libera/commonlisp - IRC Chatlog
Search
13:29:09
Shinmera
Anyway, to make the day less slow, we're now reaching the last week of the Kandria kickstarter campaign: https://www.kickstarter.com/projects/shinmera/kandria/posts/3549518
13:29:33
Shinmera
The next stretch goal is official mod support, which I think would be quite exciting for people here in specific :)
13:51:25
Shinmera
I've used jsown before and shasht. There's a new one I'd like to use but it's still not on ql, so I cannot.
13:52:46
Shinmera
Jonathan made my image memfault when I tried it once and that's enough for me to stay away forever.
13:53:15
Josh_2
Yes a potential problem with all the speed 3 safety 0 declarations. I have my images locked at safety 3
13:57:09
Nilby
the fist time i had to switch json libraries, i fixed my code to allow swapping them. but i've never had stop using one of Shinmera's libraries :)
14:01:54
yitzi
I think that names like cl-json are silly. I like unique and easy to find names. A hint to the etymology is in the description "Common Lisp JSON reading and writing for the Kzinti."
14:04:25
jackdaniel
ACTION wonders whether cl-foo name conventions is still obsolete, or has it moved to classic/retro stage already ;)
14:04:52
yitzi
Specifically, I used pretty much all of the JSON libraries in my Jupyter work and found them all deficient. Usually in encoding escape sequences, etc.
14:18:26
semz
Same. Ended up writing my own, considered publishing it, but decided against it in the end because good god there are enough already
14:26:32
Nilby
Have I forgotten again, or is there really still no fast way to convert between integer and bit-array?
14:35:13
Nilby
The code in sbcl i'm looking at seems to define bit vector operation on single bits :( but I haven't look at the compiler yet
17:04:56
Shinmera
Nilby: (read-from-string (format NIL "#*~b" 101)) is very fast to type at least ;)
17:07:53
Nilby
Shinmera: And one library even does it close to that way. Thank you, but I've already hacked something quick enough with sbcl internals. Now I have to figure it out on any other implementation. Sigh.
17:14:59
Shinmera
I'm trying to think of situations where having a bit vector would make things faster, rather than just having an integer and using ldb etc.
17:18:56
Nilby
Bike: yes, that's one reason. random 2d access seems easier too. i imagine i might go back to the the ldb way if need be
17:27:56
Nilby
I would wish that the number and array would end up being equivalent speed, and one could just choose the most convenient representation, but I guess i'll have to see
17:33:21
pjb
Nilby: there's almost no difference between number and bit-vector access times: https://termbin.com/wmxs
17:33:22
Nilby
pjb: Thanks. That looks pretty fast. That is very similar in structure to what I did with %*vector-raw-bits. Perhaps I can do that for other implementations.
17:35:20
Nilby
Yes, my code is very mutating, which is one reason why I chose the vector way to begin with
17:36:03
pjb
And, if you have very big vectors, you can also use multithreads to process separate parts.
17:38:36
Nilby
yes, most of the problems i'm working on have natural parallelism, so it can be split that way