🦋 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 00:06 simcop2387 left, perlbot left 00:07 simcop2387 joined 00:08 perlbot joined
rf So turns out you need === defined for CArray[my_cstruct_type] to work 00:28
=== (Str $foo, my_cstruct_type $bar)
00:33 derpydoo joined 01:03 xinming left 01:06 xinming joined 01:11 clsn_ joined 02:11 evalable6 left, tellable6 left, bloatable6 left, statisfiable6 left, benchable6 left, squashable6 left, sourceable6 left, shareable6 left, bisectable6 left, releasable6 left, unicodable6 left, committable6 left, quotable6 left, nativecallable6 left, reportable6 left, linkable6 left, notable6 left, greppable6 left, coverable6 left, quotable6 joined, nativecallable6 joined, sourceable6 joined, linkable6 joined, notable6 joined 02:12 tellable6 joined, shareable6 joined, committable6 joined, reportable6 joined 02:13 greppable6 joined, squashable6 joined, bisectable6 joined, statisfiable6 joined, coverable6 joined, bloatable6 joined 02:14 benchable6 joined, evalable6 joined, releasable6 joined, unicodable6 joined 02:25 MasterDuke left 02:31 rf left 02:38 codesections left 03:31 Xliff left 03:46 swaggboi left 04:05 swaggboi joined 05:05 coverable6 left, releasable6 left, benchable6 left, evalable6 left, quotable6 left, bisectable6 left, shareable6 left, sourceable6 left, greppable6 left, unicodable6 left, linkable6 left, reportable6 left, notable6 left, committable6 left, tellable6 left, bloatable6 left, squashable6 left, statisfiable6 left, nativecallable6 left, sourceable6 joined, nativecallable6 joined 05:06 benchable6 joined, quotable6 joined, coverable6 joined, notable6 joined 05:07 releasable6 joined, bisectable6 joined, greppable6 joined, committable6 joined, unicodable6 joined, squashable6 joined, evalable6 joined, statisfiable6 joined, shareable6 joined 05:08 linkable6 joined, reportable6 joined, bloatable6 joined, tellable6 joined 05:11 wbvalid joined 05:18 wbvalid left 05:23 jpn joined 05:28 jpn left 06:00 reportable6 left 06:02 reportable6 joined 06:49 teatime joined 06:52 teatwo left 07:16 Sgeo left 07:29 Max51 joined 07:30 Max51 left 07:45 jpn joined 08:04 jpn left 08:06 jpn joined 08:24 jpn left 08:35 abraxxa joined 08:37 simcop2387 left 08:38 simcop2387 joined, perlbot left 08:40 perlbot joined 09:04 discord-raku-bot left 09:05 discord-raku-bot joined 09:41 ab5tract joined 09:45 jpn joined
tbrowder__ g'day, all. does anyone have a working workflows/windows.yml for modules on github? 11:02
11:13 linkable6 left, evalable6 left 11:14 linkable6 joined, evalable6 joined
Nemokosch wouldn't bet my life on that, good sir. But hope dies last 11:16
tbrowder__ 👍🏻 11:48
11:50 petro-cuniculo joined 11:53 gcd left 11:57 petro-cuniculo left 12:00 reportable6 left 12:03 reportable6 joined, abraxxa left 13:03 linkable6 left, evalable6 left 13:06 evalable6 joined, linkable6 joined 13:43 rf joined
rf Morning folks 13:43
13:44 jgaz joined 13:46 jpn left 13:49 jpn joined
Anton Antonov @rf Morning, you, Haskel apologist ! 13:50
And monad-promoter… 13:52
Voldenet Promises are monads and they're everywhere 13:53
monad-ish 13:54
rf Anton :P 13:56
13:57 jpn left 14:00 jpn joined 14:05 jpn left
Voldenet say (await Promise.kept(Promise.kept(42))).WHAT 14:10
evalable6 (Promise)
Voldenet this looks more monadish than js impl, that would just return 42 in that case
14:10 jpn joined
[Coke] would appreciate if someone could review the "is it a bug" question in github.com/Raku/doc/issues/4271 14:18
lizmat my question would be: did it recently change, or has it always been this way? 14:20
14:20 abraxxa-home joined
Nemokosch why would if ever topicalize? 🤔 14:22
lizmat yeah, it feels like an implementation detail
Nemokosch > The with statement is like if, but tests for definedness rather than truth, and it topicalizes on the condition, much like given: 14:26
so sounds like the documentation contradicts itself 14:27
> You may intermix if-based and with-based clauses. this is the interesting part... 14:28
m: if 0 { .say } orwith Nil { .say } else { .say }
Raku eval Nil
Nil 14:29
Nemokosch perhaps this is what it's trying to say
m: if 0 { .say } orwith Nil { .say } elsif 12 { .say } else { .say } 14:30
Raku eval (Any)
Nemokosch this seems surprising to me, though
the former else clause ran, as an elsif clause, and this time it un-topicalized
14:32 jpn left 14:34 jpn joined
rf Voldenet: Monads are a container with a map and bind 14:36
(and return) but that isn't super important
14:39 jpn left
rf Not sure if promise fits it perfectly 14:39
dutchie do you not need return to do the bind/join equivalence 14:40
Woodi rf: but monad-ish can mean "clousure" too ;)
14:41 simcop2387 left 14:42 perlbot left, perlbot_ joined, simcop2387 joined
rf Woodi: Not sure what you mean by that 14:42
Woodi rf: just trying to abuse meanings becouse of some similiarities :) 14:43
not even sure what "bind" is, too lispy :)
14:43 perlbot_ is now known as perlbot
rf bind : M a -> (a -> M b) -> M b 14:43
Woodi so M is domain of values ? 14:44
and result is which part ?
