|
Arcsech posted:oh and give it an incredibly lovely build tool sbt is the worst thing about scala
|
# ? Jul 22, 2015 00:53 |
|
|
# ? Jun 6, 2024 16:46 |
Notorious b.s.d. posted:broken garbage collection Interested in hearing more about these two.
|
|
# ? Jul 22, 2015 00:57 |
|
VikingofRock posted:Interested in hearing more about these two. python has no garbage collector. it uses ref counting. substantially all python programs leak memory. pypi. jesus. what a clusterfuck. it's exactly as bad as rubygems and cpan. compiling poo poo on the fly on the install target is not a viable deployment scenario.
|
# ? Jul 22, 2015 01:27 |
|
i totally get the temptation that spawned cpan/rubygems/pypi perl/ruby/python code is distributed as source, because source is all there is. they are interpreted languages. shipping a tarball of source code plus some metadata seemed like a really good idea the problem is that then people started shipping C bindings that necessarily included C source code. so now the scripting language "package" systems are distributing source that has hidden dependencies on system libraries, source code that gets compiled on the spot, that sometimes, horror of horrors, downloads duplicate library source from the internet and compiles it the best part of this, the real poo poo frosting on a blood sundae, is that you never get that poo poo patched, because you don't even know it exists. pretty much every ruby/python user has at least one unpatched libxml2 running around from where nokogiri or lxml decided to build one on the sly
|
# ? Jul 22, 2015 01:31 |
|
fritz posted:everything except embedded a restricted subset of C++ is fine for embedded, especially since embedded no longer means "a couple hundred bytes of RAM at a couple MHz at most" for most cases. better abstraction tools coupled with aggressive optimization can be way less error-prone than going full DIY for everything like most uses of C will wind up
|
# ? Jul 22, 2015 02:00 |
|
eschaton posted:a restricted subset of C++ is fine for embedded, especially since embedded no longer means "a couple hundred bytes of RAM at a couple MHz at most" for most cases. weird, that's basically my entire embedded world
|
# ? Jul 22, 2015 02:02 |
|
Notorious b.s.d. posted:python has no garbage collector. it uses ref counting. substantially all python programs leak memory. ref counting is a form of garbage collection
|
# ? Jul 22, 2015 02:02 |
|
Bloody posted:weird, that's basically my entire embedded world how's your decrement-and-branch-if-zero instruction working out?
|
# ? Jul 22, 2015 02:03 |
|
eschaton posted:ref counting is a form of garbage collection yeah, that's why I won't answer the phone when your mom calls any more
|
# ? Jul 22, 2015 02:12 |
|
eschaton posted:how's your decrement-and-branch-if-zero instruction working out? not building that for sure and otherwise questionably!
|
# ? Jul 22, 2015 02:23 |
|
Nippashish posted:oh cool. keras is one of the "fancy new nnet libraries" I was thinking of. have you figured out a nice way to make rnns in torch? ive tried like 4 or 5 different ways and all of them end up making some part of the process really annoying. are you writing your own or are you using andrej's code like everyone else?
|
# ? Jul 22, 2015 05:57 |
|
modern c++ actually owns and is a good language that is only getting better. if you code like its 2003 it's an abomination that should be avoided at all costs
|
# ? Jul 22, 2015 06:11 |
|
recently first day on a new project i got a mass fwd saying hey guys stop forgetting the brackets in delete[], its causing memory leaks.. lol
|
# ? Jul 22, 2015 07:23 |
Flat Daddy posted:recently first day on a new project i got a mass fwd saying hey guys stop forgetting the brackets in delete[], its causing memory leaks.. lol
|
|
# ? Jul 22, 2015 07:57 |
|
Joe Law posted:modern c++ actually owns and is a good language that is only getting better.
|
# ? Jul 22, 2015 08:02 |
|
the abstractions are actually becoming worth paying for, and the fact that they're optional makes it a pro choice wherever you can't just throw absurd amounts of hardware at your problems
|
# ? Jul 22, 2015 08:03 |
|
or you'd like to throw absurd amounts of hardware at the problem and actually get most of the hardware applied to your problem
|
# ? Jul 22, 2015 08:03 |
|
Notorious b.s.d. posted:the really bad stuff about python: it's true, this is all bad but I don't get your beef with reference counting. are you like a mark and sweep kind of dude?
