00:00:09 kmc: Right, it's an actual change. 00:00:13 kmc: Well, it's a relatively trivial change, requires no changes to the language itself, and GHC does it... 00:00:15 * shachaf has yet to figure out how to read that smiley. 00:00:17 this is "GHC rejects some programs which should be accepted" 00:00:18 what's the name of this new extension we're discussing? 00:00:25 kallisti, it doesn't have a name, that's the problem 00:00:27 kallisti: -X 00:00:36 Defeated Monster X! 00:00:38 (That's -X followed by a segmentation fault.) 00:00:39 ...wat 00:00:48 the "extension" is that GHC removed the Eq and Show superclasses on Num 00:00:53 regardless of whatever -X flags you specify 00:01:09 so it's *not* standards-compliant and code written for Haskell 98 or Haskell 2010 will break 00:01:14 and this has already been observed in standard libraries 00:01:33 now, I think it might be technically difficult to make this controllable by an -X fla 00:01:35 but it's way better 00:01:36 so cool 00:01:41 good job GHC 00:02:08 cause it's bad if some code writes an instance for base:Prelude.Num and other code needs an instance for haskell2010:Prelude.Num 00:02:34 but I think this suggests we should live with the problem 00:02:41 which after all people have been working around for years 00:02:44 it's not hard to work around it 00:03:08 kallisti, I'm worried that in a few years, GHC will have fifteen divergences from standard Haskell which can't be disabled 00:03:18 kmc: What's the practical difference between GHC springing a breaking change on everybody all in one go, and Haskell 201X doing so? 00:03:26 and are undocumented, or mentioned in passing in obscure paragraphs all over the manual 00:03:36 elliott, ^^ 00:03:55 I don't think it's typically a good idea to perpetuate an existing flaw, building an ever-increasing base of code that will break if any changes occur, all for the sake of "not breaking code" 00:03:57 kmc: Well, I think that's a bit of a slippery slope argument. 00:04:07 elliott, some slopes are slippery, though 00:04:08 shrug 00:04:12 i think i've made my case 00:04:15 kmc: Are you sure it isn't already in the docs? 00:04:15 i don't need to say it over and over 00:04:21 elliott, I didn't say it isn't 00:04:31 Okay. 00:04:48 it might be "nobody's problem" since it's technically in a library, not GHC itself 00:04:53 and i mean how many people read the haddocks for Prelude.Num 00:05:14 actually what code would that break. 00:05:15 you should not have to read the library documentation for the entire standard library to know the ways it differs from the published spec 00:05:24 removing the superclass constraint on Num 00:05:25 kallisti, if you have «f :: (Num a) => ...», and you use Eq or Show constraints within f 00:05:34 ah 00:05:38 and it's not "would", this has already been observed in standard libraries, which need patching to compile under GHC HEAD 00:05:47 the solution of course is to never use type signatures in your code. :> 00:05:51 this ensures future compatability. 00:05:55 hehe 00:05:56 yes 00:06:06 I think I would care a lot more about GHC being standards compliant if the Haskell standards process actually moved at a reasonable pace and added in widely-used extensions. 00:06:18 there should be something like a type signature that's ignore. oh wait that's just a comment. 00:06:24 Right now, I don't think it's practical to try and get a breaking change in through the standard at all. 00:06:25 from now on I'm going to comment out all of my unecessary type signatures 00:06:36 to prevent my code from breaking due to an explicit type signature that later needs to be changed. 00:06:41 elliott, if GHC produced a Haskell Report-quality specification of the language they implement, I think it would remove most of my objections 00:06:46 but they don't 00:06:55 not that the Haskell Report is super high quality either 00:07:19 I think the Report just needs to be less conservative, and come out a lot more often (I think they're already trying to do the latter). 00:07:35 but it's much closer to a language description than Section 7 of the GHC Manual 00:08:33 i just wanna know what language i'm using 00:08:39 I think that change is simple enough that you could literally just write a script that fixes all the broken code. 00:08:40 even if it's specified unilaterally by some compiler writers 00:08:45 kallisti, do it 00:10:19 -!- Jafet has quit (Ping timeout: 252 seconds). 