but assumed functions... 14:45
rf M is a monad, a is the type held within the monad
Nemokosch let's keep it simple 14:48
which operation returns a monad, and which a value?
rf Bind always returns a monad 14:49
Woodi whay it is doubled ? a -> M b -> M b ? 14:50
rf (a -> M b) is another function 14:51
Woodi then what -> means ? 14:52
rf en.wikipedia.org/wiki/Partial_application 14:53
exp lol a wikipedia page on computer science is not going to make things any more understandable 14:54
they unironically care only about number of facts expressed, not how many people understand what's written
Woodi so bind is function that returns monad that changes values into ... ? 14:55
I thinked about bind in Lisp like some kind of pointer... 14:56
Nemokosch oh
14:56 jpn joined
so bind is the one that takes a function that constructs the new monad directly 14:56
dutchie if we stick to just talking about promises, bind corresponds to the thenmethod
then method
Nemokosch the then method, when you directly return a Promise in the callback 14:57
dutchie yeah exactly
the callback is the a -> M b
Woodi partial application describes curring ? 14:58
dutchie the invocant is the M a which gets "unwrapped" and fed into the callback
14:58 perlbot left, simcop2387 left
Nemokosch Promise.resolve(42).then(x => { const funky = Math.random()*x; return Promise.resolve(funky); }) 14:58
in pseudocode that absolutely isn't Javascript ^^ 14:59
dutchie Woodi: they are closely related yes. a "curried" function takes multiple args by returning another function with those args "partially applied"
some people are more precise than others in keeping the two terms distinct
Woodi dutchie: my math teacher said: understand and then memorize or memorize and then understand :) 15:00
Nemokosch yeah I guess think of Haskell 15:01
from what I know, Haskell only has functions that take one argument 15:02
Woodi so looks curring use partial application or even is p.a. ...
Nemokosch: only one ? crazy :) 15:03
tellable6 Woodi, I'll pass your message to Nemokosch
Nemokosch I don't know Haskell syntax but going by this logic, a function that "takes several parameters", would be called like f(1)('asd')(True)
15:04 perlbot joined
where f would return a new function that would return a new function that would return.... you get the idea 15:04
Woodi sounds in order ;)
15:05 simcop2387 joined
Nemokosch and on each call, the current parameter is built into the returned function 15:05
at which point it's just a matter of approach if you say "it has n unbound variables" or you say it's an nth order function 15:06
15:10 Sgeo joined 15:11 tbrowder_ joined
rf Nemo bind will "unwrap" the first monad and feed the unwrapped value to a new function (the second parameter) which returns a new monad 15:12
github.com/rawleyfowler/Monad-Resu...lt.rakumod
^ That repo implements a monad if you;re interested Woodi 15:14
Also Nemo you are correct a function call in Haskell is like f(foo)(bar)(baz) 15:15
Anton Antonov @Voldenet "Promises are monads and they're everywhere" -- you are on record, I will verify the monad axioms on promises (and be vocal if you wrong.) 15:21
@Voldenet "monad-ish" -- nice escape (from rigorous feedback.) 15:22
rf Hahahaha 15:23
15:25 grondilu joined
Anton Antonov @rf I considered working on a post that criticizes your monad approach. Diced to postpone it indefinitely. 15:25
rf I am not opposed to counter ideas, though, I haven't heard a compelling one against monads yet. 15:26
Voldenet I once said about that about js since "ye it's mostly monads" but then it wasn't using composition properly
because then(a).then(b) is different depending on whether return value is Promise or not 15:27
rf then is map
Voldenet from then(x=>a(b(x)))
Woodi rf: checking
Anton Antonov @rf My point of view on monads is how much a monadic system (e.g. a Raku package) makes the code written with it to have algebraic properties.
Voldenet hence my test above
say (await Promise.kept(Promise.kept(42))).WHAT
evalable6 (Promise)
Voldenet it's at least not as bad as js 15:28
rf Anton: I am more interested in abstracting side-effects than algebraic properties
It benefits the consumers of code to use Monadds as well so you can describe the intent of the code 15:29
Voldenet I bet you can sort of do algebraic effects in raku if you like pain 15:30
Anton Antonov @rf Sure. But, I leverage the algebraic properties when I make translations from natural language DSLs into programming language DSLs. (And vice-versa.) Hence, the algebraic properties for me are important.
Voldenet and .throw/.resume combo
Anton Antonov @Voldenet Dully noted.
