|
the PL discourse in 2016, a twitter ad in two replies
|
# ? Nov 4, 2016 04:17 |
|
|
# ? May 22, 2024 14:18 |
|
remember when toyota was forced to reveal how horrifying their code was
|
# ? Nov 4, 2016 08:34 |
my stepdads beer posted:remember when toyota was forced to reveal how horrifying their code was Didn't it come out that their code was so horrible that it probably did kill those people who sued about the uncontrollable acceleration?
|
|
# ? Nov 4, 2016 08:57 |
|
VikingofRock posted:Didn't it come out that their code was so horrible that it probably did kill those people who sued about the uncontrollable acceleration? if the people were killed after they sued then it probably wasn't uncontrolled
|
# ? Nov 4, 2016 09:56 |
|
for real though yes, toyota lost the court case and had a big ol' fine. only a few select experts were allowed to look at the code in a closed environment after a protracted court battle.
|
# ? Nov 4, 2016 09:56 |
|
like we're talking 10k global variables and a thread monitor running in the same thread as the poo poo that it was supposed to monitor bad
|
# ? Nov 4, 2016 09:59 |
|
As I remember it, the code was of such low quality that the auditors were unable to determine whether it was responsible for the accident. "Security by obscurity doesn't work", they say.
|
# ? Nov 4, 2016 10:26 |
|
one of the processes was just called "task X" and the investigator wasn't allowed to talk at all about what it did. maybe it cheated emissions tests lol
|
# ? Nov 4, 2016 10:42 |
|
the one real opening i can see when it comes to challenging java industry dominance is in static analysis tools, which is the case where java implementation and practice is actually working against it (code generation, dynamic loads, etc.), and the state of the art has progressed to a point where it may become a properly huge thing very little hard and fast guarantees in such analysis of course. still, type systems are great and all, but 100% guarantees on a small subset of errors is not really better than e.g. a 60% chance of finding errors of a type twice as prevalent
|
# ? Nov 4, 2016 13:12 |
|
just link the thing http://embeddedgurus.com/barr-code/2013/10/an-update-on-toyota-and-unintended-acceleration/
|
# ? Nov 5, 2016 01:27 |
|
fart simpson posted:like what I could maybe quibble about some little things, but the giant elephant in the room there is macros. macros are poison to tooling of all kinds. the simplest things imaginable fall apart with macros. like "find all references to this variable." and then you can't build on those foundations to build bigger and better things, like renaming refactorings.
|
# ? Nov 5, 2016 01:51 |
|
Malcolm XML posted:in other news, im liking elixir a lot. too bad it has nothing like pandas or numpy or a jvm bridge ((((( Is JInterface or awesome-elixir insufficient?
|
# ? Nov 5, 2016 02:35 |
|
where has this been all my life?
|
# ? Nov 5, 2016 03:21 |
|
Hed posted:where has this been all my life? jesus really? check out this fresh website: http://www.jerkcity.com (neither are bad, its just surprising that youve never come across it)
|
# ? Nov 5, 2016 03:23 |
|
MALE SHOEGAZE posted:Well, it's just an entirely different experience to program in. Go will not allow you to create the abstract class / indirection hell of an enterprise java project. You will never wonder where or how variable is defined, and you will never have to track down the definition of a function, interface, or type across multiple generated XML files. The worst you will have to deal with is generated code, but generated code is still limited to the (very limited) features of the language, so generally simpler to understand than say, a complex generic. usually it's not a big deal but people in my experience abuse these reflection based mechanisms to make it difficult to figure out what's going on instead of just using some loving functions it's much worse than rust-like/lisp-like macros b/c it relies on runtime behavior and can easily defy discovery and helpful tooling
|
# ? Nov 5, 2016 19:29 |
|
also w/ rust-like macros since they expand at compile-time theoretically you could have Sufficiently Smart tooling to support it (at least much easier than any dynamic reflection based shenanigans)
|
# ? Nov 5, 2016 20:48 |
|
so this is pretty cool http://witheve.com/ i'm sure a large project would run slowly and be hell to maintain, but i like the vision in the presentation its nice to see something more novel than just "javascript but we changed the syntax"
|
# ? Nov 8, 2016 12:37 |
|
how is this any different than literate programming combined with a multi-pass compiler literate programming sucks btw
