freenode/#lisp - IRC Chatlog

7:46:39 beach markasoftware: I think NCONC is specified to modify the list, unless it is the empty list of course.

7:47:05 beach aggin: Thanks!

7:47:32 beach Hello ldbeth.

8:03:10 ck_ I believe the regex quote is in the unix hater's handbook as well

8:03:25 moon-child markasoftware: nconc is defined in terms of rplacd, so it would have to be in-place

8:03:33 ck_ but it might be for a specific tool, like sed

8:08:13 ldbeth sometimes it is easier to write a proper parser

8:13:01 moon-child imo raku (née perl6) has the best regex implementation out there

8:13:52 moon-child not least because it's really well integrated with the rest of the language. And whitespace isn't significant, so you don't have to squish all the code together

8:13:56 moon-child ACTION wishes for something like that in cl

8:15:23 aggin what would be the best implementation of CASE that uses string= and also supports the OTHERWISE key ?

8:15:54 flip214 aggin: ALEXANDRIA:SWITCH

8:16:07 aggin flip214: ok will check it out

8:26:53 splittist You can break a (cl-ppcre) regex into mulitple lines and provide line comments. That can reduce the write-only-ness of complex mangling.

8:49:36 semz moon-child: cl-ppcre has an alternative sexpr syntax, though I find it pretty cumbersome and idiosyncratic personally

8:50:44 semz I can really go without writing :non-greedy-repetition over and over to get optional matches

9:08:50 moon-child https://github.com/logicmoo/wam_common_lisp this is interesting

9:34:42 SAL9000_ ** NICK SAL9000

10:06:04 antoszka_ ** NICK antoszka

10:33:03 cods_ ** NICK cods

12:24:00 deltab_ ** NICK deltab

12:36:45 phoe am I screwing something up here, or is it cheaper to sort lists than arrays on SBCL? https://plaster.tymoon.eu/view/2334#2334

12:38:44 flip214 phoe: you're not mistaken.

12:38:52 flip214 I was wondering about that too.

12:40:07 flip214 phoe: though I'm trying to be at or above one second for TIME, to reduce noise a bit.

12:41:00 phoe sure, let me try with 3M and bear with the fact that SHUFFLE is quadratic for lists...

12:41:51 flip214 phoe: phoe: well, put the shuffled data into a special and do that only once

12:42:29 phoe eh, it's not *that* bad; still finishes in under a minute

12:42:31 mfiano Well you're not timing that, so just put the shuffle on the array

12:48:58 Nilby I think it's because it with the element type it actually has to do more copying. And quicksort may have a slight throughput advantage to mergesort.

12:51:30 phoe flip214: https://plaster.tymoon.eu/view/2335#2335

12:52:36 Xach i was looking for this, maybe you were too: https://web.archive.org/web/20200212080133/http://home.pipeline.com/~hbaker1/

12:53:31 Nilby I imagine there might be some way to get that mergesort a smidgeon faster, by maybe having an architecture specific version. But of course the advantages of parallelizing are likely greater.

12:53:47 phoe ACTION adds to minion

12:53:50 phoe minion: tell Xach about baker

12:53:50 minion Xach: look at baker: https://web.archive.org/web/20200212080133/http://home.pipeline.com/~hbaker1/

12:54:08 Xach i know about baker! i told *you* about baker! what is this!

12:54:21 phoe you just told me about baker so now I know about baker!

12:55:07 phoe Xach: https://i.imgflip.com/1xvnfi.jpg

12:58:31 adeht ** NICK _death

13:00:19 _death flip214: I submitted a PR for the shuffle fix a while ago, maybe you missed it

13:05:53 phoe _death: on the clnet gitlab?

13:06:22 _death phoe: no, on github.. we already had a discussion about that

13:06:40 phoe OK, looking for that PR now...

