freenode/lisp - IRC Chatlog

3:01:41 thmprover I'm actually having some difficulty getting into the hang of literate programming in org-mode.

3:04:10 lotuseater so what problem specifically?

3:06:04 thmprover Literate programming (in the Knuth sense) is difficult, but getting org-mode to export the correct HTML is tedious. It feels like I'm fighting against the tide.

3:06:30 lotuseater do you know pandoc?

3:06:55 thmprover Yes, I dislike it.

3:06:59 lotuseater why?

3:07:34 thmprover I take advantage of a lot of custom TeX macros, and they don't play nice with pandoc.

3:07:59 lotuseater ah yes that's a tradeoff

3:08:41 thmprover I mean, I can see pandoc's appeal to the generic user, but I'm the "crazed TeX survivalist"-type user.

3:08:56 no-defun-allowed Yes, pandoc doesn't properly run TeX. So I use htlatex and fix up a bit of the nonsense it emits with a script that runs sed.

3:10:05 thmprover I'm starting to explore htlatex to translate this manuscript I'm writing on theorem provers to HTML, it's a nifty program, but so...niche.

3:10:56 no-defun-allowed Specifically, we use https://gitlab.com/cal-coop/netfarm/documentation/-/blob/master/make-htlatex.sh and that ends up making https://cal-coop.gitlab.io/netfarm/documentation/ which is...bearable.

3:11:02 lotuseater a friend of mine also showed me this half a year ago, maybe it's of use for anyone else: https://latex.now.sh/

3:11:36 lotuseater no-defun-allowed: nice look!

3:12:36 no-defun-allowed It seems likely thmprover wants to use a LaTeX document to make HTML.

3:12:41 thmprover lotuseater: the rendering is, I can't quite put my finger on it, the fonts are slightly mismatched. The CMR is too light compared to the monotype font weight.

3:12:57 lotuseater ok

3:13:11 thmprover It's nifty, but not for me.

3:14:18 lotuseater how do i in org-mode set globally to tangle all blocks? except the ones with ":tangle no"

3:16:34 thmprover "#+OPTION: :tangle <file>"

3:16:37 thmprover (I think)

3:16:49 lotuseater ah yes thx

3:17:10 thmprover np

3:17:22 lotuseater just taking org-mode and all its features is soo much :D

3:18:23 thmprover I'm trying to work with it to write my notes on earth science, but it's...I don't know...

3:19:04 thmprover I sense I am doing things wrong with org-mode, but I don't know what "doing things right" looks like, and the only talks on it seem to be excited users talking about how great it is.

3:21:20 thmprover Maybe I should go complain on #org-mode about org-mode

3:30:07 fiddlerwoaroof Hmm, I used custom TeX macros all the time with pandoc

3:31:34 thmprover fiddlerwoaroof: I'm doing pretty tricky macro stuff, which takes advantage of the dynamic scoping of TeX (and I think confuses pandoc)

3:31:47 fiddlerwoaroof Interesting

3:32:38 fiddlerwoaroof The most complicated I had were Tikz diagrams

3:34:22 fiddlerwoaroof But, I was writing humanities papers, which are somewhat less involved

3:40:10 no-defun-allowed Well, I don't know if pandoc does any macroexpansion. It just ignored my \Defun commands.

3:53:01 fiddlerwoaroof Yeah, I mostly relied on latex being passed through in md -> latex conversion

3:53:09 fiddlerwoaroof I've never tried latex as an input format

3:55:54 fiddlerwoaroof One thing I find interesting about SLIME, is that it doesn't seem to make much use of generic functions or EIEIO

3:56:28 fiddlerwoaroof Maybe it's too old, but I've found it pretty useful for managing my emacs configuration

3:56:43 fiddlerwoaroof (eq it 'SLIME)

4:10:57 beach Good morning everyone!

4:11:04 fiddlerwoaroof morning, beach

5:37:27 saganman Morning beach

5:37:32 saganman Morning Everyone!

5:37:41 beach Hello saganman.

5:38:00 saganman Hey beach.

6:27:50 badcfe in 'on lisp' fig 14.3 there's a aif2 that seems wrong to me. but i see this is not in the errata, and i'm not really a lisper, hence i am guessing i'm wrong

6:28:54 beach Let me check

6:30:16 badcfe on the forth line there, i see (if (or it ,win) ,then ,else)) .. and i am thinking it should be (if ,win ,then ,else)

6:31:05 beach I guess that depends on how he specified what should happen with the values of the test.

