🦋 Welcome to the MAIN() IRC channel of the Raku Programming Language (raku.org). Log available at irclogs.raku.org/raku/live.html . If you're a beginner, you can also check out the #raku-beginner channel!
Set by lizmat on 6 September 2022.
00:00 reportable6 left 00:02 reportable6 joined 00:05 rf joined 00:29 ProperNoun joined, epony left 00:36 lichtkind left
leont Skarsnik: you can work around the dot in @INC issue by setting the PERL_USE_UNSAFE_INC environmental variable 00:48
01:04 xinming left, xinming_ left 01:05 xinming joined 01:06 xinming_ joined
Nemokosch a fun little task (I hope): 01:28
what does the following snippet achieve?
$descriptor.comb.skip >>ne>> '-' andthen 01:29
$_>>.Int.join.parse-base(2).base(8)
inb4 I'm absolutely sure there can be corner cases this doesn't account for, but if you can notice that, you probably won 😛
01:50 deoac joined 01:59 razetime joined 02:02 razetime1 joined 02:03 razetime left, razetime1 is now known as razetime 02:38 epony joined 03:27 rf left 03:56 deoac left 04:37 sftp left 04:38 sftp joined 05:04 [Coke]_ joined, [Coke] left 05:16 alfonsox joined 05:44 Kaiepi left 06:00 reportable6 left 06:03 reportable6 joined 07:03 greppable6 left, evalable6 left, bloatable6 left, coverable6 left, committable6 left, statisfiable6 left, releasable6 left, bisectable6 left, benchable6 left, unicodable6 left, notable6 left, sourceable6 left, squashable6 left, quotable6 left, nativecallable6 left, reportable6 left, tellable6 left, linkable6 left, shareable6 left, bloatable6 joined, statisfiable6 joined 07:04 committable6 joined, tellable6 joined, evalable6 joined, releasable6 joined, greppable6 joined, linkable6 joined, shareable6 joined 07:05 notable6 joined, quotable6 joined, reportable6 joined, nativecallable6 joined, benchable6 joined, coverable6 joined, squashable6 joined 07:06 sourceable6 joined, bisectable6 joined, unicodable6 joined 07:13 Homer_Simpson joined 07:19 abraxxa joined 07:23 abraxxa left 07:25 abraxxa joined 07:28 abraxxa1 joined 07:29 abraxxa left 07:40 xinming left, xinming joined
jaguart nemokosch: This is the updated grok of a compunit gist.github.com/jaguart/9c69db9d8d...c379756dcf 07:53
the CompUnit contains a Distribution and a set of namespaces aka Packages and Modules. In this example Foo is a unit module, with two roles and two classes 07:55
One interesting thing I discovered is that auth/ver/api attributes are only populated on top level modules. If I change from `unit module Foo:auth<me>` to `unit module Foo::Bar:auth<me>` the :auth is no longer present/available (or at least I cant spot it in the MOP) 07:58
08:13 Sgeo left
Skarsnik nvm, I still got the multiple redefinition of symbol error for mid 2018 to 2019 09:09
09:13 dakkar joined
2020.02 fixed this x) 09:24
09:25 jpn joined 09:26 sena_kun joined 09:33 Homer_Simpson left 09:35 lichtkind joined 09:37 xinming left 09:57 xinming joined 10:05 sena_kun left 10:10 sena_kun joined, sena_kun left 10:20 grondilu joined 10:44 Kaiepi joined
grondilu any reason `reduce * ~^ *, ` would not be the same as `[~^]` ? I can't replace the former with the latter in github.com/grondilu/pbkdf2-raku/bl...F2.rakumod 11:01
nine jaguart: I think it depends more on how a package gets created whether it inherits those attributes or not. Support was just added to one code path. Grep for pkg_create_mo in rakudo src 11:02
grondilu Also unfortunately I can't reproduce the error in a simple one-liner.
