libera/#commonlisp - IRC Chatlog
Search
11:08:08
lisp123
First hit on Google was https://rdrr.io/cran/qdapDictionaries/man/GradyAugmented.html
11:48:46
hayley
lisp123: What I did around this time last year, as we used it for university work, was to run Minecraft with shaders on. That way the GPU would warm your computer room up.
11:49:38
hayley
In a pinch, model checking or fuzz testing can work, though in my experience CPUs make for worse heaters.
11:50:35
lisp123
i wonder if you could do some cryptomining for the same and make some $$ on the side
11:52:30
hayley
They also tend to be tuned in a way that makes interactive use of the computer unbearable, too. So best stick with the classics, if you want to use the computer.
13:10:53
jackdaniel
zgaduj zgadula: should `#+ (or) (badum +tss+ #<compute-me>)` in a file cause reader to signal an error?
13:12:31
Bike
i don't think it's possible in general to figure out where a #<whatever> ends, since the whatever could have arbitrary text
13:14:05
Bike
in clasp we used to have a test that did something like #+(or) #gblablabla, #g being an undefined reader macro. but there's obviously no way to handle that well, since the #g could introduce arbitrary syntax and the reader can't tell when the expression reads
13:14:47
Guest74
::notify lisp123 I think I'm going to make a github with all the Moby files as I had to dig them out of archive.org. They're also at project gutenberg but they rather stupidly present the readme's as the text so you have to dig and get the links to an ftp site.
13:23:16
Guest74
I'm starting to think this Ward guy might have been a hoarder. I'm not sure I see the general purpose of a word list with entries such as AAAA AAAS AAPSS. Nor do I understand why he basically chose a different format for every list.
13:25:41
Guest74
I'm starting to think composable dictionaries of specific points of view would be better. Any thoughts? Do i really want to check for spelling mistakes against acronyms if I'm not using any acronyms?
13:28:20
Bike
trying to spell check acronyms seems like a fool's errand since anyone can introduce a new one
13:28:53
Guest74
I wish there was an easy way to see how much space an object was taking up. writing out the combined dictionaries in a structured format came out to ~100megs, about 3/4 of that was white space from pretty printer.
13:31:56
beach
Guest74: You can probably find a pretty good approximation. But I suspect there is nothing to worry about. In memory, you can probably also do a lot of sharing.
13:32:37
beach
Guest74: Like the trie data structure can save a lot of space, and there are a bunch of fairly common encoding schemes that will make it even more compact.
13:37:31
Guest74
I still see a use in pov dictionaries for definitions. the word salt in context of cook, chemist, geologist is different.
13:39:15
Guest74
I'm going to load things into tries today and see how much difference it makes. At least one of the trie libraries out there has huge dependencies, so maybe it might be better to stick with hashtables if there's not much difference.
13:46:26
Guest74
TIL there are 9 different ascii encodings of IPA + arpabet and dictionary writers use them all, not to mention the proprietary phonetic encodings of companies. I've got a translater going now so I'm thinking of storing pronounciations in IPA with some sort of marker for language/dialect. Or wrap it in a struct with encoding type, dialect etc...
13:46:27
Guest74
and then you can choose to translate to your preferred encoding (cause most of them look like gobbledygook)
13:59:16
Guest74
beach: The main concern with size is github file size limits. Which would be exceeded adding definitions. I haven't tested the thesaurus yet, but I suspect the way to deal with that is pointers to sets of word, which I'm starting to suspect is what wordnet is all about.
14:20:09
hayley
Isn't there some "large file storage" for Git? Though a large dictionary is not a good fit for version control, yes.
14:44:33
SAL9000
there's also git-lfs -- iirc github doesn't support git-annex although I might be wrong
15:21:32
Guest74
some 'thesaurus' entries on 'sad" , "Quaker-colored' "contemptible" "creamy" . Wait, what?
15:29:45
Guest74
it seems creamy might be coming from a connection to subtle colours connected to blue connected to sad. Going to have to figure out a way to structure those connections. though the all inclusiveness doesn't seem bad for creative writing.
15:36:55
Guest74
it seems to include entire groups if there's any relation. "Love" includes every roman god.
15:40:40
Guest74
I confused that with "Amor" which was listed under "Love" and has lots of gods. I don't recall a god named Amor.
15:50:30
Guest74
The sunglasses make me think, that kids got a bright future. Just cruising along in the bright sunshine. I'm going to have my irc client play https://www.youtube.com/watch?v=oqhHThWflLI everytime you join.
17:45:31
Guest74
I'm wondering if it'll have any effect on these numbers. writing everything out with pretty nil is 66megs, conspack hashtable is 74megs, conspack trie is 147, tracking-refs for trie is 72, tracking refs with hashtable is 44megs.
17:47:42
Bike
you can try it out, but i didn't really change the encoder at all. i was thinking about adding a dense packing mode where it tries to track everything as a ref, but my data were small enough without that
18:02:42
Guest74
all of this is much larger than just writing plain text separated by a token. I'm guessing referencing/indexing is a problem when you've got ~300k objects
20:28:31
dirtcastle
rate how crazy I am from 0 to 10. I'm big fan of nix os and guix. but since nix language isn't a full featured language I like guix better. but since guix is gnu's they won't support windows or linux. nixos is on apple but i don't think it's on windows. I think docker didn't solve any problem it was meant to solve. I think we need advanced package manager instead of docker containers. so I'm thinking of making a new nix/guix like package
20:34:19
morganw
I think if you look at the non-guix channel there are also tools for just wrapping existing binaries so they will run.
20:39:49
dirtcastle
true. it's just my assumption but for some reason it looks like lispers don't use guix distro because it uses guile scheme? and nyxt and lem editor is getting attention. it looks like lots of lispers want their software written in common lisp and not any other lisp.
20:40:52
dirtcastle
morganw: what u said is true but they all definitely feel like we are dirty our hands but it's usuable I think.
20:42:45
morganw
I've only just started to use it, but for me the problem is that the development workflow doesn't feel natural and loading the packages is so slow it doesn't seem possible without compiling it all ahead of time. But other people don't seem to be hampered so perhaps I'm doing something wrong.
20:45:20
morganw
I think they will always be interesting in having the tooling work well and work better, and there is so little hardware that works on the free software kernels that I think most people will already be using other channels.
20:48:20
phantomics
dirtcastle: I've thought a CL init system would be cool, both that and a package manager would be huge projects
20:51:22
morganw
I've not tried it, but it sounds a bit more like standard configuration management rather than custom tooling all the way down.