00:10:59 to prevent my code from breaking due to an explicit type signature that later needs to be changed. <-- ah but there have _already_ been ghc changes that required you to put signatures where you didn't before (local monomorphism when type families are used) 00:11:17 *didn't need them 00:11:21 oerjan: I believee I said unnecessary type signatures 00:11:27 as in the ones that aren't needed 00:11:29 for anything 00:11:35 kallisti: they _weren't_ necessary before that change 00:11:51 okay. 00:11:55 that's not what I'm saying though 00:12:15 at least those changes only apply if you're already using an extension 00:12:27 kmc: there's the Easy but maybe Bad way and then the Technically Correct in EVery Circumstance But Complicated Way 00:13:06 kmc: the easy way would be to substitute every type signature that contains a Num instance with the output from GHC's type inference command 00:14:29 but that's not always the correct thing to do. 00:15:37 i didn't ask how to do it 00:15:38 i said you should do it 00:16:03 i would think the easy way is to translate the context (Num a, ...) to (Num a, Eq a, Show a, ...) everywhere 00:16:34 no. 00:16:36 well 00:16:40 if it's all old code 00:16:41 then yes. 00:16:48 yeah, that was my assumption 00:16:58 in #haskell it's often declared that X isn't a problem because you could just write library/tool Y 00:17:04 but most of those libraries/tools never get written 00:17:36 I thought that was what they say in #perl 00:17:39 because most of the effort is all the "trivial engineering" of actually writing the damn code, debugging it, writing a test suite, documenting, packaging, publishing, etc 00:17:47 oh wait 00:17:51 no in #perl it's s/write/use/ 00:18:31 kmc: so "trivial" 00:18:36 * oerjan swats kallisti for dangerous play with fire -----### 00:18:58 anyway if I ever have a programming language that people use 00:19:02 the way I'll prevent breaking code 00:19:11 is to just make sure every change has an associated script that fixes all the old code. 00:19:20 so hard 00:19:44 either that 00:20:33 or I'll have some insane system whereby you can specify versions of the language and allows a module to be executed as though it were in the past PERFECTLY WITH NO DIFFERENCES AT ALL. 00:20:37 should be trivial to implement 00:20:54 haha 00:21:06 well, it would be reasonable with like... "normal" code 00:21:11 yes "trivial" 00:21:20 "trivial" is just a douchebag way to say "easy" 00:21:28 low-level code would of course be pretty much impossible to accomodate in such a way 00:21:33 except the things declared to be "trivial" aren't even easy, they're just theoretically uninteresting 00:21:48 a problem is "trivial" if it would take 20 people 2 years to solve that problem, but they won't get a paper out of it 00:22:05 Perl JIT? trivial 00:22:09 (this is my life goal) 00:22:11 -!- rdrp has joined. 00:22:11 right exactly 00:22:31 hello #esoteric would you like to be my 20 people and help me develop a perl 5 JIT? 00:22:45 it would basically be the best thing ever. 00:23:05 it would be "okay perl is efficient now I can stop programming in everything else" 00:23:17 (not really but it would be nice) 00:24:22 I wonder what makes LuaJIT so fast. 00:24:24 Python is much closer to having a usable JIT 00:24:32 which is that? 00:24:33 kallisti, I saw a detailed writeup of that once... but I can't find it anymore :/ 00:24:37 kallisti, PyPy 00:24:40 anyone wants to build an interpreter or a virtual machine for SumaMoito (see esolangs wiki) 00:24:42 oh right that thing. 00:24:50 a perl 5 JIT would be theoretically interesting because it's probably almost impossible 00:24:55 rdrp: do you? 00:25:01 i doubt it's impossible 00:25:12 i mean, why would it be 00:25:15 elliott: well you would need a normal interpreter as well as a JIT compiler. 00:25:19 kmc: well, perl 5 is pretty strange :) 00:25:22 i know 00:25:25 yes, but i am not an expert 00:25:32 in that codes 00:25:39 rdrp: sounds like a learning opportunity :) 00:25:47 i think atrapado mentioned having an implementation 00:25:53 yes 00:25:59 i am atrapado 00:26:35 oh :) 00:26:42 i left the laptop in the other room 00:26:57 atrapado, meet rdrp. rdrp, atrapado 00:27:02 elliott: I think you would have to basically interpret stuff like source filters and BEGIN blocks because they're necessary at compile-time 00:27:11 :) 00:28:53 I'm pretty sure this is what the perl interpreter does. before compile-time finishes (or even begins, in the case of BEGIN blocks) it has to execute Perl code. 00:29:08 well maybe i build something if i have time and motivation 00:31:33 hope it collapses 00:31:57 elliott: maybe one day Perl 6 will have a JIT. after they finish defining the language's syntax almost entirely in Perl 6. 