6:32:16 badcfe yes, and it's not even specified, but there's also a usage example, which makes me think what i mention here

6:32:30 badcfe s/even/only/

6:32:43 badcfe s/forth/fourth/

6:32:56 badcfe late for me. sorry

6:32:59 beach Ah, yes, let me look at the example.

6:33:44 beach Won't the two versions have the same effect for GETHASH.

6:33:46 beach ?

6:34:09 beach If the first value is true, then so is the second.

6:34:48 beach But, yes, I think you may be right.

6:35:13 beach Then, I just barely finished my morning coffee, so opinions of others are welcome.

6:43:10 beach It was a while ago that I read On Lisp, but now I see that AIF2 solves a major problem with the single-value version, namely that, if you respect the advice of Norvig and Pitman of using only Boolean expressions as the test in IF, then the "IT" of AIF will never be a useful value.

6:44:08 badcfe i see now, that it doesn't matter, because both are null iff we shall branch to ,else -- if we can assume that any function indicating not-found on second return value will always have a null first return value as well, in other words that we have a functional dependency (sorry to use database theory term here). however, i think my variant is more intuitive in addition to be shorter

6:44:36 beach I agree with your assessment.

6:47:50 badcfe hey thanks. maybe i should submit as errata.

6:48:36 beach Sure. It can't hurt.

6:48:52 badcfe about that advice you mention, you suggest it as practice? i'm just curious

6:49:52 beach Which one?

6:50:31 beach I try as much as possible to follow the advice of page 13 of the LUV slides by Norvig and Pitman, if that's what you are asking.

6:50:52 beach So yes, I use only Boolean values for tests.

6:51:14 badcfe yes, that's what i was asking. i guess my next read should be LUV

6:51:58 beach I like it very much. It was written by two people with a lot more experience with software development in larger teams than I will ever have.

6:52:44 beach Many #lisp participants contest their advice, but I think I know for a fact that the people who do contest it have way less experience than these two people have.

8:16:47 fiddlerwoaroof If these are the notes you're talking about, they're pretty interesting: https://www.cs.umd.edu/~nau/cmsc421/norvig-lisp-style.pdf

8:18:02 beach Yes, that's it.

8:19:42 fiddlerwoaroof I always forget CATCH/THROW exist

8:21:06 phoe regarding 4.3? that's sorta good

8:21:23 fiddlerwoaroof Yeah

8:21:52 phoe they very rarely need to be used, given that non-local GO/RETURN-FROM also exist and are slightly safer to use

8:22:17 fiddlerwoaroof Yeah, the main difference is dynamic extent of CATCH, right?

8:22:20 phoe ...unless they leak out of their dynamic scope, but that is usually not very hard to spot in code

8:22:23 phoe yes

8:22:38 phoe CATCH is fully dynamic, which also means it's a wee bit slower

8:22:52 fiddlerwoaroof I actually just had a problem that is shaped this way :)

8:22:54 phoe but THROW doesn't need to lexically match a CATCH form, which is also much more general

8:23:21 phoe so if you don't feel like passing a closure all the way to your non-local jump site, THROW might be a good option

8:36:44 fiddlerwoaroof phoe: I needed to figure out a good way to avoid doing extra work with taking here: https://fwoar.co/pastebin/3cf12f4c0037cce4021a1da07bf2da76051bc425.lisp.html

8:37:27 phoe what do you mean, extra work?

8:37:33 fiddlerwoaroof Obviously, with the LABELS implementation I can just use a block, but this is just for convenient prototyping

8:38:02 fiddlerwoaroof The lines beginning with `>>>` show every invocation of the inner function in TAKING

8:38:22 fiddlerwoaroof TAKING is supposed to only process N elements of the collection being worked on, and ignore the rest

8:38:47 phoe is the second code block good then? the outut seems consistent with what is processed

8:39:05 fiddlerwoaroof Yeah, THROW allowed me to implement EXIT-EARLY

8:39:11 phoe yes

8:39:37 phoe the only other option is to pass a closure all the way to TAKING, which might or might not be viable in your case

8:39:44 phoe or use conditions, which will be slower.

8:39:52 phoe so THROW seems like a decent fit in this case

8:40:14 fiddlerwoaroof Yeah, I thought of conditions but those also just feel semantically wrong for this

8:40:27 fiddlerwoaroof maybe a signal that doesn't inherit from condition would be ok

8:40:36 phoe you mean from warning/error?

