freenode/#lisp - IRC Chatlog
Search
19:55:44
ckonstanski
I have only used clsql. I'm interested in postmodern, but I've written too many apps that had to connect to multiple databases of differing types (within the same app). That's where clsql shines because it supports many database engines.
20:00:29
attila_lendvai
there's also hu.dwim.rdbms although it only has one thoroughly tested backend, which is for postgres. oracle has also been used in production by another team
20:27:23
attila_lendvai
so, how do I touch a file? is there anything simpler than a with-open-file for writing?
20:30:40
attila_lendvai
didn't you mean :direction :output :if-exists :append :if-does-not-exist :error (it assumes the file exists and needs to be touch'd)
20:46:29
attila_lendvai
I need the contents to remain intact. I only want to update the last modified time
20:48:45
ckonstanski
Uses /bin/sh as the shell I think. Careful on systems that actually use /bin/sh as opposed to symlinking it to /bin/bash.
20:56:27
attila_lendvai
this is incredibly annoying... I can't even find a kludge to do it without leaving the standard. there's not even (setf file-length)...
21:05:09
attila_lendvai
I ended up using UIOP's with-staging-pathname and copy the file contents to a new file. *shakes head*
21:13:02
ckonstanski
Maybe you hate the idea of (uffi:run-shell-command) but looking at its implementation might provide some ideas.
21:19:12
attila_lendvai
yes, I hate the idea of exec'ing a binary for something as simple as touch'ing a file. I hate it, and I also want it to be on record... :)
21:22:21
attila_lendvai
syscalls, except when you're on windows... I don't want to go down on that road. it's much more bumpy than a spurious copy-file
21:25:19
pjb
attila_lendvai: you can also call MS-Windows functions. Just use #+windows #-windows ;-)
1:18:24
pillton
Xach: The twitter link found in the "Planet Lisp is back on twitter" article is wrong.
2:04:29
p_l
Never expected it, but sometimes I'm pining for DISASSEMBLE on a lisp machine... For teaching new people about computation
3:32:35
aeth
(defun foo (x) (1+ x)) (disassemble #'foo) ; only three instructions in CLISP, but 12 in SBCL
3:34:41
aeth
Alternatively, you could use disassemble on a machine with a simpler to understand instruction set, rather than x86 with its Turing-complete MOV.
3:38:13
Bike
mov is turing complete because it has indirections and indexing, which are pretty common on other processors as well.
5:19:12
p_l
ACTION ponders if there's somewhere one could do PhD in AI without essentially doing matrix math and stats only :/
5:22:42
rme
When I am cranky, I say that it seems like AI / machine learning nowadays is mostly a matter of solving y = mx + b.
5:31:25
p_l
rme: what is sold under machine learning is kinda exactly that... though usually people who reach for machine learning could just as well hire a competent statistician, excuse me, "Data Scientist"
5:38:41
p_l
I need to track down the papers, but I've seen some interesting work on meshing together "nice" and "scruffy" AI
5:50:35
p_l
White_Flame: as I said, there is some interesting research out there, just not as "visible" as "write a classifier after a tutorial of Python and TensorFlow"
5:51:15
White_Flame
I've been spending a fair amount of time reading old old Lisp inference source code
5:51:47
White_Flame
lots of interesting paths to discover that never caught on, but have their own strengths
5:52:15
White_Flame
that's one problem with the AI field in general; everybody focuses on their pet aspects, and won't work with anybody who believes a different aspect is most important
5:52:38
p_l
also, fun thing: a big name in "large scale server ops", kubernetes, is essentially an blackboard architecture from good old AI
5:52:52
White_Flame
and everybody believes that their one secret sauce is the magic to AI if it could only be developed far enough
5:54:07
p_l
well, given enough computronium, AIXI should converge on "general intelligence", but it's probably prudent to shoot anybody who tries
5:57:52
p_l
White_Flame: well, IIRC the basic explanation (it's been 4 years, and even then we didn't really go into blackboard systems) you have "blackboard" that serves as common model of what you're working on, and various "experts" that use said data, and change it, working towards solution
7:32:41
fiddlerwoaroof
Hmm, I'd imagine you could just close *standard-input*, *standard-output* and, essentially, daemonize
7:33:25
fiddlerwoaroof
But that probably has issues in a multi-threaded environment, even if it's theoretically possible