13:09:47 phoe found it and recreated at https://gitlab.common-lisp.net/alexandria/alexandria/-/issues/21

13:10:49 _death my opinion is that the answer to both your questions is "no"

13:12:49 phoe I assume you meant the 2nd and 3rd questions

13:12:53 _death yes

13:13:15 phoe I think the same way, but my nitpickish nature told me to nonetheless put the questions there

13:13:58 phoe better to voice my worries out loud even if they are only going to be read by a single random person 10 years from now whose old code blew the heap because of this change

13:14:25 _death if you have a large list, you already lose with that shuffle

13:14:54 phoe yep, I can fully agree with that

13:15:24 _death so maybe that person is still waiting for it to finish, and in the meantime won't use the new shuffle

13:17:27 _death (by the way, by "large" I mean really small)

13:20:20 _death if only it performed the shuffling on disk, like they used to merge sort "big data" in the old days...

13:20:40 flip214 _death: oh sorry, got no notification? phoe: thanks, will take a look

13:21:16 _death flip214: interesting.. I also had something like that happen to me, and only noticed a PR by chance

13:22:38 Nilby Also of course the performance on a random value straight consed list is slightly faster than shuffled

13:23:37 Nilby But what's not to like about a linear shuffle?

13:26:55 _death if you can do that, sure.. often, I find that I need shuffle

13:29:50 _death many alexandria functions are good for convenience but may not be the most efficient for your particular use-cases.. e.g., I remember having to implement a map-iota-permutations function that's similar to alexandria:map-permutations but uses a compact representation, given that the sequence is [0-n)

13:33:50 _death flip214: I guess the reason for no notification is that your alexandria repo on github is a "fork"

13:34:36 _death flip214: so since you're maintaining it, you may want to recreate it as an original

15:13:22 jmercouris what symbol would you use to designate a function as a implementation detail

15:13:27 jmercouris like %salmon

15:13:28 jmercouris or?

15:14:11 jackdaniel %caribe

15:14:15 jackdaniel they are dangerous

15:15:11 jackdaniel the usual way to designate a function as an implementation detail is to not export it

15:16:01 jackdaniel %foo denotes "dangerous"

15:16:01 jmercouris I don't want to say it is dangerous

15:16:12 jmercouris I just want to say it exists for this insular purpose

15:16:15 jmercouris it is not really meant to be used

15:16:20 jmercouris outside of this context

15:16:24 phoe in my headcanon foo:bar is public, foo::bar is private, foo::%bar is an invocation to cthulhu and should never be used

15:16:25 jmercouris like you can, nothing bad will happen

15:16:38 jackdaniel then don't use %

15:16:41 jmercouris OK

15:16:44 jmercouris how about this

15:16:50 jmercouris let's say we have something called process-salmon

15:16:57 jmercouris and it processes a single salmon

15:17:02 phoe is that exported?

15:17:07 jmercouris then we have another function that processes multiple salmon

15:17:25 phoe how does this second function differ from mapcar #'process-salmon?

15:17:27 jmercouris the only reason we need the function that processes multiple salmon is because the caller expects to be able to pass a list of salmon to be processed

15:17:41 jmercouris it MAY or may not differe from (mapcar #'process-salmon

15:17:52 phoe what is the difference, then?

15:17:58 jmercouris for example, sometimes we might only (process-salmon (first salmon))

15:18:06 jmercouris I'm just wondering how to name these class of functions

15:18:25 jmercouris because you see the problem right? salmon is the singular and plural

15:18:36 jmercouris so my thought was process-salmon*

15:18:42 jmercouris to indicate the plural of salmon

15:18:42 phoe oh, I see

15:19:24 phoe I'd avoid trouble and use a singular process-salmon and indicate this in the documentation/docstring, and ask the users to mapcar/mapc #'process-salmon if they want to try and process multiple

15:19:26 jackdaniel how about process-one-%caribe;drop table users;

15:19:50 jmercouris phoe: I know, that is what I would normally do

15:19:54 jmercouris phoe: can I show you a concrete example?