Nemokosch then is kind of both bind and map, from what I understand 15:31
rf I really dislike exceptions, which is why I made Monad::Result, I think its very gross to make the caller decipher what possible exceptions can be thrown
then is just map, map : M a -> (a -> b) -> M b
Nemokosch well, then join me on the dark side and let's dislike control exceptions together 😛 15:32
rf CATCH { default } on every block is just as bad IMO
Plus it's not enforced or implied so uncaught exceptions are far too common
Nemokosch false negatives are worse than false positives with this really 15:33
when you only see that some of your assumptions didn't hold
Anton Antonov @rf You and @Nemokosch but be on the same gray side. (Or same far side gallery.)
Woodi rf: "exceptions" looks like "sudden explosions" :) but concept of shortcuts in execution flow should be usefull... if we have good behaving code like calculations... 15:34
Anton Antonov @rf @Voldenet Here is a (very schematic) flowchart of my monads-for-DSLs workflow: raw.githubusercontent.com/antononc...agents.jpg
rf That is an interesting approach 15:36
Woodi: Most software I write needs to be triple redundant and have 0 exceptions, thus why I prefer monads over exceptions. Shortcuts can simply be expressed as function composition assuming the types align 15:37
Which is also one of the main concepts behind Humming-Bird ^
Voldenet m: class Effect is Exception { has $.x is rw; }; CATCH { when Effect { .x = 42; .resume; } }; my $x = Effect.new; $x.throw; say $x.x 15:38
camelia 42
Voldenet I'm begging you, don't use the above thing
it sort of works though
rf Anton you are the first ML person I have heard even use the word Monad :D 15:39
Anton Antonov @rf Basically, if I can use "your" monads if I can put the operations in a reduce statement. For example, reduces(&my-monad-bind, my-monad-unit-object(), [&some-op1, &some-op2, &some-op3, &take-value] ) .
rf Yes that should work 15:40
Anton Antonov @rf Ok, good. (Meaning, "you are on record and will try to verify.") 15:41
@rf ML people use monads, but they do not know and/or use the terminology. 15:42
rf As long as your ops are -> M a -> (a -> M b) -> M b
Voldenet most people use monads and algebraic effects
Anton Antonov @rf Right, the associativity rule.
Voldenet they just buried in N layers of their language abstraction
s/they/they're/ 15:43
Anton Antonov @Voldenet Most Data Science people do not want to program. So, whatever simplifications are used to make the required work more palatable. 15:44
Voldenet I remember showing my data sci. code to data scientist, he scratched his head and said he didn't get the code :/ 15:46
(I tried to abstract away data science part so I could get to my programming one…)
Anton Antonov @Voldenet Right, hence, I make/use natural language DSLs for Data Science. They still say the same. 15:47
Voldenet that makes sense
Anton Antonov Hopefully, I am not overestimating the interest in this -- here is an example of data wrangling Python code generation from sequences of natural language commands: github.com/antononcube/RakuForPred...thon.ipynb 15:51
Or, if you prefer, the Raku code results version: github.com/antononcube/RakuForPred...Raku.ipynb 15:52
15:57 perlbot left 15:58 simcop2387 left 16:01 perlbot joined 16:02 simcop2387 joined
clsn_ So. Haven't worked with raku in a *long* time, and some things have changed. Right now, I can't see how it's possible to make a regex that matches a *combining* character (or set thereof). I can only match base characters and specify combining characters on them if I want, but I'm searching for the actual combining character which may be on any of many bases (and may even have other combining chars with it.) 16:54
This is not an unrealistic request, by the way. Not everything is like é where the accent isn't something you'd want to search for without the letter. I'm working with Hebrew cantillation marks, which are like punctuation that happen to be written as combining characters. 16:56
m: my $x="עֵֽינֵיכֶ֑ם"; say $x ~~ /\x[0591]/; 16:59
camelia Nil
17:02 linkable6 left, evalable6 left 17:03 linkable6 joined 17:04 evalable6 joined
Nemokosch strings are normalized according to NFC 17:09
clsn_ Yes, which is fair enough... But rakudo, from what I've seen, matches stuff according to its "NFG". How might I write a regex that can match the 0591 in that string? NFC vs NFD isn't really relevant; none of the characters there are or can be precomposed. 17:11
Nemokosch I'd expect a regex to operate on the level of characters, not codepoints 17:14
clsn_ Well, from a Unicode perspective, \x[0591] is a character, so I'm not sure what you mean. If you mean by graphemes, that sort of presumes that it doesn't make sense to search for an \x[0591] because it is written as a diacritic, yet that makes just as little sense as saying that it doesn't make sense to search for a comma in a sentence. 17:15
Nemokosch I'm not sure if it's still a character after NFC 17:17
but sure thing, definitely not a grapheme, and a high-level string has characters as graphemes
17:18 codesections joined
clsn_ NFC, as I understand it, is "combine everything that can be combined into precomposed characters," and nothing in the example string can make up a precomposed character. Am I misunderstanding you? 17:18
Well, then, how would I write a grammar to search for it? It may not be written as a spacing character, but it is exactly as reasonable to search for it as it is to search for a comma or semicolon in English text. 17:19
Nemokosch This is probably beyond me. There is stuff like this docs.raku.org/type/Uni.html 17:23
but whether it works with regex stuff, no clue 17:24
clsn_ So I could convert it to something more unicode-ish, but can I then use regex--- I see.
