|
animist posted:
|
# ? Oct 1, 2018 05:01 |
|
|
# ? Jun 3, 2024 00:25 |
|
cinci zoo sniper posted:i agree, but it’s like javascript for machine learning right now, in terms of specific field ubiquity. i wish something nicer would take the high level spot, or that psf would take their head out of their rear end for python 4 or whatever, but the outlook is not great on that one. the nim people are trying to make an inning there which is a good or a bad thing depending on how you feel about nim
|
# ? Oct 1, 2018 05:02 |
|
nimtorch iirc was dog poo poo the last time i looked at it julia would be good as hell if it didn't have lisp disease (lib maintainers don't give a poo poo about anything, nothing ever gets tested) if you need to use a non-plang use c and torch or c++ and tensorflow bindings, for neural net land. i dunno for rest-of-statistical-computing land
|
# ? Oct 1, 2018 05:33 |
|
bob dobbs is dead posted:julia would be good as hell if it didn't have lisp disease (lib maintainers don't give a poo poo about anything, nothing ever gets tested) ha, sounds about right
|
# ? Oct 1, 2018 05:52 |
animist posted:the nim people are trying to make an inning there last time i looked at nim it felt to be intended to replace octave/matlab, which is a noble pursuit but will be kind of a flop i feel. naturally, these things won’t happen overnight or even over a year or two, but it feels like there is catch 22 with big languages being bad in way A and swarm of microlangs being not a good thing in itself either. like, let’s take ostensibly “scientific computing”, quite a condom of a term, where there are concurrent new developments of - julia, nim, chapel, f#, q#, maybe dlang and rust (in the same way as f# or chapel), probably a lot more smaller fishes what im whining at is that there’s nothing in sight for a language with a feature-complete ecosystem and tooling support that could universally supersede python for numbers problems without inducing learning overhead with semi-mandatory “real programmer things” like memory management also why on earth did julia go for 1-indexing
|
|
# ? Oct 1, 2018 06:00 |
help me im reading down the julia rabbit hole
|
|
# ? Oct 1, 2018 08:17 |
|
cinci zoo sniper posted:help me im reading down the julia rabbit hole have fun indexing at one and never having the right tool for the job unless you make it yourself
|
# ? Oct 1, 2018 08:53 |
Boiled Water posted:have fun indexing at one and never having the right tool for the job unless you make it yourself don’t worry im expert at reading and thinking about being productive without actually doing anything
|
|
# ? Oct 1, 2018 08:55 |
also apparently they added custom indexing at some point
|
|
# ? Oct 1, 2018 08:56 |
|
cinci zoo sniper posted:also apparently they added custom indexing at some point neato, so your packages and libraries can break each other?
|
# ? Oct 1, 2018 09:00 |
|
cinci zoo sniper posted:what im whining at is that there’s nothing in sight for a language with a feature-complete ecosystem and tooling support that could universally supersede python for numbers problems without inducing learning overhead with semi-mandatory “real programmer things” like memory management Common Lisp could if people actually tried to use it again for the task, especially if the big matrix numeric stuff was done via FFI (as it is with Python) one big difference is that you actually could implement the numerics in Common Lisp too—if you were willing to put time into optimization—and get good results (it can perform on the same order as C and FORTRAN these days) as an advantage, you can implement the non-numeric parts of AI systems (like knowledge representation and inference) really well in Common Lisp, and also optimize those implementations well whether by clever coding, giving the compiler hints, or straight up extending the language I’d also suggest that DSLs are far more reasonable to implement atop Common Lisp than most other languages because the typical way a DSL is implemented in Lisp (Lisp-like, via macros) lets you bring the full power of Lisp’s existing tooling to bear on the DSL as if it were any other code in the host language, which isn’t something you can say for most other environments
|
# ? Oct 1, 2018 09:27 |
Boiled Water posted:neato, so your packages and libraries can break each other? i don’t know but i would like to believe it doesn’t matter on compiler level
|
|
# ? Oct 1, 2018 09:49 |
|
Gazpacho posted:UML is useful as a common language for training presentations and documentation. Better than everyone making up their own notation. brand engager posted:I immediately forgot what all the different arrow types meant in whatever course introduced them so
|
# ? Oct 1, 2018 09:58 |
|
diagrams are a ducking pain in the rear end to draw.