15:19:59 phoe sure

15:20:04 jmercouris one moment please

15:20:41 phoe is a single salmon allowed to be a list, or not? if that is the case, you can make a very simple GF of typecase that accepts either a non-list singular salmon or a list of salmon

15:20:50 jmercouris phoe: http://dpaste.com/8V4NMDJX9

15:20:58 phoe that would be kinda sloppy, but would in turn work around the sloppiness of english

15:21:03 jmercouris here is the problem, the prompter:actions ALWAYS specify a list to return

15:21:05 jmercouris it is part of our API

15:21:18 jmercouris and therefore any action MUST handle a list of objects

15:21:41 jmercouris therefore I need a sort of 'mini wrapper' to just unwrap the list, I do NOT want to put that logic into scroll-page-to-heading

15:22:06 jmercouris I would end up having to write a (if (listp heading) (first heading) heading) at the top

15:22:15 phoe alexandria:ensure-car

15:22:19 phoe but, yes, I see the issue

15:22:26 jmercouris yeah, that's fine, I could do thta

15:22:31 jmercouris s/thta/that

15:22:33 jmercouris but still

15:22:58 jmercouris I don't think it is a wise idea to overload the value of heading to accept EITHER a list or a single value

15:22:58 phoe this sounds like a question of API design - are your functions specified to accept a single object, a list of objects, or both?

15:23:08 jmercouris it /depends/

15:23:19 phoe I'd grab any of those choices (preferably the first, but that's my preference) and apply that consistently

15:23:46 phoe if that is impossible, then specify in the function contract/documentation/docstring what is accepted by a given function - a single object, or a list, or both - and then use that

15:24:14 phoe and then refactor as much as possible in order to aim for eventual consistency in the API

15:25:18 jmercouris hm

15:25:27 jmercouris interesting thoughts

15:26:20 [df]_ ** NICK [df]

15:27:23 jmercouris thanks for your feedback phoe

15:27:59 jmercouris well we have decided that the API should return a list /always/

15:28:10 jmercouris the problem is that many of our functions do not yet respect this, and they accept only single values

15:28:19 jmercouris I guess ensure-car could be a band-aid...

15:28:27 jmercouris or making the inline-command as I have done

15:28:30 phoe hey wait a second

15:28:35 phoe return a list, sure

15:28:39 phoe but we are not talking about returning values

15:28:43 phoe we are talking about *accepting* values

15:28:47 jmercouris yes

15:28:48 phoe that's the other end of the function call

15:28:53 jmercouris the functions do NOT ACCEPT values, correct

15:28:58 jmercouris sorry, a LIST

15:29:16 phoe so you have a lot of functions that return lists of things

15:29:33 jmercouris only one function that returns a list, but it is used to gather user input

15:29:40 jmercouris and this function is used extensively

15:29:51 phoe I am confused now

15:29:55 phoe < jmercouris> well we have decided that the API should return a list /always/

15:30:00 phoe I thought that you meant all of your API

15:30:10 phoe what do you mean by "API" here?

15:30:11 jmercouris phoe: yes, when calling (prompt "give me some input") prompt will ALWAYS return a list

15:30:23 jmercouris a list of things the user has inputted

15:30:49 phoe okay, and what is the problem?

15:31:01 phoe you have a single function that is used a ton of times, and it always returns a list

15:31:13 jmercouris yes

15:31:20 jmercouris it did not ALWAYS return a list, in the past it would return a single value

15:31:31 jmercouris and many functions are built around this concept

15:31:40 jmercouris so for example, the one I showed you

15:31:47 jmercouris "scroll-page-to-heading", it expects a single heading

15:31:52 phoe oh, I see

15:31:56 jmercouris it doesn't make sense that one could scroll to multiple headings

15:32:06 jmercouris you can only scroll to one, it is a mutually exclusive operation

15:32:14 jmercouris you can scroll to several in a row, but only one at a time

15:32:28 phoe then call it like (scroll-page-to-heading (first input))

15:32:39 phoe that's what I would call the clean way

15:32:40 jmercouris so the way that I've written it in my example :-D

15:32:54 phoe yes, I understand your example now

15:33:03 jmercouris OK, it is a good reassurance

15:33:07 phoe this way it's clear that HEADINGS is always a list, even if it is a singleton list

15:33:09 jmercouris and I did not know about ensure-car anyways

15:33:14 phoe so at least we have type issues out of the way

15:33:18 jmercouris indeed

15:33:21 phoe but!