(I haven't tried very hard though, ngl) 11:03
Nemokosch grondilu: well, what happens in one case and another? 11:06
11:09 sena_kun joined 11:13 jpn left
grondilu Nemokosch: It works fine with `reduce` and I get a stringification of Blob error with `[ ]` 11:22
tellable6 grondilu, I'll pass your message to Nemokosch
dakkar grondilu: random thought: `($seed, { foo } ... *)[1..$c]` may be better written `({ foo } ... *)[^$c]` 11:23
(this does not actually relate to your question)
grondilu dakkar: noted 11:24
11:26 jpn joined
grondilu dakkar: actually, no. I don't want the first element of the sequence, and &foo does take an argument so you can't just remove $seed. 11:28
dakkar grondilu: I'm not familiar with pbkdf2, what would be some arguments to run it with?
aaaaah, sorry, I had mis-parsed the sequence operator!
grondilu you can use the synopsis in the README 11:29
github.com/grondilu/pbkdf2-raku
dakkar does that
grondilu if you don't want to install MD5 you can make up a dumb prf function. It just needs to take two args that can be converted into blob8 and it must return a blob8 11:31
dakkar uh… I have replaced the line `reduce * ~^ *, (` with `[~^] (` and I get the same blob in output 11:32
grondilu really? odd.
dakkar running on v2022.06
grondilu oh you're right it does work with the synopsis 11:33
I was using the other test file 11:34
dakkar installs Digest::HMAC
Nemokosch uh oh 11:35
dakkar which zef can't find…
Nemokosch wait what? that's new
grondilu Digest::HMAC is the name of a module in the Digest repo 11:36
Nemokosch not anymore though, right?
grondilu huh
oh yeah
Nemokosch that's what sparked the whole topic of module-dependencies 11:37
for what it's worth, I still can't install Digest::HMAC because of test failures 11:39
grondilu reinstalls his Digest module
Nemokosch oh right, your module used to provide Digest 11:41
now the module vs distribution polemy bit the other way around...
grondilu well now the name of the module is just HMAC. The name of the repo is still Digest. 11:43
Nemokosch I don't mean this in a bad way, just curious. Did you know that the modules you provide are essentially the API of your distribution?
grondilu Nemokosch: I am not quite sure about what you mean 11:44
tellable6 grondilu, I'll pass your message to Nemokosch
Nemokosch and hence eliminating the Digest *module* was an API-breaking change either way?
grondilu I knew it would necessarily break things in any case.
but I really wasn't happy with how the module was organized anyway. 11:45
Nemokosch although to be frank, if something didn't specify at least the api field for a dependency, it would still break...
grondilu there is an api field?
grondilu looks up docs.raku.org/language/modules 11:46
oh yeah there is
Nemokosch Yes, not sure how it would integrate with semver tbh, maybe it's just more explicit 11:47
but where I'm getting at is that when you change the provides section, that's arguably a breaking change on the biggest possible level
grondilu damn I guess this repo will need quite a bit of cleaning
I'll submit an issue refering this conversation for memorandum 11:49
Nemokosch this may be somewhat selfish of me but I think all these precedents are very good for eliminating the whole "recommendation-manager" logic in dependency resolution 11:50
Of course there is the path of always blaming it on the author or the user...
grondilu submitted github.com/grondilu/libdigest-raku/issues/36 11:51
Nemokosch but really, what if someone decides to break up a module while leaving the rest of the modules untouched (pretty much what you did)? That's a breaking change on compunit level so you might want to reflect that with the versions. Now, simply because versioning happens at META6 level, all your other modules will jump to a version marked as breaking, even though they could be used just like before 11:52
grondilu I should have at least changed the major version number, that's for sure. 11:54
Nemokosch And again, I don't find it constructive to just blame this on the author. Things change over time, including our knowledge. It's still a best interest of everyone to only break when necessary.
My 2 cents is, what you could have done is to still provide the Digest module, as something that just re-exports the stuff from its new location. 11:55
grondilu there is no need to blame anyone for anything. There's a reason for the disclaimer in the license.
11:55 xinming left, xinming joined
Nemokosch Well, I think it's valid to blame the language spec itself for this bizarre idea inherited from Perl. 11:57
grondilu Nemokosch: at the time I didn't know how to re-export stuff without just copy-pasting.
tellable6 grondilu, I'll pass your message to Nemokosch
Nemokosch that metadata comes from distributions, installation happens on distribution level, yet dependency resolution is tied to modules
dakkar grondilu: the problem is the `:c(1)` 11:58
change that to `:c(2)` and now both `reduce` and `[~^]` work the same
grondilu dakkar: is it? Interesting.
dakkar (I'm now trying to figure out why)
essentially, `[~^] ($thing)` works differently from `reduce * ~^ *, ($thing)` 11:59
grondilu m: say [~^] "foo".encode
camelia foo
Nemokosch why does it?