This is actually for pretty much the ONLY program I've ever written in rakudo, apart from contributions I made to the actual project. And it *used* to work. Many years ago. 17:25
17:26 cfa joined
cfa bisectable6: my $x="עֵֽינֵיכֶ֑ם"; say $x ~~ /\x[0591]/; 17:26
bisectable6 cfa, Will bisect the whole range automagically because no endpoints were provided, hang tight
cfa, ¦6c (67 commits): «Nil␤»
cfa, Nothing to bisect!
Nemokosch that must have been a lot of years ago for sure 17:27
probably prior to MoarVM, and MoarVM has been the state of art runtime since like 2013
clsn_ It was a REALLY long time ago; I'm not sure I can find quite how long it was. Eh, I probably have logs someplace...
Nemokosch anyway, now I'm not convinced that it is intended to work 17:28
clsn_ Yeah, the latest commit in my repo is from December 2011.
Nemokosch there is a candidate for smartmatching Uni against Regex github.com/rakudo/rakudo/blob/2022...ex.pm6#L47 but it basically converts to Str and calls it a day 17:29
clsn_ It may or may not be "right" for it to work *as stated*, but I think there definitely needs to be some way to make it work, or you're really missing something important.
i.e. converting to some form or another that regex-matches on codepoints or something like that. 17:30
Nemokosch what I doubt, though, is that this is high-level enough to fall into regex territory
clsn_ s/i\.e\./e.g./ (can't believe I used \ for that...)
Nemokosch yeah that sounds horrible tbh, to replace a part of a grapheme 17:31
clsn_ Well, I still contend that if regexes can't do it in any fashion, then you're failing to capture or make available something very important and not unreasonable for people to want to do. I present my own program as evidence of that (granted, one might argue that I only barely qualify as "people"...) 17:32
I could easily see someone studying Hebrew or Arabic doing searches for vowel-patterns (which indicate grammatical forms).
Nemokosch well I'm just saying that it perhaps doesn't fall into regex territory
clsn_ and the Hebrew Bible cantillations are part of a well-understood and well-defined grammar. 17:33
Not certain what that really means, or if that answers. You can do that, you just can't use regexes for it? And yet it's matching patterns in a string of characters, isn't that what regexes are supposed to do for a living? Why should someone have to write up their own homegrown regex-matcher just for certain kinds of characters? 17:34
Nemokosch they are not "characters" on Str level 17:36
clsn_ My program from way back when would parse a Biblical sentence according to the structure of sentential breaks encoded by the cantillations and output a tree graph in dot format. That's parsing text with a grammar.
Nemokosch And like, regex is not meant for any pattern matching. For example, you can't just arbitrarily match binary patterns in the unicode representation 17:37
I mean, sorry for your loss
cfa here's another example, 17:38
m: say "u\x[0308]" ~~ /\x[0308]/
camelia Nil
clsn_ web.meson.org/cache/Esth:8:9.png
Nemokosch But I'm not convinced that this is a problem with the regex itself, as it clearly works on the principle that a character is a grapheme
clsn_ I can see that this is a limitation of the way rakudo has chosen to define strings and regexes. But I wonder if that choice is defensible in the face of, well, not being able to do exactly what regexes and grammars are supposed to do. 17:39
Nemokosch frankly I don't know about Unicode enough to understand what makes a "combining character" a "character", in this jargon
Again, I don't think regexes (let alone grammars) are supposed to dig this deep 17:40
clsn_ Eh, that's because "character" sounds like it should be some graphical unit, i.e. a grapheme, so it's hard to see a combining character as one.
Voldenet if you don't mind performance hit then
Nemokosch So what is it exactly, that it isn't just called a codepoint?
clsn_ But whyever not? As I said, it's a very reasonable thing to ask a grammar to do.
Voldenet m: my $x="עֵֽינֵיכֶ֑ם"; say 0x591 (elem) $x.ords; 17:41
camelia True
Nemokosch You said so yes but it didn't sound any different from saying that grammars are for binary inspection.
clsn_ In Unicode parlance, character and codepoint can be almost interchangeable. Indeed, I understand what you mean about having trouble seeing it as a character, but coming from a more Unicode-centric POV myself, I find the opposite to be true.
Nemokosch Also, you earlier made the distinction from é. ("Not everything is like é where the accent isn't something you'd want to search for without the letter.") 17:42
what backs this distinction up, that could be somehow integrated? 17:43
clsn_ I don't know. Binary patterns are not regex-fodder because they don't generally have structural meaning that's useful for pattern-matching in most strings. Combining characters do. I guess there's some fuzziness in that argument.
Ah, that's a better question...
Nemokosch Yes, this whole fuzziness
clsn_ OK, let's see if I can explain what I mean by that, and maybe I'm wrong about the distinction as well...