15:33:28 phoe you won't need ensure-car if HEADINGS is always a list

15:33:36 phoe you can just use CL:FIRST all the time instead

15:33:47 phoe ENSURE-CAR is useful in situations when you *maybe* have a list

15:33:50 jmercouris OK here is where it gets tricky

15:33:54 jmercouris I do want ENSURE-CAR

15:33:58 jmercouris because someone else could use the function

15:34:03 jmercouris and they just want to pass a single value

15:34:07 phoe then they fucked up

15:34:08 jmercouris a single heading

15:34:10 phoe and they need to fix up their code

15:34:15 jmercouris well yes, but

15:34:23 jmercouris it is counterintuitive to pass a list when we only want the first value /always/

15:34:30 jmercouris so I wouldn't blame them

15:34:36 jmercouris even if the spec shows something else

15:34:55 phoe hmmmmmm

15:34:59 phoe is this an internal function?

15:35:04 jmercouris yes and no

15:35:09 phoe why no?

15:35:10 jmercouris all functions in nyxt that are commands are external

15:35:16 jmercouris we export all commands

15:35:25 phoe okay, so in this case, SCROLL-TO-HEADING* that accepts a list of headings is exported

15:35:26 jmercouris you see that this is not a command, it is a regular defun

15:35:35 jmercouris it is not exported

15:35:52 jmercouris however, we encourage users to muck around even with unexported things

15:35:52 phoe oh, wait, SCROLL-PAGE-TO-HEADING - is it a command? it's just a function

15:35:58 jmercouris it is jus ta function

15:36:02 phoe I see MAKE-COMMAND in L6, I assumed this is what we are making

15:36:08 phoe > however, we encourage users to muck around even with unexported things

15:36:10 jmercouris we are making a command, yes

15:36:13 phoe ouch

15:36:20 jmercouris :-) it will bite us yes

15:36:21 jmercouris at some point

15:36:28 phoe well then, you should encourage users to fix their code whenever you change stuff, too

15:36:44 phoe and then move fast and break stuff fearlessly

15:36:46 jmercouris yes

15:36:56 jmercouris so the exported thing in this case is scroll-to-heading*

15:37:01 jmercouris which ONLY accepts a list

15:37:08 jmercouris you see that it does (first headings)

15:37:14 phoe yes, I see

15:37:15 jmercouris so the user of said function could very well get it wrong

15:37:38 phoe in that case I can understand the use of ENSURE-CAR

15:37:38 jmercouris thats why I wanted some convention or something to show it

15:37:46 phoe because that function is supposed to DWIM, in a way

15:37:49 jmercouris right, I could just add ensure-car to the other function

15:37:56 jmercouris I could ensure-car in scroll-page-to-heading

15:37:57 phoe and you kinda use it in both contexts

15:37:58 jmercouris and it would be cleaner that way

15:38:12 phoe sure

15:38:52 phoe ;; and then you could just (make-command scroll-to-heading* scroll-page-to-heading) if you're willing to changing your MAKE-COMMAND macro to accept a function name instead of a lambda list and body

15:39:15 phoe but then! are all of your commands single-argument functions?

15:39:21 jmercouris No

15:39:29 jmercouris Only some

15:39:36 jmercouris That’s the thing :-D

15:39:53 jmercouris I think ensure car might be the way to go here

15:40:09 phoe well then which one of them accept lists and which one of them don't accept lists, because things are getting real sloppy

15:40:13 phoe and I don't want to be the one maintaining it

15:41:05 jmercouris All of them should gracefully accept lists

15:41:15 jmercouris That’s what we decided

15:41:41 jmercouris And I can either wrap them as ive done with (first xyz) or use ensure car within the body of the actual function

15:41:54 jmercouris What should I do?