|
# ? Jul 22, 2015 08:13 |
|
python has a really easy to use threading model actually
|
# ? Jul 22, 2015 09:58 |
|
fart simpson posted:python has a really easy to use threading model actually yeah, turns out it's easier to write threaded code when the threads don't actually execute concurrently
|
# ? Jul 22, 2015 10:19 |
|
incidentally which ide do you use for lua? i'm using itorch notebook, and while it's great for poking around it feels a poor fit for developing a library my ideal would be spyder-for-lua, but i can't find anything along those lines
|
# ? Jul 22, 2015 11:19 |
|
Notorious b.s.d. posted:the really bad stuff about python: Threading isn't broken, as there is only a single thread
|
# ? Jul 22, 2015 11:22 |
|
Soricidus posted:yeah, turns out it's easier to write threaded code when the threads don't actually execute concurrently lol
|
# ? Jul 22, 2015 14:14 |
|
Soricidus posted:yeah, turns out it's easier to write threaded code when the threads don't actually execute concurrently Valeyard posted:Threading isn't broken, as there is only a single thread these guys get it
|
# ? Jul 22, 2015 14:25 |
|
Notorious b.s.d. posted:pypi. jesus. what a clusterfuck. it's exactly as bad as rubygems and cpan. compiling poo poo on the fly on the install target is not a viable deployment scenario.
|
# ? Jul 22, 2015 19:02 |
|
MononcQc posted:Isn't python interpreted anyway? sure. except for FFI bindings.
|
# ? Jul 22, 2015 19:12 |
|
like he said, it sounds like a good idea on the surface, but it isn't because you actually do have libs that depend on compiled code.
|
# ? Jul 22, 2015 19:12 |
|
Dessert Rose posted:like he said, it sounds like a good idea on the surface, but it isn't because you actually do have libs that depend on compiled code. like what's a better solution? jw
|
# ? Jul 22, 2015 19:17 |
|
MALE SHOEGAZE posted:like what's a better solution? jw distribute working binaries?
|
# ? Jul 22, 2015 19:21 |
|
Dessert Rose posted:distribute working binaries? so instead of having an insecure unpatched libxml2 that you don't know about that some module built from source, you have an insecure unpatched libxml2 that you don't know about that some module bundled as a binary? also we have to live in the real world, where a lot of python runs on linux and distributing binaries for linux is painful to say the least
|
# ? Jul 22, 2015 21:12 |
|
coffeetable posted:incidentally which ide do you use for lua? i'm using itorch notebook, and while it's great for poking around it feels a poor fit for developing a library i use atom atm. sublime is also ok. intellij actually has a p good lua plugin that can do some (limited) refactoring like renaming variables and whatnot, but intellij is totally unusable with a network drive. its a shame that tooling in lua is so profoundly bad. i really miss pycharm.
|
# ? Jul 22, 2015 21:14 |
|
Dessert Rose posted:like he said, it sounds like a good idea on the surface, but it isn't because you actually do have libs that depend on compiled code. such as anything that wants to be fast
|
# ? Jul 22, 2015 21:42 |
|
Soricidus posted:so instead of having an insecure unpatched libxml2 that you don't know about that some module built from source, you have an insecure unpatched libxml2 that you don't know about that some module bundled as a binary? one of my coworkers has been using conda somewhat successfully for their python projects
|
# ? Jul 22, 2015 21:44 |
|
code:
|
# ? Jul 22, 2015 22:31 |
|
there was this python library that needed to compile some dumb C code and it was super broken on windows, I tried to talk with the author about it (because I didn't just want to rewrite the drat thing) to make this compilation step optional (which wasn't only on windows because...?) he basically told me to gently caress off open source is dispiriting sometimes
|
# ? Jul 22, 2015 22:43 |
|
mongo db is JavaScript: the database
|
# ? Jul 22, 2015 23:00 |
|
MALE SHOEGAZE posted:like what's a better solution? jw use a language that isn't so slow that it needs c libs for totally mundane poo poo?
|
# ? Jul 22, 2015 23:06 |
|
NihilCredo posted:or you can do this and let windows do all the work I tried using this once and found problems w/ it that I cant remember right now. it did clean it up for you and that was cool .
|
# ? Jul 22, 2015 23:17 |
|
Notorious b.s.d. posted:i totally get the temptation that spawned cpan/rubygems/pypi the thing that makes that concept even worse is that java lets you do the same thing, but with a better structure since source and javadocs can be in separate packages for the same artifact. the one thing those languages kind of had going for them is done 1000000x better in java.
|
# ? Jul 22, 2015 23:22 |
|
|
# ? Jun 6, 2024 16:46 |
|
oh, I remember what my problem w/ the file autodelete thing was. I was creating a temp file that was going to be passed as an argument to an external application, but when you create an autoclosing handle like that it creates it in exclusive read or something so only your process can read it. that meant the external app couldn't read the file while mine was running and if I ended mine then it deleted the file. pwnt. I ended up using a more thorough IDisposable temp file.
|
# ? Jul 22, 2015 23:25 |