Nemokosch that even though "combining characters" fall back into being codepoints and hence just binary data specified by Unicode, they can matter on textual level sometimes apparently 17:44
clsn_ An é is, in a sense, a letter in itself. That's (kinda) why it has a precomposed codepoint, or at least why it was thought at some point to be worth encoding precomposed and Unicode inherited it. And even if considered as an e plus an acute accent, there's nothing in common between e+acute and a+acute. They're independent of one another. 17:45
Nemokosch > é oof 17:46
clsn_ It's not like it's completely impossible, but it would be an odd situation wherein you'd want to search for words with 3 or more accents or something.
Do my unicode chars not come through okay?
Nemokosch not really. I mean, this is just universally sad. Here we are in 2023 and the best we could get is like, semi-cover fairly similar languages in IT 17:47
Voldenet The problem is that one grapheme can be respresented by multiple codepoints 17:48
Nemokosch anyway. What I think should (and might?) exist is still something like "capture this letter containing codepoint XYZ"
clsn_ OTOH, Hebrew and Arabic vowels, for example, or even Devanagari combining vowel marks, are more related to themselves and each other than to the letters they are on. á and é have nothing in common, particularly, but का and गा rhyme, both might represent similar grammatical constructions, etc. 17:49
17:49 cfa left
clsn_ Ideally not "containing codepoint XYZ" but "containing a regexp(?) of these codepoints" or at the very least "containing a codepoint out of this set". 17:49
Nemokosch ngl this also sounds to me that Unicode itself is either misunderstood or contains problematic concepts 17:51
teatime it is complex for sure
clsn_ It's even more so in Hebrew and Arabic. A word that is CONSONANT + QAMATS(05B3) + CONSONANT + PATAH(05B7) + CONSONANT is very distinctly third-person singular masculine past tense, simple construction.
Nemokosch like, if this \x[0591] is so useful on its own and an acute accent isn't, why aren't they distinguished on any conceptual level? 17:52
clsn_ I don't need to know what the consonants are, but that's what that word means (there are exceptions and phonological concerns and blahblahblah but to first approximation.)
0591 represents the chief sentential pause in the middle of a Biblical verse.
web.meson.org/cache/Esth:3:12:.svg is an even more extreme example (the longest verse in the Hebrew Bible) 17:55
The cantillations define and determine that tree. Just as one might parse an English sentence on periods and commas and semicolons (but the cantillations are more precisely-defined and fine-grained.)
From a Unicode perspective, I guess combining characters are combining characters (they do have combining classes, though), and they don't try to distinguish ones which are more or less important than others, probably because they're not suppressing the ones of lesser importance. But here, NFG *does* "suppress" them, in some sense, in that you can't conceive of them without their bearers, and that sucks in the ones that have independent meaning as well. 17:59
17:59 abraxxa-home left 18:00 reportable6 left 18:01 reportable6 joined
clsn_ For that matter, I don't think you can even search for "some hebrew letter followed by a TSERE" or whatever (i.e. use a character class for the base.) 18:03
Nemokosch my point is that if they are so important, perhaps they should stand on their own, just like nobody would pretend that a comma or a dot is a combining character
or any punctuation for that matter
lizmat clsn_: :ignoremark ?
Voldenet probably ignoremark won't work 18:04
lizmat docs.raku.org/language/regexes.html#Ignoremark
why wouldn't it ?
clsn_ I tried ignoremark.
lizmat example?
clsn_ That ignores the mark. But I don't want to ignore the mark! I want to search for a specific mark!!
Voldenet m: my $x="עֵֽינֵיכֶ֑ם"; say $x ~~ / .<?{ 1497 (elem) $/.ords }> / 18:05
camelia ｢י｣
lizmat well, then search for the char with :ignoremark, and then check whether it is followed by a TSERE ?
Voldenet there's more than one way to do what you want
<?{ }> is not very elegant solution, but a solution
Nemokosch a not very elegant solution to a not very elegant task 😅 18:06
Voldenet in fact
m: my $x="עֵֽינֵיכֶ֑ם"; say $x ~~ / .<?{ 1497 == $/.ord }> /
camelia ｢י｣
clsn_ Maybe they should stand on their own. But Unicode considers combiningness from the point of view of graphics, not semantic sense. By adopting that, rakudo has placed ALL the combining characters in the same bucket. If there's a distinction that should be made, it will need to be made in rakudo.
Nemokosch > But Unicode considers combiningness from the point of view of graphics, not semantic sense. Holdya holdya. So far, all you said was how you have the Unicode perspective. 18:07
Voldenet current combining characters situation is probably a tradeoff, since combining characters turn elegant constant-time algos into monsters
clsn_ I can certainly search codepoint-by-codepoint and find the characters I'm looking for. But then, once more, didn't God create regexes precisely to do this kind of job? I'm looking for the word that contains a \x[0591] in a string of words. How can I do that?
Voldenet but the above one _is_ the regex 18:08
… :)
Nemokosch the only problem with it is that it's slow-ish, really
Voldenet you can compose it and put more regexes in it
clsn_ That's how I understand what I think Unicode is doing; maybe I'm wrong about that.