00:32:36 i expect partial optimizations, even optimizations 00:32:41 kallisti: WTF do you think Parrot is? 00:33:06 elliott: an interpreted bytecode language? 00:33:37 http://esolangs.org/wiki/Parrot of course 00:35:01 the example resembles Python way more than it resembles Perl. 00:35:31 but it uses <> and $_ !!!! 00:36:22 ah good they renamed "Parrot magic cookie" to something less stupid sounding. 00:36:33 now it's "polymorphic container" 00:39:15 but parrot magic cookie is a good name 00:40:26 is parrot actually good? 00:40:31 I don't know its isntructions 00:41:13 kallisti: polly morphic cracker 00:42:02 they should suddenly switch from Parrot to LLVM 00:42:05 it would be such a good idea 00:42:09 trivial to implement too 00:44:15 elliott: so you have your very own reduceron yet? 00:45:41 I'm guessing it's the target platform for @ 00:50:08 no @ uses a nanobot network http://xkcd.com/865/ 00:50:50 trivial 00:50:59 just like google's failed llvm jit for python was trivial 00:52:15 -!- Nisstyre has quit (Ping timeout: 240 seconds). 00:57:31 kmc: Trivial trivial trivial. 00:57:41 elliott: an interpreted bytecode language? 00:57:52 kallisti: Ah, it's one of those things where you're wrong. 00:57:59 lols 00:57:59 ??? 00:58:38 * elliott wants a tool that's like LMGTFY, but also automatically clicks on an article linked to in the results and highlights a certain phrase. 00:58:45 In this case it would be Parrot's Wikipedia article. 01:00:04 so what is hoogle? 01:00:11 let me hoogle that for you 01:00:16 i can tell it means haskell google.. but that doesn't quite 01:00:20 @hoogle hoogle 01:00:21 Can't think of anything more interesting to search for? 01:00:21 package hoogle 01:00:36 does it search haskell libraries? 01:00:54 http://www.haskell.org/hoogle/?hoogle=%28a+-%3E+b%29+-%3E+%5Ba%5D+-%3E+%5Bb%5D 01:00:54 anyway.. it gives me an idea 01:01:10 itidus21, http://lmgtfy.com/?q=hoogle 01:01:54 the idea it gives me is 01:02:00 kmc: You can't assume if you learn. 01:02:11 a source code search in google and bing 01:02:21 google had one; they shut it down 01:02:31 Not yet: http://www.google.com/codesearch 01:02:31 oh 01:02:35 derp 01:02:40 It's up for a few more days. 01:02:46 heh 01:02:48 wow 01:03:21 god damn it are there no concepts in the world that have yet to be implemented 01:03:21 does this mean that itidus21 getting ideas actually ends up wiping them out from reality? 01:04:00 itidus21: space elevator 01:04:00 itidus21: There are, you just haven't thought of any. 01:04:29 brainfuck on that codesearch would be nice 01:05:27 itidus21, well you'd have a better chance coming up with new ideas if you learned more of the existing ones 01:05:35 like i think everything i've seen you say about haskell is wrong 01:05:55 http://www.google.com/codesearch#search/&q=%22,[.,]%22&type=cs 01:05:59 also you were unwilling to even use google to find out what hoogle is 01:06:13 that's a bad sign for the kind of intellectual curiosity needed to come up with new concepts 01:06:45 -!- Nisstyre has joined. 01:06:57 kmc: Before Haskell, it was the lambda calculus. Unfortunately, we taught him too much lambda calculus for him to use it as a symbol of whatever he needs a symbol of. 01:07:26 haha 01:07:31 luckily haskell is a much larger language 01:07:44 it's possible to go on not-learning Haskell indefinitely 01:07:58 Argh. "oocalc" (and "socalc") were relatively unambiguous, you could get to them with the characters then [tab], but "localc" is /maximally/ ambiguous, you have to type the whole thing >_< 01:08:03 -!- rdrp has quit (Ping timeout: 268 seconds). 01:08:05 local c 01:08:12 INDEED 01:08:13 lol cock? 01:08:22 dude 01:08:27 cock isn't spelled calc 01:08:28 and 01:08:30 lol has an l on the end 01:08:35 * Gregor nods sagely. 01:08:44 there's no i in threesome 01:09:20 i found the most compact language to be unary lambda calculus 11111111111. next in line is binary lambda calculus 0010101010. and next is SK combinatory logic (SK(SKSKSK(S((SS(SSS or something. 