12:00 reportable6 left
dakkar m: say [~^] (Buf.new(1..10)) 12:00
camelia Stringification of a Buf is not done with 'Stringy'. The 'decode'
method should be used to convert a Buf to a Str.
in block <unit> at <tmp> line 1
dakkar m: say [~^] Buf.new(1..10)
camelia Stringification of a Buf is not done with 'Stringy'. The 'decode'
method should be used to convert a Buf to a Str.
in block <unit> at <tmp> line 1
grondilu wait it returned a Str?
dakkar no?
grondilu m: say [~^] "foo".encode # should return a blob 12:01
camelia foo
dakkar oh
grondilu m: say [~^] "foo".encode xx 2
camelia utf8:0x<00 00 00>
Nemokosch > [~^] utf8.new(102,111,111)
does return a Str
grondilu well that is wrong, I think
dakkar m: say ([~^] (Buf.new(1..10), Buf.new(1..10),)).raku;
camelia Buf.new(0,0,0,0,0,0,0,0,0,0)
dakkar no, it's a buf, `say` calls `.gist` on it that may stringify it 12:02
Nemokosch if I do my &op = * ~^ *; [[&op]] utf8.new(102,111,111) - that returns utf8
12:02 reportable6 joined
grondilu say doesn't stringify a blob, that's precisely what the error message said in the first place 12:03
m: say "foo".encode
camelia utf8:0x<66 6F 6F>
grondilu hang on
that's an utf8 blob. I guess that's special 12:04
dakkar m: say (reduce * ~^ *, ( Buf.new(1..10) )).raku;
camelia Buf.new(1,2,3,4,5,6,7,8,9,10)
Nemokosch utf8 is a specific blob
dakkar m: say ([~^] ( Buf.new(1..10) )).raku;
camelia Stringification of a Buf is not done with 'Stringy'. The 'decode'
method should be used to convert a Buf to a Str.
in block <unit> at <tmp> line 1
grondilu ^here
dakkar m: [~^] ( Buf.new(1..10) )
camelia Potential difficulties:
Useless use of [~^] in sink context
at <tmp>:1
------> <BOL>⏏[~^] ( Buf.new(1..10) )
Stringification of a Buf is not done with 'Stringy'. The 'decode'
method should be used to convert a Buf to a …
dakkar no say, no print, no anything, still explodes
grondilu another case of degeneracy issue 12:05
Nemokosch it's a warning probably
12:05 xinming left
dakkar m: [~^] ( Buf.new(1..10) );say "was a warning" 12:05
camelia Potential difficulties:
Useless use of [~^] in sink context
at <tmp>:1
------> <BOL>⏏[~^] ( Buf.new(1..10) );say "was a warni
Stringification of a Buf is not done with 'Stringy'. The 'decode'
method should be used to co…
dakkar m: my $a=[~^] ( Buf.new(1..10) );say "was a warning" 12:06
camelia Stringification of a Buf is not done with 'Stringy'. The 'decode'
method should be used to convert a Buf to a Str.
in block <unit> at <tmp> line 1
dakkar nope, it's an exception and it terminates the program
12:07 xinming joined
Nemokosch okay 12:07
Buf.new is still kinda morbid imo
dakkar m: my $a= reduce &infix:<~^>, ( Buf.new(1..10) ); 12:08
camelia Stringification of a Buf is not done with 'Stringy'. The 'decode'
method should be used to convert a Buf to a Str.