|
# ? Oct 1, 2018 10:02 |
|
lot of bad opinions amounting to "just write the code lol" itt atm granted i have largely never done anything except draw haphazardly on a whiteboard then written the code, but obviously uml, and pseudocode, does something rather incomparable with the "real" code
|
# ? Oct 1, 2018 10:09 |
|
Cybernetic Vermin posted:lot of bad opinions amounting to "just write the code lol" itt atm or, you know, write the reqs and docs with human language
|
# ? Oct 1, 2018 10:20 |
|
Wheany posted:well, except for
|
# ? Oct 1, 2018 10:32 |
|
Gazpacho posted:Theres four arrow types of consequence, tattoo them on your arm if that’s the only way you can remember them i have now done this, but there are 20 people in this conference room that haven't done it. please advice
|
# ? Oct 1, 2018 10:46 |
|
Wheany posted:i have now done this, but there are 20 people in this conference room that haven't done it. please advice tattoo it on them too
|
# ? Oct 1, 2018 10:54 |
|
https://blog.joda.org/2018/09/do-not-fall-into-oracles-java-11-trap.html oracle jdk no longer free for commercial use lol
|
# ? Oct 1, 2018 12:21 |
Wheany posted:https://blog.joda.org/2018/09/do-not-fall-into-oracles-java-11-trap.html wasn’t that sorta expected in that folks should use openjdk or something? also i wonder how oracle legal will detect what jdk i use
|
|
# ? Oct 1, 2018 12:38 |
|
I have a 6 hour flight tomorrow, what's a good java book to download and work through
|
# ? Oct 1, 2018 12:53 |
creatine posted:I have a 6 hour flight tomorrow, what's a good java book to download and work through oracle java development kit terms of service
|
|
# ? Oct 1, 2018 13:15 |
|
Wheany posted:https://blog.joda.org/2018/09/do-not-fall-into-oracles-java-11-trap.html otoh they have brought openjdk into parity at the same time, so we now get flight recorder and mission control for free, which is pretty neat. also zgc is a huge improvement overall this is a lot less bad than what i expected from oracle so far
|
# ? Oct 1, 2018 13:31 |
|
zgc is very on-topic for the thread really: http://cr.openjdk.java.net/~pliden/slides/ZGC-Jfokus-2018.pdf having done a lot of huge-heap java stuff in the past it is actually a pretty exciting addition
|
# ? Oct 1, 2018 13:35 |
|
creatine posted:I have a 6 hour flight tomorrow, what's a good java book to download and work through that's an interesting gang tag, friend
|
# ? Oct 1, 2018 14:21 |
|
animist posted:
i know i've worked with people trying to build products without even this level of understanding the problem
|
# ? Oct 1, 2018 14:24 |
|
Captain Foo posted:that's an interesting gang tag, friend
|
# ? Oct 1, 2018 14:34 |
|
Shinku ABOOKEN posted:or, you know, write the reqs and docs with human language You're talking to programmers here
|
# ? Oct 1, 2018 14:47 |
|
eschaton posted:Common Lisp could if people actually tried to use it again for the task, especially if the big matrix numeric stuff was done via FFI (as it is with Python) they vectorize the knowledge representation and inference too nowadays and common lisp is older than python, if someone really wanted good ffi numerics in it they would've done it by now
|
# ? Oct 1, 2018 14:54 |
|
creatine posted:I have a 6 hour flight tomorrow, what's a good java book to download and work through Practical Common Lisp
|
# ? Oct 1, 2018 21:42 |
|
bob dobbs is dead posted:they vectorize the knowledge representation and inference too nowadays you’re assuming they haven’t
|
# ? Oct 1, 2018 21:45 |
|
eschaton posted:you’re assuming they haven’t knowing the numerics libs in common lisp 5 years ago, i'm pretty sure the answer is they haven't, yeah. the blas poo poo is pretty dog poo poo in actual docs and testedness compared to how good numpy is
|
# ? Oct 1, 2018 21:46 |
|
creatine posted:I have a 6 hour flight tomorrow, what's a good java book to download and work through java language spec skip the boring parts they won't bite you in the rear end for 6 months at least
|
# ? Oct 2, 2018 01:24 |
|
Wheany posted:https://blog.joda.org/2018/09/do-not-fall-into-oracles-java-11-trap.html christ, who wants to deal with this nonsense? use .net core instead.
|
# ? Oct 2, 2018 01:30 |
|
i like dotnet core
|
# ? Oct 2, 2018 02:13 |
|
brap posted:i like dotnet core its good
|
# ? Oct 2, 2018 03:19 |
|
brap posted:i like dotnet core i read this as "i like don't care" at first lol
|
# ? Oct 2, 2018 03:20 |
|
akadajet posted:christ, who wants to deal with this nonsense? use .net core instead. it's all about milking the install base. oracle doesn't give a gently caress whether hipsters get turned off on java, they weren't going to use it anyway
|
# ? Oct 2, 2018 03:41 |
|
|
# ? Jun 3, 2024 00:25 |
|
Kevin Mitnick P.E. posted:it's all about milking the install base. oracle doesn't give a gently caress whether hipsters get turned off on java, they weren't going to use it anyway yeah, c# is totally a hipster lang
|
# ? Oct 2, 2018 03:43 |