|
# ? Nov 8, 2016 14:36 |
|
comedyblissoption posted:how is this any different than literate programming combined with a multi-pass compiler how dare u impugn the name of the shakespeare programming language
|
# ? Nov 8, 2016 16:10 |
|
literate programming is really good for certain use-cases, mostly when you have truly complex code to write and the comments get substantial enough that some additional typography really is called for doesn't happen that often in my area at least, but now and then i do write something properly complicated and properly formatting the documentation in-line is the best bet
|
# ? Nov 8, 2016 17:53 |
|
LeftistMuslimObama posted:how dare u impugn the name of the shakespeare programming language i taught a three-week "intro to programming" class one time and the way the schedule works there's a two-hour loving-around period so we just talked about esolangs and then did a dramatic reading of a shakespeare program
|
# ? Nov 8, 2016 18:57 |
|
Cybernetic Vermin posted:literate programming is really good for certain use-cases, mostly when you have truly complex code to write and the comments get substantial enough that some additional typography really is called for thats not literate programming. literate programming is when the language tries to actually be correct english (or some other language) and u program it like you're talking to the enterprise's computer or some poo poo. it never works
|
# ? Nov 8, 2016 19:53 |
|
LeftistMuslimObama posted:thats not literate programming. literate programming is when the language tries to actually be correct english (or some other language) and u program it like you're talking to the enterprise's computer or some poo poo. it never works If this is right (I have no idea), that doesn't seem like what eve is? It really just seems like a programming language with nicely-formatted comments.
|
# ? Nov 8, 2016 19:56 |
|
LeftistMuslimObama posted:thats not literate programming. literate programming is when the language tries to actually be correct english (or some other language) and u program it like you're talking to the enterprise's computer or some poo poo. it never works this is not right
|
# ? Nov 8, 2016 20:03 |
|
leftist heap posted:this is not right huh. id never seen it used in this context before and had always seen it used for what i said. but i stand corrected. still, that eve thing doesnt seem to add much except nice formatting of comments so its not nearly the revolutionary tech it's claiming to be.
|
# ? Nov 8, 2016 20:12 |
|
the average programmer being able to clearly explain their code would be revolutionary
|
# ? Nov 8, 2016 20:17 |
|
LeftistMuslimObama posted:thats not literate programming. literate programming is when the language tries to actually be correct english (or some other language) and u program it like you're talking to the enterprise's computer or some poo poo. it never works no no, i am not sure the movement you describe has a name (but the cobol, sql, and applescript sort of tradition), but literate programming is knuths toy thing, with WEB which is basically pascal and tex interspersed, and CWEB and so on does make a lot of sense when you do heavy algorithm stuff like knuth does, where it really is true that each line of code will get read and considered many more times than it is rewritten
|
# ? Nov 8, 2016 20:43 |
|
Literate programming might have a use for writing library API documentation.
|
# ? Nov 8, 2016 20:56 |
|
natural language programming (nlp) is the other one
|
# ? Nov 8, 2016 21:33 |
|
knuth's thing is garbage because when you're lookin at it in an editor (as people are wont to do) now it just looks like tex and code, aka code and code, aka ugly as hell. eve is nice enough to always be pretty. they've got some interesting ideas in there and raise some valid points
|
# ? Nov 8, 2016 21:35 |
|
the eve stuff looks like swift playgrounds imo
|
# ? Nov 8, 2016 21:39 |
|
knuth's original version of literate programming is absolutely hosed, since you write your entire codebase as a set of deeply nested macros in arbitrary order inside a tex document and a preprocessor fishes out all the pieces and puts them in order so that it compiles. it's something only an academic could come up with. literate haskell (and eve since it looks to be the same style), i would argue, is not even literate programming. it's just a way to typeset source code. very rarely is it useful to actually produce a publication (except maybe a journal article or thesis). on the other hand, TeX's original source code is literally a full book, laid out like any other book.