I'm sorry, I'm not seeing how that's working. Expecially since the thing you're matching is a letter without any diacritics. 18:09
Nemokosch m: my $x="עֵֽינֵיכֶ֑ם"; say $x ~~ / .<?{ 1425== $/.ord }> / 18:10
Raku eval Nil
Nemokosch meh, why ord
clsn_ Here... here's the whole verse. Please tell me a regex I can use to find the word with the 0591 under it: "כִּ֚י יֹדֵ֣עַ אֱלֹהִ֔ים כִּ֗י בְּיוֹם֙ אֲכָלְכֶ֣ם מִמֶּ֔נּוּ וְנִפְקְח֖וּ עֵֽינֵיכֶ֑ם וִהְיִיתֶם֙ כֵּֽאלֹהִ֔ים יֹדְעֵ֖י טֹ֥וב וָרָֽע׃"
Nemokosch m: my $x="עֵֽינֵיכֶ֑ם"; say $x ~~ / .<?{ 1425 (elem) $/.ords }> /
Raku eval ｢כֶ֑｣
Nemokosch this was the better one
clsn_ That's the right letter, yes. Maybe one can do this after all? Placing other dummy letters around it? 18:11
Nemokosch this literally does "take the letter and check what it's made of"
clsn_ (It's Genesis 3:5, btw; I just picked it arbitrarily when trying this out.)
hm. so then could I say... 18:12
Nemokosch in either case, thank you for the journey at least
Voldenet m: my $x="כִּ֚י יֹדֵ֣עַ אֱלֹהִ֔ים כִּ֗י בְּיוֹם֙ אֲכָלְכֶ֣ם מִמֶּ֔נּוּ וְנִפְקְח֖וּ עֵֽינֵיכֶ֑ם וִהְיִיתֶם֙ כֵּֽאלֹהִ֔ים יֹדְעֵ֖י טֹ֥וב וָרָֽע׃"; say $x ~~ / (\w<?{ 1497 (elem) $/.ords }>) /
camelia ｢י｣
0 => ｢י｣
Voldenet perhaps this, but my terminal outputs it all as spaces
Nemokosch I wouldn't have thought for the life of me that something that has zero length can be this significant
Voldenet that… doesn't help
Nemokosch funky, it turned backwards 18:13
clsn_ m: my $x="כִּ֚י יֹדֵ֣עַ אֱלֹהִ֔ים כִּ֗י בְּיוֹם֙ אֲכָלְכֶ֣ם מִמֶּ֔נּוּ וְנִפְקְח֖וּ עֵֽינֵיכֶ֑ם וִהְיִיתֶם֙ כֵּֽאלֹהִ֔ים יֹדְעֵ֖י טֹ֥וב וָרָֽע׃"; say $x ~~/<:Lo>*.<?{ 1497 (elem) $/.ords}<:Lo>*/;' 18:14
camelia ===SORRY!=== Error while compiling <tmp>
Unable to parse expression in metachar:sym<assert>; couldn't find final '>' (corresponding starter was at line 1)
at <tmp>:1
------> ay $x ~~/<:Lo>*.<?{ 1497 (elem) $/.ords}⏏<:Lo>*/;'
…
clsn_ bah, sorry, my rakudo regex-fu is very weak, it's been a looong time.
The "turning backwards" is probably an artifact of the Bidi algorithm at work in your terminal, which is the cause of much headache and profanity. 18:15
Voldenet m: my $x="כִּ֚י יֹדֵ֣עַ אֱלֹהִ֔ים כִּ֗י בְּיוֹם֙ אֲכָלְכֶ֣ם מִמֶּ֔נּוּ וְנִפְקְח֖וּ עֵֽינֵיכֶ֑ם וִהְיִיתֶם֙ כֵּֽאלֹהִ֔ים יֹדְעֵ֖י טֹ֥וב וָרָֽע׃"; say $x ~~/<:Lo>*.<?{ 1497 (elem) $/.ords}><:Lo>*/;'
camelia ===SORRY!=== Error while compiling <tmp>
Unable to parse expression in single quotes; couldn't find final "'" (corresponding starter was at line 1)
at <tmp>:1
------> :Lo>*.<?{ 1497 (elem) $/.ords}><:Lo>*/;'⏏<EOL>
expecting …
Voldenet m: my $x="כִּ֚י יֹדֵ֣עַ אֱלֹהִ֔ים כִּ֗י בְּיוֹם֙ אֲכָלְכֶ֣ם מִמֶּ֔נּוּ וְנִפְקְח֖וּ עֵֽינֵיכֶ֑ם וִהְיִיתֶם֙ כֵּֽאלֹהִ֔ים יֹדְעֵ֖י טֹ֥וב וָרָֽע׃"; say $x ~~ /<:Lo>*.<?{ 1497 (elem) $/.ords }><:Lo>*/; 18:16
camelia ｢כִּ֚י יֹדֵ֣עַ｣
Voldenet apparently it works
clsn_ Not really, it's the wrong work.
word.
Still, it's catching a whole word... um, a whole PAIR of words... which is... is it better than just a letter? 18:17
Wait, 0591 is 1425, not 1427
Voldenet right :D 18:18
clsn_ 1427 is HEBREW ACCENT SHALSHELET, 0593, which is a VERY rare cantillation and certainly not found in this verse.
You can write 0x0591, right? With hex notation? That'll be less confusing.