15:42:18 phoe accept lists as which arguments

15:42:24 phoe first, second, third, nth?

15:42:29 jmercouris doesn't mater

15:42:33 jmercouris it is always the first

15:42:37 jmercouris s/mater/matter

15:42:39 phoe oh, I see

15:42:50 jmercouris so IDEALLY

15:42:56 phoe then ensure-car all the things and pray that you never get a list as a valid argument

15:42:58 jmercouris scroll-page-to-heading should accept a list OR a value

15:43:13 jmercouris and like I said I can either wrap it (as I've done)

15:43:15 phoe because if (1 2 3) is a valid argument to one of your functions then (ensure-car '(1 2 3)) ;=> 1

15:43:17 jmercouris or use ensure-car within its body

15:43:30 phoe and so your users will need to work around your sloppiness and manually wrap their lists

15:43:41 jmercouris I'm not sure

15:43:46 phoe so they'll need to call it like (do-command '((1 2 3)))

15:43:48 jmercouris when would they need to manually wrap their lists?

15:43:58 phoe when a command would normally accept a list of things

15:44:10 jmercouris let's say I write a new function

15:44:13 jmercouris download-links

15:44:18 jmercouris as you can imagine, this accepts a list of links

15:44:20 phoe yes

15:44:23 jmercouris I'm not going to write ensure-car within download-links

15:44:26 jmercouris that woud be nonsensical

15:44:35 jmercouris so the user can just do (download-links '(link1 link2))

15:44:42 jmercouris there is no manual wrapping

15:45:02 phoe then what about (download-links 'link1)

15:45:20 jmercouris yeah, that's not possible

15:45:24 jmercouris should it be possible?

15:45:34 phoe no idea, you tell me

15:45:43 jmercouris well, the name is plural

15:45:48 jmercouris so no, it should not be possible

15:45:56 jmercouris if if it was (download-link 'link1) then it would be fine

15:46:00 phoe it's weird to me, because (download-links '(a b c)) operates on A B C whereas (scroll-to-heading '(a b c)) operates on A and nothing else

15:46:11 jmercouris yes, that is inconsitent

15:46:22 jmercouris while it is true that (scroll-to-heading '(a b c)) is technically correct

15:46:27 jmercouris we are always expecting a list of a single value

15:46:43 jmercouris the only reason it is a list is becuase of PROMPT

15:46:54 jmercouris no other reason, prompt ALWAYS returns a list of values that the user selects

15:47:18 jmercouris if I could declare a type of a list of a single value

15:47:19 jmercouris I would

15:47:25 jmercouris then it would be clear

15:47:27 phoe you can

15:47:32 phoe (cons t null)

15:47:39 phoe this is a proper list containing exactly one value

15:47:39 jmercouris ah

15:47:42 jmercouris I did not think of that

15:47:46 phoe s/a proper list/type of a proper list/

15:48:09 jmercouris is the problem clear?

15:48:16 jmercouris it does not make sense to jump to a list of headings

15:48:20 jmercouris but prompter ALWAYS gives a list

15:48:44 phoe I'd honestly always explicitly unwrap this list whenever I want to get its first element

15:48:53 jmercouris that is what I will do

15:48:56 phoe and signal an error if the list contains zero or two+ elements

15:49:00 jmercouris with scroll-to-heading*

15:49:01 phoe just for type safety

15:50:48 phoe because if prompter is allowed to return a list of N elements (including one), then we need to always verify if we get a list of exactly 1 element

15:50:59 phoe then extract it via CL:FIRST, and then operate on it

15:51:07 jmercouris yes

15:51:15 phoe at least that's what I'd do to keep it as magic-less as possible

15:51:17 jmercouris that is why we do not use it directly, and wrap it

15:51:19 phoe because magic means bugs

15:51:26 jmercouris indeed, there is no magique aqui

15:51:45 jmercouris you see in my example, I wrap it and extract the first function quite explicitly