|
# ? Nov 8, 2016 22:30 |
|
gonadic io posted:a little late, but here's what i know as literate programming: let's all just pretend I posted this here in thefirst place
|
# ? Nov 9, 2016 00:09 |
|
Edison was a dick posted:Literate programming might have a use for writing library API documentation. Literate programming is about describing the implementation, not the interface. So, no. There is a reason it is used exceedingly rarely. It's also a bitch to maintain, and I found it impossible for exploratory programming.
|
# ? Nov 9, 2016 00:13 |
|
today was the bay area forth group's annual event many of the sessions came off as backward-looking, summarizing the good forth-based ideas of years past. there were some innovative presentations in the afternoon regarding the multicore chips based on chuck moore's design. some berkeley researchers presented a c compiler for them that can target each unit of a program to a different core. http://pl.eecs.berkeley.edu/projects/chlorophyll/ a guy demonstrated forth running in an excel spreadsheet. he made it to use in his job because previously he used a forth runtime coded in Delphi and the boss didn't like him using such a marginal technology. so forth running in excel is less marginal. at the event last year chuck had said that he had moved away from chip development and was learning to program the iphone. this year he said it's too hard because he can't get at the hardware and he can't get forth running on it. so he gave up on that and bemoaned the overengineering of modern computer systems. this is not a new thing for chuck and it follows the general tendency of EEs to underestimate software there was lively discussion about the risks of closed proprietary hardware (secure boot, graphics controllers etc.) and where engineering talent will come from if people can't hack. this is a real concern imo but in that setting it gets entangled with the EE-centric worldview. i don't see any solution other than a political one of funding independent educational institutions and programs. nothing is going to come along and just flatten the HW/SW stack, however much they might wish. e: oh yah in the discussion someone mentioned mirai DNS attack and how "they still don't know what caused it" (wtf). i didn't get a chance to point out that excessively open and hackable systems are exactly what caused it, and hence "openness" as an absolute principle is not a solution Gazpacho fucked around with this message at 10:34 on Nov 20, 2016 |
# ? Nov 20, 2016 09:27 |
|
Gazpacho posted:forth gross quote:e: oh yah in the discussion someone mentioned mirai DNS attack and how "they still don't know what caused it" (wtf). i didn't get a chance to point out that excessively open and hackable systems are exactly what caused it, and hence "openness" as an absolute principle is not a solution iot vendors treating security as an afterthought is what "caused" it. the openness of a system is irrelevant if you can login with admin/admin creds
|
# ? Nov 20, 2016 13:24 |
|
The Forth and APL communities can be a bit depressing. Both languages cling to life in weird niches isolated from mainstream software engineering and computer science. There are some brilliant people with neat ideas in each camp, but there's a huge chasm between those worlds and the ones everyone else programs in. They have a tendency to come across as impossibly quaint and sheltered. The more time I spend studying them the more certain I am that both have written themselves into extinction patterns. The languages are good, but the communities may be unfixable.
|
# ? Nov 20, 2016 16:12 |
|
There is a large amount of APL users who are not professional programmers, and may not even think of themselves a programmers. To them, an APL environment like Dyalog is more of a immensely powerful calculator. If these people are led away by other tools like Matlab, R, or Julia, they will not be replaced. I'm not even sure how much their numbers are replaced. Most APL programmers I encounter are middle-aged or older. I think one problem is that APL proper is just too drat big. It's very C++-ish in that it has undergone several spurts of growth, and not all have been pretty. There is a reason for the size of the Dyalog APL reference manual. K may be better in this regard.
|
# ? Nov 20, 2016 17:36 |
|
travelling wave posted:gross IJ the tendencies you mention are extreme in that group but they're not unique to it. i've seen them wherever EEs try to be authorities on software. it's as if they think software is just a playground for people who aren't as smart enough to make chips & boards Gazpacho fucked around with this message at 20:09 on Nov 20, 2016 |
# ? Nov 20, 2016 19:39 |
|
|
# ? May 22, 2024 14:18 |
|
By what standards of what makes a programming language good are Forth and APL any good?
|
# ? Nov 20, 2016 20:53 |