Voldenet m: my $x="כִּ֚י יֹדֵ֣עַ אֱלֹהִ֔ים כִּ֗י בְּיוֹם֙ אֲכָלְכֶ֣ם מִמֶּ֔נּוּ וְנִפְקְח֖וּ עֵֽינֵיכֶ֑ם וִהְיִיתֶם֙ כֵּֽאלֹהִ֔ים יֹדְעֵ֖י טֹ֥וב וָרָֽע׃"; my regex etnahta { .<?{ 1425 (elem) $/.ords }> }; say $x ~~ /<:Lo>*<etnahta><:Lo>*/; 18:19
camelia ｢עֵֽינֵיכֶ֑ם｣
etnahta => ｢כֶ֑｣
Voldenet you could simply do this
it's probably more sane when you want to compose it
you can use 0x591 if you want, the `{ 1425 (elem) $/.ords }` is regular subroutine 18:20
clsn_ Ugh, hard to read because of the Bidi stuff. But still. That... looks right, actually. 18:22
Still smacks slightly of not-ideal, but requiring you to use a subroutine just to pick out the combining character you want isn't THAT unreasonable. (though actually, I need to be able to check for any member of a *set* of combining characters, but that's probably generalizable from this.) 18:24
Voldenet I didn't test this for performance, maybe some form of checking substrings of .encode would've been faster 18:25
clsn_ What you have here is maybe clumsier than it once was, but still has some elegance, thank you.
Meh, I'm not terribly fussed about performance. Computers are fast enough that even slow for them is still fast, when dealing on the scale and number of instances I'm worried about. 18:26
I'll have to see if/how I can shoehorn this in to my old program, but it looks like a promising path. 18:27
18:28 grondilu left
clsn_ Anyway, so thanks very much, and maybe it's something for you to ponder as well. 18:35
Voldenet I've sort of given up from expecting much from unicode 18:36
m: "ł".NFD.say
camelia NFD:0x<0142>
Voldenet common polish letter, l with a stroke, is defined as character, so it would never match l anyhow… 18:37
doesn't put high confidence in the standard itself
clsn_ Yeah, Unicode has plenty of st00pid in it. Some of it comes from the fact that encoding letters is just plain more complicated than it sounds, but much of it is... well... yeah, st00pid. 18:39
They have some tables, I think, for dealing with stuff like what you're talking about in SOME cases, but I'm pretty sure not in that case. Whatever; I'm not here to defend Unicode. I am fully aware of its flaws (some of them; I'm sure it has more I don't know about yet) and will not dispute faults you find in it. 18:42
18:49 teatwo joined 18:52 teatime left
[Coke] I'm late, but if you're looking for the accent, then you probably want a different normalized form (with the combining chars split out), and then look for that. 19:15
m: say <e á é a>.NFD.grep: 0x0301 19:17
camelia (769 769)
[Coke] m: say <e á é a>.map(*.NFD).grep(*.grep: 0x0301).map(*.Str) 19:19
camelia (á é)
[Coke] there you go, that's more useful.
you could replace that inner grep with a \c[] with the combining char's name (or the decimal codepoint) or whatever. 19:22
This should also work if any of the graphemes have multiple combining chars. 19:23
19:29 derpydoo left 19:48 jpn left
[Coke] www.perlfoundation.org/the-perl--r...rence.html is only showing last year 19:49
19:51 jpn joined
clsn_ It's not an accent, and it isn't like I can list all the letters it might be on. And it isn't an NFC/NFD thing, because it isn't something that can be precomposed anyway. But thanks! 19:54
[Coke] then you should be able to see it in the ords for that grapheme, no? 20:03
(you should be able to skip the NFD step if it doesn't need decomposing, I mean.) 20:04
20:09 jpn left
clsn_ You would think. Hm, so use grep instead of ~~? But is that looking through codepoint by codepoint? Which might not be a bad thing, to be fair. 20:24
So long as it is done a bit more elegantly than just a for-loop through the whole string! :)
20:32 jpn joined 20:39 jpn left 20:43 jpn joined 20:49 jpn left 20:50 rf left
[Coke] I think this is a raku bug. Tried to install my own module, App::Unicode::Mangler, and got an error line like: 21:19
[App::Unicode::Mangle] Please u
[App::Unicode::Mangle] se uniparse instead.
I think something is trying to print "nice" whitespace there and failing.
21:24 perlbot left, simcop2387 left, perlbot joined
[Coke] m: "e̸".ords.say # see, this has the ords already - if it was combinable, you'd get the combined char here. 21:24
camelia (101 824)
21:25 simcop2387 joined 21:30 perlbot left 21:33 perlbot joined 21:40 perlbot left, perlbot joined 21:54 jpn joined 22:01 jpn left
guifa is nqp big integer the same as a Raku Int? 22:13
[Coke] ¡nʞɐɹ# 'oʃʃǝH 22:14
No nqp types are exactly the Raku types.