8:40:49 phoe all conditions, by definition, inherit from type CONDITION

8:41:03 phoe and those non-error ones would be semantically valid to use here

8:41:05 fiddlerwoaroof Sorry, I think I was confused

8:41:20 phoe you signal a situation that you found a match of some sort

8:41:35 fiddlerwoaroof Yeah, I meant using SIGNAL instead of ERROR/WARN

8:41:40 phoe and it's up to the code around this call to handle or ignore it. this includes optionally performing a non-local jump.

8:42:13 phoe so using HANDLER-CASE over SIGNAL would be equivalent to THROW/CATCH, but also very likely it would be slower because of walking the handler stack and performing runtime type checks.

8:42:36 phoe see https://gist.github.com/nikodemus/b461ab9146a3397dd93e

8:42:49 fiddlerwoaroof Yeah

8:43:10 phoe so you're good with THROW/CATCH in this case, it seems like the proper primitive for bailing out early in this case.

8:44:15 fiddlerwoaroof Yeah, I just never actually think about it

8:44:37 phoe well now you did think about it, most likely because you actually found a use case for it

8:44:49 phoe and CL delivers. :D

8:45:08 fiddlerwoaroof One reason I like CL is that every programmer/team has their own dialect

8:45:22 fiddlerwoaroof This is true in most languages, but I find the variations in CL more interesting

8:49:53 phoe that's both a boon and a challenge

8:50:08 phoe synchronizing different minds is real tricky in Common Lisp because of that trait.

8:56:31 no-defun-allowed Hypothetically, you at least have a definition of all the dialect-y stuff in a common language, which may not be the case for human languages.

9:29:22 ldb good evening

9:29:35 beach Hello ldb.

9:31:35 ldb my spell check program prototype takes very long time to input a dictionary of 10,000 words

9:32:03 phoe I remember that beach and I collaborated on a spellchecker once; https://github.com/robert-strandh/Spell

9:32:25 beach ldb: I suggest you save as a Common Lisp data structure so that you can just do a READ.

9:34:36 ldb it is due to the data structure I use requires a lot of comparison between nodes when build from a list of sorted words

9:34:49 phoe ldb: what is your data structure?

9:35:10 beach ldb: So don't build it every time you load it.

9:35:42 beach ldb: Build it once from the source data, and save the built data structure using PRINT.

9:36:42 phoe beach: currently, Spell saves it to a FASL. It has no facilities for printing that data.

9:36:44 Nilby as you may known, things with lists can unintentionally get O(n^2) or worse very quickly

9:37:23 ldb it requires 2 minitues on a test input of 10,000 words, but my actual input is 500,000 words, and building from larger dictionary takes extra comparsions

9:38:28 beach ldb: Maybe you have some non-linear algorithm in your build code.

9:38:43 ldb phoe: https://www.researchgate.net/publication/220355507_Incremental_Construction_of_Minimal_Acyclic_Finite-State_Automata

9:39:47 ldb beach: yes, I realized every time a new node is added it has to be compared by almost every existed nodes, to minimize the data structure

9:41:28 beach Not surprising then.

9:45:05 phoe that's painful - and O(n²)

9:46:42 ldb ok, instead of a naive list, I can keep a lookup table with a bool and the number of children as index keys to do a pre-filtering

9:49:29 ldb is there a quick way to compare two assoc lists? the order doesn't matter, then only requirement is that each pair with #'eql keys should be has #'eq values

9:49:48 ldb and the number of pairs are same

9:51:15 fiddlerwoaroof compare?

9:51:53 Nilby Usually if you have more than about 12 elements in a alist it will be faster with a hash table.

9:52:11 fiddlerwoaroof You're going to have to check every key/value pair, no?

9:52:20 fiddlerwoaroof It'll be O(n) regardless

9:53:02 ldb *check equality

9:53:29 ldb Nilby: mostly 1~5 pairs

10:13:28 liberliver1 ** NICK liberliver

10:28:53 Nilby Hmm. I have 1gb databse with 10m words that takes about 20 secs to load and full regexp search, and I thought it's way too slow.

10:31:07 ldb 1gb takes long time to do IO

10:52:51 Nilby The raw IO for 1gb only takes about 1.6 secs for me, and that's longer than it should.