in block <unit> at <tmp> line 1
dakkar ok, so it's got something to do with the operator… probably its identity value
12:08 jjido joined
dakkar because we're reducing a 1-element list with a binary operator 12:09
Nemokosch now that's a great observation
dakkar so `reduce` has to guess what to pass to the operator as its first argumnent, and it picks the operator's identity value… which I'm not sure how to access 12:10
m: dd [~^]
camelia ""
dakkar aha! there's our string 12:11
grondilu: so, solution is: don't pass a 1-element list to the reduce stop
step
(because the result would be very unclear anyway…)
Nemokosch look: github.com/rakudo/rakudo/blob/2022...gy.pm6#L77
dakkar (it's now lunchtime for me, see you in an hour or so)
also github.com/rakudo/rakudo/blob/2022...m6#L66-L69 12:12
Nemokosch anyway, it seems to me that ~^ is not even meant to be used here 12:14
UNLESS you do want to get a string 12:15
oh right, it was line 68, I miswrote it to prefix 12:19
meh, now I can see that there is no other buffer operator for binary stuff 12:22
I think this is something to fix, if not two things to fix. Whether a random Buf role should be coerceable with .Stringy is one question - but that the default shouldn't be an empty string *for all types*, that's fairly certain 12:27
Anton Antonov I want to (re-)implement a spellchecker that produces suggestions of misspelled words taking into account the surrounding words. What would be a good namespace for it? I consider "ML::Spellchecker", "ML::PhrasalSpellchecker", "ML::ContextualSpellchecker". (For a brief time I considered using "Lingua::*", but the design I have in mind "learns" from textual data.) 12:28
lizmat ML::ContextualSpellchecker feels good to me 12:35
12:39 simcop2387 left 12:40 simcop2387 joined
Anton Antonov @lizmat Thanks! "ML::ContextualSpellchecker" is my favorite, but wonder to what degree "Contextual*" is implied by "ML::". 12:40
lizmat so you're implying that ML implies learning from context... hmmmm 12:41
Anton Antonov @lizmat Yeah, that is a long implication chain... 🙂 12:42
lizmat if you feel strongly about that, then indeed, I would drop the Contextual part :-)
Anton Antonov Ok, sounds good. Of course, "ML::Spellchecker" is going to have "dumb", non-contextual spellchecker(s). 12:43
12:50 jpn left 12:52 jpn joined 12:58 jpn left
Nemokosch grondilu: apparently if you take two utf8 buffers, they will turn into Str with ~^ either way 12:59
12:59 guifa left
although I think this is a crazy feature, it definitely seems deliberate, like there is no other real reason for it to exist 13:00
13:00 guifa joined
I wonder if it's possible to beat the .Stringy part down, at least 13:00
13:18 jpn joined 13:32 jpn left 13:42 jgaz joined 13:46 jpn joined
Skarsnik Finally I can run my benchmark, I should do a blog about this xD 13:59
lizmat hehe.. please do! 14:00
14:07 Homer_Simpson joined
Homer_Simpson anyone know how to view a functions signaure in comma IDE 14:07
eg say print put prompt readline get close open etc
14:12 epony left
Skarsnik 2018.04 give me a moar panic, fun 14:14
14:22 epony joined
err, there are some perf regression since 2020 for my benchmark 14:30
lizmat could well be... I'd be interested to see compared to 2022.12 and blead, though 14:34
Skarsnik <gist.github.com/Skarsnik/2e472a97d...e556fc> 14:38
Nemokosch Gumbo gives me NativeCall flashbacks 14:41
Skarsnik If I remember my profiling attempt, the nativecall stuff is very unsignifiant 14:42
14:44 jjido left
I wonder if I still have my leak tests x) 14:44
Anton Antonov @lizmat Here is an MVP of "ML::Spellchecker" : github.com/antononcube/Raku-ML-Spellchecker 14:45
lizmat nice, but what decides to go for "arise beautiful butterfly" ? 14:46
14:48 razetime left
leont I'm trying to use $*IN.Supply in a react { whenever ..}, but it seems to hang, anyone got a clue? 15:04
15:05 rf joined
leont Like, it will only receive the data on EOF 15:06
That makes the whole thing rather useless from an asynchronous POV 15:07
dakkar leont: `.Supply(:size(10))` or something? 15:10
the docs say that it batches 64KB at a time…
leont Yeah, I had just come to that conclusion. Or probably :1size and then .lines really.
dakkar has vague memories of Supply.lines and IO::Handle.lines behaving pretty differently 15:11
lizmat stackoverflow.com/questions/750750...-an-stdout perhaps ? 15:14
leont It's really a weird way of going about though. It forces one to read byte-for-byte, which is quite inefficient. What one really wants is to read a large buffer, but if the result is small read that. It's read(2) versus fread(3) semantics really. 15:21
Anton Antonov @lizmat [nice, but what decides to go for "arise beautiful butterfly" ?] -- yeah, to answer this question (with probabilities and evidence) a more dedicate ML algorithms is needed. (Using contexts.) 15:22
lizmat $*DEFAULT-READ-ELEMS perhaps ? 15:25
15:25 Kaiepi left
leont lizmat: it's not about the number, it's about the retries. 15:25
Anton Antonov Basically, I use the tries-with-frequencies data structure to find most probably completions. For example, "cooking ice in the kitchen" should get a correction suggestion for "cooking rice in the kitchen", although none of the words in the former phrase are misspelled.
lizmat dann bin ich uberfragt
Anton cool!