01:09:31 what about gzipped perl 01:09:42 kmc: I'm not sure gzip will help. 01:09:46 next i guess is P" 01:09:52 i bet i can write a web server in gzipped perl which is shorter than one in unary lambda calculus 01:09:54 Perl's already pretty high in entropy. 01:10:06 pikhq_, bah, it's all ASCII 01:10:21 that's a 12.5% waste right there 01:10:22 gzipped unary lambda calculus 01:10:23 Ah, right, so there's some room for Huffman to help. 01:11:16 haskell fanboys who don't know the language are probably less annoying than haskell bashers who don't know the language 01:11:22 don't know which one is more common 01:11:35 latter I think 01:11:55 yeah i think so too 01:12:02 but it depends how you define "know the language" 01:12:10 i found the most compact language to be unary lambda calculus 11111111111. next in line is binary lambda calculus 0010101010. and next is SK combinatory logic (SK(SKSKSK(S((SS(SSS or something. 01:12:14 unfortunately I'm not knowlegeable enough to discredit them. :P 01:12:15 binary lambda calculus is obviously more compact than unary 01:12:26 realistically very few programmers will ever see a payoff from learning haskell well enough to write practical production software in it 01:12:35 oh.. i see where i went wrong 01:12:40 i was measuring compact in alphabet size 01:12:50 unary gzipped perl 01:13:35 and the SK( one doesn't count since it has a variable length alphabet 01:14:15 base256 flogscript 01:14:22 itidus21: erm there are just 4 letters, S, K, (, ) 01:14:29 sure you can choose to exclude variables from your alphabets 01:14:40 http://answers.yahoo.com/question/index?qid=20090312200046AAJkBeI 01:14:41 but deep down you know you need them 01:14:44 itidus21: the alphabet is SK() 01:14:58 your mistake is thinking that variables like "x" and "y" are part of the SK calculus 01:14:59 I don't think I can imagine what base 256 would even look like. For the 36 years I have been in the computer business, I have seen base 16 (hexadecimal) representation used and on older machines octal (base 8). Base 256 would mean that you would have to come up with 256 unique single characters for each digit of the number. 01:15:03 they're not, they're used in reasoning about it 01:15:10 ^, this guy doesn't have much to show for his 36 years. 01:15:11 actual SK calculus expressions contain only S, K, and balanced parentheses 01:15:29 codepage 437 is printable at almost every codepoint 01:15:40 hmm 01:15:51 i think the only confusion is 0 vs 32 vs 255 01:15:54 who needs parentheses when you have trees 01:15:54 thats bizzare 01:16:07 and it's the default on DOS / PC BIOS stuff 01:16:13 which is great 01:16:20 because you can just print raw datastructures for debugging 01:16:24 ^bizarre 01:16:28 kmc: As long as you're superhuman :P 01:16:29 monqy: who needs trees when you have parentheses 01:16:33 :0 01:16:36 elliott, hardly 01:16:40 :1 01:16:53 in a lot of cases it'll be pretty obvious 01:17:11 can parentheses make pears i don't think so 01:17:18 ya 01:17:20 pearentheses? 01:17:33 elliott: ahh thats the plunge where it loses the ability to make analogies about it.. when you get rid of variables.. i hate that plunge 01:17:35 ((2)(3)) pears 01:17:48 itidus21: It's called learning. 01:17:52 You seem to have an aversion to it. 01:19:07 hmm 01:19:30 itidus21: imagine you've got these symbols. now imagine you've got these rules that transform groups of symbols into other symbols. 01:21:43 now repeat those rules over and over a bunch until you can't do them anymore. 01:21:56 * kallisti good at explain 01:22:27 i can imagine that the best reference for lambda calculus is the writings of alonzo church. 01:22:42 but that kind of reference is not usually practical 01:23:01 since its not broken into lessons and exercizes 01:23:16 when will people learn to write white papers as lessons and exercizes :P 01:23:41 i don't think your problem is the references 01:24:36 my problem is big. experience in trying to learn things tells me that on some level i'm refusing to work for the understanding 01:25:08 maybe you should do some reduction practice to get used to ski; have you tried that? 01:25:20 I recommend (S I I (S I I)) 01:25:27 IIIIIIIIIIIIIII is good 01:25:32 so really it is more important for me to learn to learn than it is for me to learn lamdba calculus 01:26:14 you can either learn by osmosis (this is what I try to do most of the time), or learn by doing, or learn by people beating you over the head. 01:26:29 it starts with the very small details... like my reluctance to do the dishes 01:27:20 i wont give everyone the run around again on this. :D 01:27:55 -!