15:51:51 phoe yes

15:51:52 jmercouris s/function/value

15:52:00 phoe and that's how I'd do it, I guess

15:52:23 jmercouris 1

15:52:25 jmercouris thank you phoe

15:52:43 jmercouris I'll be sure to curse your name some years from now when it all blows up :-D

15:52:47 phoe but, just to be sure

15:53:01 phoe make-command scroll-to-heading* is the thing that is going to be user-facing, right?

15:53:09 jmercouris that's correct

15:53:14 splittist Just stick :before and :after methods to ensure-car all the things. What could go wrong?

15:53:20 phoe splittist: god no

15:54:25 phoe jmercouris: I'm actually thinking of one more thing

15:54:35 jmercouris OK, what?

15:54:39 phoe what is your prompter like?

15:54:46 jmercouris what do you mean?

15:54:47 phoe does it always allow one to input multiple values?

15:54:53 jmercouris sometimes

15:54:58 phoe what does it depend on?

15:55:03 jmercouris if you invoke it with multiple-selection-p

15:55:12 jmercouris now, sorry

15:55:14 jmercouris that's not the full story

15:55:20 jmercouris a prompter contains multiple sources of information

15:55:32 jmercouris each source decides if multiple-selection-p is allowe

15:55:40 jmercouris and each source has its own list of actions

15:56:03 phoe my naïve attempt to rewrite this would be https://dpaste.com/3QFEMD9AL

15:56:17 phoe here I explicitly tell the prompter that I want single selection

15:56:22 phoe and so my function accepts only one argument

15:56:32 phoe that's likely not going to hold in practice because of what you just said

15:56:38 jmercouris phoe: yes

15:56:44 jmercouris I wouldn't need to wrap anything in your example

15:57:02 jmercouris I would just say (prompter:actions (list 'scroll-page-to-heading))

15:57:24 jmercouris the actions list can be a command, function, or symbol

15:57:42 phoe but these multiple sources of information in the prompter, they happen one after the other, right?

15:57:52 jmercouris they happen in parallel

15:57:58 phoe how does it work?

15:58:01 jmercouris the prompter queries each source simultaneously

15:58:23 jmercouris so if I say (prompt "select tomato?" source1 source2 source3)

15:58:26 phoe does it mean that I can get three popups at the same time, each asking me for different stuff?

15:58:27 jmercouris and source1 is some farmer in italy

15:58:33 jmercouris source2 is some farmer in france

15:58:37 jmercouris source3 is some farmer in spain

15:58:48 jmercouris the prompter will call up each farmer and ask them for a list of their tomatoes

15:58:57 jmercouris maybe the spanish farmer only sells a single tomato at a time

15:59:05 jmercouris but maybe the french farmer will allow you to buy 3 tomatoes

15:59:13 jmercouris the user selects the tomato from the selection and presses RET

15:59:27 phoe so it merges these sources of information in some way and displays a merged list of all them that the user can select from?

15:59:30 jmercouris at this point the prompt will gather up all of the tomatoes you've selected and return them

15:59:37 jmercouris phoe: yes, exactly

15:59:41 jmercouris let me show you a screenshot

16:00:24 jmercouris phoe: https://i.imgur.com/xzvyZo3.png

16:00:34 jmercouris there are two sources here

16:00:42 jmercouris "New URL" and "Global History"

16:00:47 phoe yes, i understand this

16:01:10 jmercouris when I press RET it will grab ALL selections from ALL sources

16:01:18 jmercouris and return them as a list

16:01:24 jmercouris even if one of the sources only allows single selection

16:01:36 phoe yes; it's not a problem of a source allowing single selection or not

16:01:54 phoe it's a problem of you not being able to configure the *prompter* to only allow a single selection across *all* the sources

