccloud.com] has quit [[NS] Quit: Connection closed for inactivity] |
| 17:01 | | Emmy [Emmy@Nightstar-9p7hb1.direct-adsl.nl] has joined #code |
| 20:02 | | Derakon[AFK] is now known as Derakon |
| 20:46 | <&McMartin> | https://github.blog/2019-12-10-multiple-git-vulnerabilities-in-2-24-and-older/ |
| 20:46 | <&McMartin> | Mostly but not entirely on Windows clients, but if you use Git on Windows, update it, folks |
| 21:49 | <@sshine> | I just realize I have a pretty cool data analysis task at hand. |
| 22:16 | <@sshine> | my employer has this "check your salary" page where you can enter some job title, and it produces an average of what people who say they have that jobtitle say they earn. |
| 22:16 | <@sshine> | so it's a bit speculative, but very entertaining. |
| 22:17 | <@sshine> | the search engine is very crap because it was made in an inefficient way. |
| 22:17 | <@sshine> | since it was made we now run Apache Solr, and all the available data is in there |
| 22:17 | <@sshine> | so I can improve the precision of the queries greatly (even though they're still speculative) |
| 22:18 | <@sshine> | so it involves taking a bunch of CVs and sorting them by how much they resemble the ad-hoc CV parameters in the search query (job title, geo area, educational level, age, experience level) |
| 22:19 | <@sshine> | the current sorting mechanism has some hardcoded weights it uses to calculate a distance score. |
| 22:19 | <@sshine> | those weights were determined experimentally by a currently broken training module. |
| 22:21 | <@sshine> | so I thought I'd recreate those weights... I bet they're off after 10 years of having been hardcoded. |
| 22:22 | <@sshine> | also, I was thinking, maybe having hardcoded weights isn't necessary... |
| 22:23 | <@sshine> | right now, for example, "age" has a fixed weight across all jobtitles. but maybe age doesn't affect pay equally for all jobtitles. |
| 22:23 | <@sshine> | really, probably not. |
| 22:37 | | Kindamoody [Kindamoody@Nightstar-eubaqc.tbcn.telia.com] has quit [Client exited] |
| 22:38 | | Kimo|autojoin [Kindamoody@Nightstar-eubaqc.tbcn.telia.com] has joined #code |
| 22:38 | | mode/#code [+o Kimo|autojoin] by ChanServ |
| 22:39 | | Kimo|autojoin is now known as Kindamoody |
| 23:23 | < Emmy> | sshine, sounds fascinating. is that specific to your employer, or does it contain reference data across a group/country? |
| 23:27 | <@sshine> | Emmy, it's data entered into the CV section on my employer's site. |
| 23:28 | <@sshine> | Emmy, and yes! I think it's a very nice task because my boss expects me to "just speed up the current thing" (replace SQL calls with Solr calls). I've done that in 10 minutes, so anything beyond that is just indulgence. |
| 23:32 | <&Reiver> | Solr calls? |
| 23:32 | <@sshine> | so I'm toying with the idea of not having fixed weights for each dimension, but rather, for every jobtitle in a query, (1) calculate mean/dev for each dimension, (2) determine some threshold at which point to discard outliers for any one parameter, (3) return average. so no sorting at all |
| 23:32 | <@sshine> | Reiver, Solr search queries? |
| 23:32 | <&Reiver> | Don't know them, probably should |
| 23:32 | <@sshine> | Reiver, I think Solr is kinda like Elasticsearch. but I'm very ignorant about this. |
| 23:33 | <@sshine> | at least, they both claim to be based on Apache Lucene... |
| 23:33 | <&Reiver> | I see |
| 23:33 | <&Reiver> | I know even less about that too >_> |
| 23:33 | <&Reiver> | Is it SQL-like, or a totally different set of syntax? |
| 23:33 | <@sshine> | urm |
| 23:34 | <@sshine> | the search syntax is more like what you'd expect from an over-engineered search engine query box... |
| 23:34 | | * Emmy shrugs, is totally stuck in the kiddy-pool with Jet- and T-SQL |
| 23:35 | <@sshine> | I just know that if I fetch 10.000 objects via the ORM/SQL, that takes time, whereas fetching them from Solr doesn't. I don't know why, I guess it's partly NoSQL hype, and partly some full-text search algorithms that beat SQL 'LIKE' by a cybertruck's length. |
| 23:35 | <&Reiver> | oh, yeah, LIKE is awful |
| 23:36 | <&Reiver> | I understand /why/, but it does not negate the awful~ |
| 23:36 | < Emmy> | anyway. naptime. time to dream about code that just works on first drafts and queries that never time-out |
| 23:37 | < Emmy> | Ya know, totally unrealistic, dreamlike, idealist stuff |
| 23:37 | | Emmy [Emmy@Nightstar-9p7hb1.direct-adsl.nl] has quit [Connection reset by peer] |
| 23:39 | <@sshine> | what we do right now is: get a bunch of CVs, sort them by distance to the search query CV, discard "outliers" (hardcoded for each dimension), and use only the 20 closest CVs for generating the average/histogram. |
| 23:40 | <@sshine> | I'm thinking: use better outlier discard method (discard top and bottom percentile of every query), and use every remaining CV in result, making sorting superfluous. |
| 23:40 | <@sshine> | wait. no. |
| 23:40 | <@sshine> | arf. sorry, I'm just in full rubberduck mode right now. |
| 23:42 | <@sshine> | I can't. the CVs I get are only filtered for jobtitle, so if someone put in "20 years of job experience", I really want to exclude the resumes with "2 years" listed. |
| --- Log closed Sat Dec 14 00:00:48 2019 |