- pikhq_ has quit (Ping timeout: 240 seconds). 01:28:04 -!- pikhq has joined. 01:28:28 itidus21: I think if you focused as much effort on the actual problem at hand rather explaining to yourself why you can't do something because of X which is like Y ... 01:28:45 itidus21: i have this hunch that on some level there's a part of you that is blocking any action that might improve your self-confidence. 01:29:02 (i think i may have a similar part.) 01:29:04 you can either learn by osmosis (this is what I try to do most of the time), or learn by doing, or learn by people beating you over the head. 01:29:10 your parenthical ended up in the wrong place 01:29:19 also i think that it is often forgotten how rare knowledge of lambda calculus in the world truely is :D 01:29:25 elliott: the third one is just the one that happens the most often. on this channel. 01:29:46 are you sure you don't try it 01:29:48 and in #perl, or so i hear 01:30:04 i doubt anyone i have met in my life knows it 01:30:08 no they tried to beat me over the head but I won't let them. 01:30:09 except at school 01:30:33 it's not learning by beating-over-the-head if you're not learning anything. 01:30:59 #perl is just really really really eager to solve my XY problem. 01:32:12 learning, when it comes down to it, requires work and effort. if there is no work and effort there will be no learning 01:32:18 elliott: do I have some kind of reputation now? :P 01:32:48 kallisti: Yes; it's 31, which is much lower than mine. 01:32:56 itidus21: I can learn the basics of some things with very little effort. mastery takes much more time. 01:33:09 kallisti: You do have a really big X-Y problem, though. 01:33:18 you mean like... in general? 01:33:26 or are you referring to a specific XY problem. 01:33:34 * oerjan wonders if he has a big X-Y problem 01:34:23 i am biding my time until i can incorporate computation into my world view 01:34:45 * kallisti computes his matrix of solidity on a daily basis. 01:34:59 itidus21: are you familiar with the 10000 hour rule 01:35:33 elliott: I don't think it generally makes sense for someone to have a chronic XY problem. it really depends on what they're doing... 01:35:39 according to a wiki page, Platonic abstractions are not spatial, temporal, or mental they are not compatible with the later Idealism's emphasis on mental existence. Plato's Forms include numbers and geometrical figures, making them a theory of mathematical realism; they also include the Form of the Good, making them in addition a theory of ethical realism. 01:35:47 people can be chronically bad at identifying root causes 01:36:30 oerjan: i don't think you need to spend 10000 hours to learn the lambda calculus. 01:37:01 elliott: well it's mainly that you have to actually use those 10000 hours on _something_ 01:37:13 mathematical notations are an application of symbols 01:37:17 kmc: this is not true for all questions I've asked on IRC, but there are many times when people try to solve an XY problem when there is none to solve. 01:37:24 oerjan: IRC! 01:37:27 I really just want them to answer my actual question, and leave me alone. :P 01:37:42 kallisti: well everything you have said about your irc bot thing's design so far 01:37:44 has been terrible 01:37:45 elliott: ooh 01:37:48 elliott: go on 01:37:51 symbols can be percieved through any of the senses 01:38:00 so I'm inclined to think that yes, you really do have a problem. 01:38:17 elliott: plugins are bad? 01:38:19 this is a weird channel 01:38:25 but symbols are not the world 01:38:40 kmc: what did you expect? :P 01:38:47 kallisti, right, it's bad if channels see XY problems where there are none; that was my experience in #python 01:38:58 they get a lot of clueless noobs with terrible ideas 01:39:11 and so if you ask how to do something weird, they assume you are a clueless noob with a terrible idea 01:39:15 I think some people actually just look for an XY problem to solve because they assume that's the case 99% of the time. 01:39:35 in my case i had fairly good reasons for doing the weird thing, which took a while to explain 01:39:45 -!- pikhq_ has joined. 01:39:49 (i was asking which Python text-templating library would be best for generating C++ code) 01:39:53 itidus21: smell symbols? taste symbols? 