16:02:01 jmercouris that's correct

16:02:12 phoe so info sources aside, you cannot tell the prompter that you want to get a single result from it

16:02:20 jmercouris not reliably, no

16:02:25 phoe and that's the architectural bug that I see here

16:02:47 jmercouris well, it is a compromise, to allow the behavior that we want

16:02:47 phoe I'd like to have a prompter that forces the user to only select one and only one thing, only to return it

16:03:10 phoe but then also has the multi-select behavior as an option to allow opening multiple URLs, for example

16:03:29 jmercouris we did have a debate about this as well

16:03:52 jmercouris https://github.com/atlas-engineer/nyxt/pull/1157

16:04:02 splittist ACTION wonders why jmercouris wants to eat parakeets

16:04:02 jmercouris starting with "When I wrote this I realized it did not work. Then I tried to do"

16:04:14 phoe currently the prompter returns very simple metadata about the values that it returns

16:04:17 phoe namely, the length of the list

16:04:47 phoe but it's not the prompter returning that IMO is the problem, it is the prompter getting invoked

16:05:06 phoe I'd like to invoke the prompter in a way that always returns one element, because then I can depend on that element being singular

16:05:24 jmercouris right, explicitly asking for a single return value

16:05:31 phoe yes, when the prompter is invoked

16:05:50 jmercouris I'm not sure that fixes the problem though

16:05:55 jmercouris ACTIONS operate on the SOURCE level

16:06:18 jmercouris hm, maybe it would work

16:06:24 phoe hm

16:06:33 phoe this has become a strangely specific discussion

16:06:38 phoe maybe let's move it to #nyxt

16:06:44 jmercouris OK

16:06:53 gitgoood ** NICK gitgood

17:37:05 mfiano So if we have a class with a few slots that is not exported, and then we have a macro

17:37:08 mfiano oops

17:37:17 mfiano MOP question: If I have an internal class that is not exported which has a few slots, and then I have a macro that is exported that the user uses to define a class that is a subclass of this class (among other details), can our internal code reliably use STANDARD-INSTANCE-ACCESS to access the inherited slots, without fear of the vector positions changing when the user redefines the subclass?

17:38:34 mfiano The internal class is never going to be changed itself, only the user's subclass, which we want to S-I-A in our internal code to access the inherited slots

17:40:21 mfiano Seems like a question Bike or beach would know best :)

18:14:32 fiddlerwoaroof ACTION just spent an hour or so because of SET-PPRINT-DISPATCH

18:14:59 fiddlerwoaroof It turns out that if you do (SET-PPRINT-DISPATCH 'string ...), you can break ASDF/UIOP

18:16:15 fiddlerwoaroof For portability, they should probably copy the pretty-print dispatch table and add it to the WITH-SAFE-IO-SYNTAX macro

18:16:26 fiddlerwoaroof (or similar)

18:17:08 fiddlerwoaroof Or, just NIL out *PRINT-PRETTY*

18:17:23 _death fiddlerwoaroof: you should use your own dispatch table.. I have a patch for slime that lets you have a repl-specific dispatch table

18:17:42 fiddlerwoaroof Yeah, the issue was that I accidentally modified the global one

18:18:17 fiddlerwoaroof So, user error, but it seems like the sort of problem a library like ASDF should think about

18:19:03 _death fiddlerwoaroof: as I wrote in that patch's commit message, their assumptions may not be warranted, but this is the real world ;)

18:43:33 waleee-cl Will a newer sbcl re-use the cached files in ~/.cache/common-lisp from the previous version on the system?

18:44:04 waleee-cl I got a package error when building nyxt that went away after wiping the cache, hence the question

18:44:12 _death waleee-cl: it's best to delete the cache

18:45:28 _death waleee-cl: sbcl's fasls have versions, but even with the same version I don't know if you can practically rely on that

18:45:39 Bike mfiano: i think you can rely on the slot locations being the same unless the user can use multiple inheritance or affect the inherited slots in certain ways

18:45:54 waleee-cl _death: yeah, I guess. It worked without wiping it for about 4 months of sbcl's straight from the git repo

18:46:05 Bike mfiano: "For a given class, the locations increase consecutively, in the order that the directly accessible slots appear in the list of effective slots."