Anton Antonov "dann bin ich uberfragt" - I can spellcheck that if find large enough German text. 15:27
Nemokosch leont: which one is the useful behavior here: read(2) or fread(3)? (for context, I don't know either lol) 15:29
leont read(2) IMNSHO 15:30
read(2) semantics is "read up to X bytes", fread means "read exactly X bytes unless EOF".
The latter can't be combined well with something like .lines, because it will read all those 65536 characters before even considering splitting them up. 15:31
(or .words, or anything else interesting) 15:32
15:32 Kaiepi joined
Nemokosch so what happens if there is no data to retrieve, for example? 15:33
leont It won't emit anything until there is data 15:34
I'm not exactly sure about the details, I would hope a blocking read inside a supply{} (in IO::Handle.Supply) doesn't block anything else from running but I'm not sure about that) 15:35
Anton Antonov Are there rules or policies for having / providing / distributing large resource data files in Raku modules? 15:37
Nemokosch <@694526400488669234> (Anton) this would make a great FAQ question 15:38
Anton Antonov For example, the "ML::Spellchecker" module I am working on will have some large text corpus (say ≈20MB), which is used for _training_ of the spellchecker; the trained, ready to use spellchecker has much smaller operation data. I can use two module repositories: "ML::Spellchecker" and "ML::Spellchecker::Training". (Although, I am not very happy with that separation.) 15:39
Nemokosch leont: Supplies and react blocks still seem to be something that kind of require pioneering from a user point of view 15:40
this sort of stuff could be a killer feature or at least a solid offer 15:41
grondilu is back 15:47
15:47 frithnanth joined
grondilu dakkar: ok but that doesn't tell me why it worked fine with reduce * ~^ * 15:49
dakkar grondilu: because, when presented with something that's not an operator, reduce doesn't try to guess the identity value 15:50
grondilu is that specced?
I mean maybe I'm dense, but I don't see why reduce * op * would behave any differently than [op] 15:51
dakkar docs.raku.org/routine/reduce look for "identity"
grondilu ok I get it. 15:53
dakkar `[op]` works the same as `reduce &infix<op>`
15:54 Sgeo joined
grondilu * ~^ * is an adhoc function so it can't be assigned an identity. Got it. 15:54
leont I have been bitten by supply/react edge cases without fully understanding before, I'm slightly suspicious about them. They're mainly poorly documented though. 15:55
dakkar btw, the "identity" is just "call the code with 0 arguments, see what happens"
grondilu oh, I thought it was an attribute or something. 15:56
dakkar I'm sort-of guessing by looking at the definiton of the op (I linked it earlier)
Nemokosch I kind of have an OCD for issues (also bad experience) so I didn't want to take that burden just yet. BUT. 16:03
16:03 frithnanth left
github.com/2colours/Raku-ideas/blo.../Issues.md points 19 and 20 try to point out the things collected from this discussion 16:03
16:05 melezhik joined
melezhik .tell Skarsnik: if you play with Rakudo benchmarks I can setup a simple SparrowCI pipeline to run your tests against the 2022.12 and the blead Rakudo , we have everything we need - sparrowhub.io/plugin/sparkyci-pack...o/0.000010 16:10
tellable6 melezhik, I'll pass your message to Skarsnik
melezhik ^^ lizmat
sjn .seen zostay 16:12
tellable6 sjn, I saw zostay 2020-09-21T20:22:21Z in #raku: <zostay> I have a sort of production project that was built on Smack + Cofra which is a application framework I started building, but both are only half-baked.
16:15 melezhik left
Nemokosch oof 16:19
sjn: what's up?
16:21 epony left 16:22 epony joined 16:26 Homer_Simpson left 16:27 sena_kun left 16:28 Homer_Simpson joined
Skarsnik blogs.perl.org/users/sylvain_colin...-slow.html 16:28
Homer_Simpson .
how come doing my $fh = open "test.txt", :r; my $contents = $fh.slurp; print($contents); does not print £ ¬ or ¦ properly 16:30
I get 1234567890-=[];'#,./!"£$%^&*()_+{}:@~<>?\|`¬¦
Skarsnik Can you terminal display these?