10:53:35 Nilby (on an old laptop, but with an ssd)

12:19:15 liberliver1 ** NICK liberliver

13:17:15 marcoxa Just pitched in some Euros for CLCS :)

13:17:29 phoe <3

13:18:09 marcoxa I swear. It'a a gift for my mother in law :)

13:19:43 phoe does your mother in law know Lisp?

13:22:46 marcoxa she know how to order people around.

13:27:08 puchacz hi, is hash-table in abcl synchronized?

13:27:34 surabax_ ** NICK surabax

13:39:24 Nilby puchacz: It looks synchronized to me from reading the code, but I haven't tried it. If you want your code to work on other implementations you could wrap in bordeaux-threads:with-lock-held .

13:39:34 puchacz Nilby - thanks

13:39:46 puchacz the manual does not say anything, but the code is better than manual

13:40:01 puchacz and yeah, wrapping with bt will solve ambiguities

13:42:06 Nilby The problem is it's still a race condition if your code does a (gethash val table) and then later does a (setf (gethash val table) ..) presuming the results of the previous gethash are still valid.

13:43:02 puchacz yes

13:47:48 Nilby I looked in abcl/src/org/armedbear/lisp/HashTable.java in case you want to check.

13:48:00 puchacz tks

14:09:20 phoe puchacz: hmmmm, maybe easye can answer that

14:11:39 puchacz phone - if undocumented, for more serious work I would rather wrap it

14:11:50 phoe so would I, yes

14:12:12 phoe but still, maybe that needs to be clarified in the abcl docs.

14:14:38 puchacz sorry, Michał, typo, "phoe" of course :)

14:21:41 phoe :D

14:31:06 nwoob is there a better programming experience for CL from emacs?

14:31:21 beach Better than what?

14:31:27 nwoob better than emacs

14:31:39 beach Oh, better THAN Emacs?

14:31:46 beach Not at the moment I think.

14:31:55 beach But we are working on it. :)

14:31:58 phoe nwoob: depends on your definition of "better"

14:32:02 beach Give us few more years.

14:32:16 phoe there are swank/sly clients for VSCode, Atom, and Sublime Text

14:32:21 phoe there's also slimv/vlime for Vim.

14:32:32 nwoob I looked how to compile CL program and found C-c C-k to evaluate buffer, but C-c C-k is undefined

14:32:50 beach Are you using SLIME?

14:32:53 nwoob yes

14:33:25 nwoob I have a scratch opened and it says Lisp interaction at the bottom right

14:33:53 nwoob in there i'm typing CL code and trying to compile the function so that I can use that function in aline-repl

14:33:59 beach No, it should be a buffer *slime-repl...*

14:34:29 phoe you should be able to type something in the scratch buffer, like (defun foo (x) (+ x 42)) and then hit C-c C-c on it

14:34:43 phoe it should get compile-and-loaded and you should be able to call (foo 24) in the REPL.

14:35:10 nwoob phoe: that's what I'm trying to do. but C-c C-k is undefined

14:35:39 beach nwoob: That command is valid in a Lisp source code buffer.

14:35:40 phoe in the lisp buffer, M-x lisp-mode <RET>

14:35:55 beach nwoob: In the REPL buffer, just type forms.

14:41:13 marcoxa Where can I find a link for swank/slime for VS Studio 2019?

14:41:38 nwoob Turns out I need to save scratch buffer code. I saved code in test.lisp and then did C-c C-k and it is working

14:41:44 phoe marcoxa: https://marketplace.visualstudio.com/items?itemName=rheller.alive

14:41:51 phoe nwoob: oh, good

14:42:13 nwoob I thought I could evaluate buffer without saving

14:42:21 phoe only individual expressions within a buffer

14:42:31 phoe C-c C-c should work in there

14:42:40 phoe but C-c C-k requires a saved file.

14:42:46 nwoob in scratch buffer, right

14:42:51 marcoxa Thanks phoe

14:42:58 nwoob which emacs puts by default in lisp mode

14:44:05 hendursaga Besides SBCL, what other implementations have the disassemble function disassemble to assembly language, not bytecode or C?

14:45:28 phoe CCL?

14:46:10 puchacz btw, why do people choose CCL over SBCL?

14:46:22 phoe blazing fast compiler compared to SBCL's

14:46:38 phoe also a nice macOS IDE built-in.

14:46:45 X-Scale` ** NICK X-Scale

14:46:57 phoe and AFAIK some Cocoa/ObjC bindings that SBCL does not have.

14:47:30 Nilby CCL aslo has no memory limits, and can call C code with structs by value without libffi