18:46:39 mfiano Bike: I was told the MOP does not guaranteee slot locations, in that the "youngest" slots could be placed first.

18:46:47 mfiano Which would definitely be against my idea

18:46:52 Bike hmm yeah i might be misremembering

18:46:59 Bike "Direct slot definitions coming from classes earlier in the class precedence list of class appear before those coming from classes later in the class precedence list."

18:47:33 Bike so any new slots would displace the old ones

18:47:50 mfiano Does that mean superclass slots would appear first?

18:48:32 Bike no, superclasses will be later in the CPL

18:48:37 Bike well i suppose i can just test what sbcl does

18:49:48 Bike seems the slot locations are consistent. wonder what i'm misreading here

18:50:12 Bike oh, wait, i see, "The result of compute-slots is a list of these effective slot definitions, in unspecified order."

18:50:39 Bike so you can't rely on an ordering of effective slots

18:50:52 mfiano Fun times. Back to the drawing board. Thanks!

18:51:30 Bike if this is important, you could define your own metaclass with a compute-slots that does define the order/locations

18:52:21 mfiano I would have to underatand that better to see if the obsoileted instance edge case would bite me

18:52:37 mfiano I'll read more into it tomorrow I guess

18:53:06 Bike i don't think redefinition poses a special issue

18:53:39 mfiano Also I feel like I would totally screw that up. I'm not very familiar with defining custom metaclasses

18:53:54 mfiano D you have an example of doing something similar?

18:54:00 mfiano Do*

18:54:13 Bike something similar in the sense of keeping locations consistent, no

18:54:22 mfiano Fair enough

18:56:51 mfiano Bike: is beach's example here doing just that? https://common-lisp.net/~mevenson/lisp/beach/CLOS-MOP-HTML/instance-structure-protocol.html

18:57:42 Bike yes, although as you can see you have to explicitly specify the slot-order for every class, which is probably not what you want

18:57:52 Bike (also i think that example is from AMOP)

18:59:59 mfiano Really all I want is for 2 slots in a private class that is the sole direct superclass (enforced by a macro) to always by first in the storage

19:02:44 Bike sure. plenty of ways you could do that. all you need to do is make sure compute-slots always returns those two slots first

19:03:26 mfiano The example I linked to says it will not work for subclasses

19:03:32 mfiano I am not sure what must be done to do such

19:05:23 Bike it's simple. say your slots are named A and B. one thing you could do would be to (call-next-method), get your list, look for the slots named A and B, and reorder them to the head of the list, and then return that.

19:06:57 mfiano Sounds simple enough, and gives me enough of an idea to play around with it. Thanks

19:07:18 Bike no problem

19:11:28 mfiano Also, looks like there is a typo in that example

19:11:59 mfiano The test function is comparing A with A instead of A with B

19:12:58 Bike right you are.

19:13:55 mfiano Can you dumb it down for me, as far as how your suggestion of using c-n-m is different than the use of it there?

19:14:18 Bike in the example they sort the entire list with this predicate that uses the (custom) class-slot-order property

19:14:32 Bike you'd want to just... you know what i'll just write it out

19:14:53 mfiano Ah, hmm, ok

19:15:59 Bike (defmethod compute-slots ((class whatever))(let* ((prev (call-next-method))(a-slot (find 'a prev :key #'slot-definition-name))(b-slot (find 'b prev :key #'slot-definition-name)))(list* a-slotb-slot(remove a-slot (remove b-slot prev)))))

19:17:25 mfiano Ok, reformatting and parsing

19:20:23 Kingsy https://github.com/xach/buildapp <- is this considered the easiest way of buildign an executable when using commonlisp ?