16:30 jpn left
Homer_Simpson yes 16:31
my terminal is powershell
17:08 epony left 17:10 epony joined 17:12 archenoth joined, sena_kun joined
Nemokosch I'm still somewhat doubtful if a powershell window uses UTF-8 encoding 17:27
17:28 alfonsox left
<@106510366283952128> (Skarsnik) definitely weekly material 🙂 17:29
by the way... it's fairly suspicious that 2021.10 is around the time new-disp got merged 17:33
17:35 dakkar left
Skarsnik is that power shell terminal or the new Windows 10+ terminal? 17:52
tonyo Homer_Simpson: what should it say? 17:53
skarsnik: that's great - can you bisect and figure out which commit set that performance back? 17:55
Skarsnik Not today x) 17:56
Maybe tomorrow or this week end, not sure
17:57 lichtkind_ joined 17:59 lichtkind left 18:00 reportable6 left 18:03 reportable6 joined
rf Would anyone be interested in a functional programming with Raku guide/blog? I feel like Raku suites a functional style so well! 18:07
moritz my experience as a former Perl 6 blogger: yes, lots of people would be interested 18:08
and yes, lots of great features for FP in raku
(as long as you don't want to statically type your functions, that can get ugly pretty quickly)
rf Great! It'll probably be a while but I'll figure something out 18:09
tonyo the typing is what prevents me from doing fp with raku 18:12
moritz honstly, I don't have much experience with statically typed FP; Haskell and OCaml come to mind where it's supposed to be less daunting 18:15
Nemokosch yes and the ML end of things starts to look more like a cult than reasonable design 18:23
Raku would probably be more like Erlang or some Lisp. The data types, though... we still don't know a lot about Daniel Sockwell's work. 18:25
Skarsnik Every modern language has anonymous functions/lambda/whatever. Aside from strong type contract thing functionnal languages are kinda useless x) 18:26
I wish Raku has a pragma that say "don't ignore method signature at compile time"
Nemokosch I doubt it's "ignored", considering that dispatch relies on actual types, and quite surely there is some JITting at the very least, involving dispatches 18:30
Skarsnik If you write like class A { method foo(Int $a)} ; A $a .= new; $a.foo("plop"); this fail at run time since it assume you can always add method at runtime 18:32
If you write like class A { method foo(Int $a)} ; A $a .= new; $a.foo("plop"); this fail at run time since it assume you can always add method at runtime 18:33
btw does the multiline error repport get finally fixed? or it's more something for RakuAST? 18:34
Nemokosch what is the "multiline error report"? 18:37
ironically, this will fail at compile time because this is not valid Raku code 18:38
18:45 Maylay left
Anyway, I'm not sure if the "bytecode" from the compilation even cares about those types. 18:47
Skarsnik multierror thing is if you have like a multiline instruction and something bad happens in like the last line, ti will repport error on the first line 18:53
rf Raku FP is a lot like Clojure tbh\ 18:56
You have all of the immutability you want when you need it (Map, List etc) but without implicit partial application 18:57
18:58 jpn joined
Nemokosch And that's why Daniel's "Persistent Data Structures" are so long awaited 18:59
That could really help with performance of immutable stuff
19:04 jpn left 19:06 jjido joined 19:10 jpn joined 19:16 Maylay joined
tonyo i disagree about the strong type contract being the only thing FP is good for 19:23
monads are great, it at least hasn't been bastardized to the point of OO yet. in haskell they'd be better without thunking 19:24
19:25 Maylay left 19:33 deoac joined 19:41 jpn left 19:51 deoac left
rf Monads are very good, even if they aren't statically typed. I agree 19:52
20:01 abraxxa1 left 20:04 abraxxa-home joined 20:32 melezhik joined 20:36 Homer_Simpson left
Nemokosch monads are decorators in a costume, change my mind 😛 20:36
melezhik .seen rawleyfowler 20:37
tellable6 melezhik, I haven't seen rawleyfowler around
melezhik .tell rf 20:38
tellable6 melezhik, I saw rf 2023-01-11T19:52:01Z in #raku: <rf> Monads are very good, even if they aren't statically typed. I agree
melezhik .tell rf: you probably need to add a simple sparrow.yaml to Monads to make CI works
tellable6 melezhik, I'll pass your message to rf
melezhik the same way you did for Humming Bird ... 20:39
Nemokosch so is rf rawleyfowler, at the end of the day? 20:43
melezhik I guess so, I tell from the comment s I've found fir this alias )) I might be wrong though , sorry ahead of time )) 20:48
rf Yes. That is me. Sorry melezhik I was about to add that before my work needed me!
melezhik No worries, and nothing to say sorry for ))) 20:49
I just keep my eye on sparrowci builds and try to help whenever I can ...