01:39:58 monqy: yup 01:40:12 actually many questions I've asked on #perl were asked while in the process of figuring out /what/ I want to do. so I go in, ask some insane question, and they respond with disdain and THEN ask what I'm trying to do. 01:40:16 it should be the other way around. 01:40:26 (which is a pretty WTF question, but I had a good reason 01:40:50 as for taste, the army developed a device which can touch the tongue and feed data into the brain from a camera which can be percieved as a black and white image in the visual field 01:40:52 something like that 01:41:08 elliott: so far the only bad idea I've had is attempting to hotswap my code sanely. 01:41:16 er rather 01:41:31 bad idea I've had and mentioned extensively on this channel, about my IRC bot. /anti-lawyering 01:42:07 kallisti: also the thing where you used hideous eval tricks to do plugins 01:42:13 also the "let's run plugins in their own directory" thing 01:42:22 "hideous eval tricks" = perl's internal import mechanism 01:42:23 monqy: about smell.. if you can discriminate smells then when exposed to a series of different smells you could detect patterns 01:42:24 speaking about #haskell because i have experience there, some specific questions are almost always XY problems 01:42:29 also the "oh i don't like perl 5's oo system... so instead of like using one of the alternate ones I'll just do more insane metaprogramming" thing 01:42:30 like "how do I convert IO Int to Int" 01:42:32 also... everything 01:42:39 or "how do I do with type classes" 01:42:48 s/// 01:42:52 this i can understand.. but LC is not so easy 01:43:00 elliott: insane metaprogramming is basically what I do with a slightly complicated project in a dynamically typed language. 01:43:03 it's great. 01:43:04 "How do I define a typeclass?" "Don't. Whatever you're doing stop." 01:43:09 pretty much 01:43:11 kallisti: Right. So you have an X-Y problem, and it's yourself. 01:43:12 -!- pikhq has quit (Ping timeout: 268 seconds). 01:43:16 elliott: noep 01:43:46 I am not writing an IRC bot to distribute to people. I am writing an IRC bot to play around with and have fun. 01:44:07 I'm free to do terrible things at my leisure. I might discover something good. 01:44:26 it is quite surprising to me that computation is really just a branch of mathematics.. 01:44:45 that is a big mental shift 01:45:05 kallisti: that's utterly irrelevant to the question of whether or not you're doing things that you really shouldn't and should instead do 01:45:17 which is what the X-Y problem is all about; it's not "just for you" if you're asking other people to help with your problems. 01:45:19 something which can be drawn from this is the idea that brains are not necessarily computers 01:45:41 itidus21, that's not a justified inference 01:45:41 elliott: sure it is. because what should I do? what is right in a programming situation? in this case something is an XY problem if the thing I'm asking about is actually /incapable/ of solving the problem at hand. 01:46:11 elliott: if it works but disgusts people. I... don't really care. 01:46:17 brains discovered more mathematics than what can be computed 01:46:18 sorry. 01:46:31 kallisti: god this is stupid 01:46:34 it suggests to me that minds/brains are beyond computers 01:46:40 "I don't have an X-Y problem because I don't care" 01:46:47 so nihilist and edgy, man! 01:46:52 brains discovered more mathematics than what can be computed <---- false 01:46:53 elliott: no. I also explained why it's not an XY problem. 01:47:03 if you say so 01:47:10 because what is the correct thing to do depends on why you are programming something. 01:47:14 there is why as well as a how and a what. 01:47:26 >:-) 01:48:11 kmc: who the fuck uses - dashes in an ascii arrow? 01:48:13 you're sick 01:48:16 haha 01:48:28 er 01:48:29 *4 dashes 01:48:35 elliott: also I do use OO just not fully. my plugin code is not "eval tricks". and I'm still not really sure why changing the CWD is a bad idea. 01:49:13 kallisti: i forget which stage comes after denial but i look forward to it 01:49:26 elliott++ 01:49:39 ? 01:49:45 anger. 01:49:48 elliott: fuck you 01:50:10 I can't believe you exposed my terrible XY problem. 01:50:12 great; i can deal with anger 01:50:51 all these eval tricks and changing of current directory so that plugins can easily refer to state within their directory 01:51:02 all not what I wanted 01:52:19 (next stage is bargaining) 01:52:41