Nemokosch holy based 20:51
rf melezhik: Should be good I think 20:52
melezhik yup 👍 20:53
rf Thanks again for SparrowCI it works very nicely :D 20:55
20:58 melezhik left 20:59 abraxxa-home left 21:10 grondilu left
Nahita rf: i had no idea you could escape > in a user defined operator, thanks for that 21:11
i was using the nonascii version of that
21:11 melezhik joined
melezhik . 21:11
rf - success - ci.sparrowhub.io/report/2565 21:14
21:15 melezhik left
rf Nahita: No problem :D 21:22
I'm going to try to implement the Option monad next, then I want to see if I can do Lightweight Threads 21:25
21:37 jpn joined 21:40 melezhik joined
melezhik . 21:40
rf - welcome
21:41 melezhik left 21:43 jpn left
Anton Antonov Monads allow your code to have algebraic structure. (Which can be leveraged in variety of ways, e.g. readability, making re-interpreters, etc.) Decorators -- I agree very similar in function -- do not generally provide that ability. The other thing about monads is that the "process of decoration" itself is programmable -- monads are based on "programable semi-columns." 21:45
21:47 n1to joined
Nemokosch I think you pointed out the thing that makes monads controversial. It's just so. much. abstract. boilerplate. I used to like math at high school, I was even competitive but programming is about solving problems, at the end of the day, not to describe some overblown algebraic concept with code 21:49
tonyo nemokosch: monads are decorators in what way 21:52
tellable6 tonyo, I'll pass your message to Nemokosch
Nemokosch in a way that adds a costume to them :kappa: 21:53
tonyo oh, monads are so much more than that. and they're more than overblown algebraic concepts. monads can easily be used to model business processes with failure modes congruent with operations 21:54
the boilerplate around creating them is far less than littering an otherwise useful process with `if err go here`
Anton Antonov My monad is very problem-solving oriented based. You can see how I use monads in the movie I posted today in the showcase channel. 21:55
Nemokosch it's rather that the sole concept of a monad is basically something one can come up with by common sense - except the overabundance of crippling category theory jargon, and some axioms taken from that 21:56
inb4 "there is a difference between doing something like that and doing it exactly right"
Anton Antonov * ... my monad usage is problem solving based ...
tonyo they're really not 21:57
Nemokosch hah, they all claim that 😆 22:01
tonyo have you ever used monads practically? 22:02
Nemokosch perhaps it would be good to have a command for tellable that goes like - "if my first message is this command, drop the messages stacking up immediately"
because it will be really hell on earth if I log in once 22:03
the question is: has anybody used monads practically, WHILE calling them monads? 22:04
22:10 xinming left, xinming joined 22:11 discord-raku-bot left, discord-raku-bot joined
tonyo yes 22:15
monads are not some glorified class
Nemokosch well, depends on the implementation of the class 😛 22:17
rf I do agree most developers use monads without knowing they're monads. But I have worked with people who know they are monads. 22:18
tonyo personally i've used them in projects where business processes were important so it could handle failure modes for things like `Order.get(id=5).update_status(status='shipped', tracking_no='123').email_customer()` .. without monads there's a lot of `if order == nil { .. handle this here }` and this introduces a lot of branching. `update_status` on a monad where the side effect is the order doesn't 22:20
exist handles this case without branching and email_customer still can do the right thing, which is not email a customer that their order id doesn't exist
i hated monads and also thought they were a dumb way of thinking about data relationships but after using them, they're far easier to grok and use practically where clarity in process is important 22:21
22:21 jgaz left
tonyo i'd like to see a class that can handle that without littering the code logical branches to handle failure modes 22:22
rf I work at a bank, we use monads for the critical stuff that I'm not allowed to touch :D 22:23
tonyo stephen diehl (another bank person), i think, has a good talk on all that stuff 22:24
where he's highly critical of monads and more so of thunking
Nemokosch I think monads as in software architecture bricks may very well be useful. It's rather the discourse that it involves, this carried weight, that fires back. 22:27
tonyo agreed, how difficult it is to actually understand/teach what they are is a huge barrier 22:28
22:29 discord-raku-bot left 22:30 discord-raku-bot joined
Nemokosch My own experience is that I could never understand even banally simple things when I started programming. Like, what an interface even is, or what polymorphism is about. And then I faced problems that these means could help with. Same with decorators: couldn't for the life of me understand them, until one day it came naturally as a solution to various checks performed for command execution. 22:30
rf That's sort of the running joke. Once you learn Monads you can't teach Monads
Nemokosch Yes, and it's fairly logical that the joke has some actual truth to it. One cannot use a half-monad or quasi-monad. It's either the thing or not. And to make sure it is the thing, you need to enter a discourse that requires a much more axiomatic approach that you'd be used to with software architecture most of the time. 22:31
rf Monads are also really hard to teach without a strong type-system, which most languages don't have 22:32
22:32 jpn joined
Nemokosch that's why I can only say: decorators, at least in the Pythonic sense, perhaps I get; and I have a strong impression that a monad is something like that, with some additional formal criteria I guess? 22:35
rf A monad is simply a box that may or may not produce a side-effect upon looking inside of it. 22:36
An empty List is the side effect of viewing a List (lists are Monads)
22:37 jpn left
Nemokosch well, what if a side-effect does happen? 22:38
rf Nothing typically, what happens right now if you try to map an empty list?
Nemokosch well, Raku aside, I would say I get an empty list back 22:40
rf Yes, the side-effect is hidden from you and you continue on your way safely
Nemokosch you know, it trips me up that you said that a value is a side effect 😅 how does that add up?
rf A monad has to have some kind of side-effect, could be async-io, could be an exception, could be an empty list 22:41
Promise is a monad, it's side-effect is that it might fail
Or not resolve (whatever the term for promises is)
Nemokosch that would have been the next question (can a valid Promise in this context never return) 22:42
rf That is where bind and map come in (>>= and >>=?) 22:43
They're just functions that hide the unwrapping of the object from you, you can review my code for an example: github.com/rawleyfowler/Monad-Result
Nemokosch well the map operation is indeed a lot like Promise#then in JS 22:50
rf .then is actually just .map 22:51
It's just named differently because JavaScript
22:53 discord-raku-bot left, discord-raku-bot joined
Nemokosch well, thank heavens they didn't mean it map, that's something else already 22:53
and you are lucky if it's only one other thing because associative arrays are also often called maps 22:54
tonyo the monad itself both describes the side effect and hints the compiler on how to tell you that you've forgotten to handle some case 22:56
Nemokosch and this bind really seems like "map but things might go wrong", in this Result context at least 22:58
tonyo the latter is the biggest rub, i think, with things that claim strong typing but really have types without strong typing (like go)
well, it's kind of more than that too. i mean, if you think about something as simple as Maybe(List) and think about calling map on Maybe (which is a container for a `List|None`) then the compiler should force you to understand your code needs to handle both cases, remember you're calling it on a Container and not the list itself. In both cases you may prefer to have it return the list or an empty 23:03
list so your monad handles the side effect of `you gave me a nil value`, or you may prefer it just continues to give you `Maybe(List)` which will still unwrap to `None`. This gives you the ability to go a step further and start linking method calls on the Container with the safety of knowing you're not going to use the unwrapped `List|None` in a way that blows. practically it's a little more
complicated (usually)
Nemokosch I'm not so fond of static typing but maybe if they were actually robust, like supposedly in certain functional languages, they could make a good appeal 23:04
tonyo i found it a very crappy experience at first but thought better of it after using it for a while 23:07
otherwise agree though, go's typing kind of stinks 23:08
python's leaves a lot desired 23:09
23:10 n1to left
rf OCaml is probably the best truly statically typed language I have written 23:11
tonyo looks pretty interesting 23:14
esp with that beautiful vim logo on their main page
Nemokosch :DDD 23:16
23:26 jpn joined 23:30 jpn left
tonyo rf: i wonder if that Monad-Result could be generic-ized a bit to make Monads definable in just the course of writing raku 23:33
23:35 vrurg_ is now known as vrurg 23:37 sena_kun left 23:49 hythm left
Anton Antonov I have a proof -- please watch this movie : www.youtube.com/watch?v=Xy7eV8wRLbE . (I referred to it above.) 23:53
guifa o/
tonyo that guy is using your name 23:56

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!