guifa how can I convert any old Int into a big int for nqp use? I'm trying to find the fastest way to shift the char codes of a string by X 22:15
22:17 simcop2387 left, simcop2387 joined
lizmat guifa: why would you need bigints for that ? 22:28
guifa errr, I guess there are actually two separate ops there and my brain is a bit tired hahaha 22:29
step one is to do some math on big ints (because I don't want to error if numbers are two big)
step two is then to shift the char codes by X
lizmat how would that look in Raku ? :-) 22:30
guifa the second part, @str.ords.map(* + $adjust-value)>>.chr.join 22:32
I've been testing around to see the fasest method
sorry $str
lizmat you realize that .ord will only produce the first codepoint of a grapheme 22:33
guifa Yeah -- in this case, it's a guarantee that it's a single codepoint
lizmat ok, check
so, if $adjust-value is 13, you're doing something like a rot13 22:34
guifa $str.trans( <0 1 2 3 4 5 6 8 9> => <a b c d e f g h i j>) is the fastest native Raku method, but has a huge start up penalty, so unless numbers are regularly 100+ digits, the current winner is $new := $new ~ ($_ + 49).chr for ^$a.ords;
yup
lizmat m: use nqp; say nqp::strfromcodes("foo".NFC) # does this give an idea ? 22:39
camelia foo
Voldenet `$new := $new ~ ($_ + 49).chr for ^$a.ords` 22:41
doesn't it malloc for every character? 22:42
No idea how can this be faster
lizmat m: use nqp; my int32 @a; @a.push($_ + 3) for "foo".NFC; say nqp::strfromcodes(@a)
camelia irr
guifa my int32 @temp; nqp::strtocodes($str, nqp::const::NORMALIZE_NFC, @temp); @temp[$_] += $adj; $str := nqp::strfromcodes(@temp)
^^ that's basically about 3% faster than the trans method 22:43
lizmat only 3% ?
guifa that's why I think there should be a faster way
lizmat += is generally not the fastest
guifa also when I tried nqp::for(…, …) it says it expects a block, but I give it one 22:46
lizmat nqp::for is an interesting beastb :-) 22:48
Voldenet the faster way would be to use cstring, then avx256 sum it with 0x3131313131313131
reject sanity, embrace xs
22:51 japhb left, japhb joined
guifa Voldenet: ha, yeah. I mean, I get I'm basically doing something that's solving a problem Raku wasn't made to solve hahaha 22:51
Voldenet: ha, yeah. I mean, I get I'm basically doing something that's solving a problem Raku wasn't made to solve hahaha 22:52
22:52 ugexe left
guifa It's just killing me I can't speed up number formatting by much more and probably 40% of it is not being able to do math on strings (understandable, Raku abstracts away a lot of that stuff intentionally) and 40% of it is wanting to supper arbitrarily large numbers 22:54
Voldenet actually I think that stuff like `$str.ords.map(* + 49).map(*.chr).join` could be rewritten into vectorized form 22:56
22:57 jgaz left
Voldenet on the optimizer leve 22:57
Nemokosch not sure if it would help here but did folks officially give up on moving away from libtommath? 22:59
in MoarVM that is
23:01 ugexe joined
guifa Thankfully for formatting with West Arabic digits I can skip the rot'ing, but with any others I'll need to add them in (thankfully, that's an easy optimization) 23:04
23:12 jpn joined
guifa okay this is ugly as sin but it's def faster 23:12
nqp::strtocodes($str, nqp::const::NORMALIZE_NFC, @temp); nqp::bindpos_i(@temp,$_,nqp::add_i(nqp::atpos_i(@temp,$_),$adj)) for ^@temp; $str = nqp::strfromcodes(@temp) 23:13
23:17 derpydoo joined 23:18 jpn left
guifa oh nice 23:19
changing out that for ^@temp with a my int32 $temp = nqp::elems(@temp); while($temp--, { ^^thatmess upthere }); knocks off another 15-20% 23:20
[Coke] is there a way to find out if your grapheme will render? 23:21
m: "d͖̤ᷛ᷼f͚ͯᷬ̒ ".uninames.say 23:22
camelia (LATIN SMALL LETTER D COMBINING RIGHT ARROWHEAD AND UP ARROWHEAD BELOW COMBINING DIAERESIS BELOW COMBINING LATIN LETTER SMALL CAPITAL G COMBINING DOUBLE INVERTED BREVE BELOW LATIN SMALL LETTER F COMBINING DOUBLE RING BELOW COMBINING LATIN SMALL LETTER…
[Coke] in my local terminal, that's a box with a ? in it. It's valid unicode, but my terminal can't display it. 23:23
guifa not from Raku at least -- you'd need to come up with some way to query the terminal, know what font it will use, and then figure out if the font has that character in its inventory 23:24
I think dwarren has some moduels for the font side of stuff 23:26
[Coke] s͔᷹o̟ᷔ ̵̢ę͚a̴̔s᷻́y͖ᷗ ̟᷽t̝̦o̵͡ ̯᷍gᷙ᷅o᷆̽ ̠ᷦoᷖᷪfᷧ̀f᷺ᷝ ᷇̏t̜̊hᷗ͘e̠͑ ̱ᷰṟᷮa᷇ͪíͭl̲ᷤs̩͍ 23:28
so easy to go off the rails 23:29
made some slight improvements to github.com/coke/raku-unicode-mangler - at least it doesn't generate invalid characters now, just a lot of unprintables. :)
Nemokosch 😄 23:32

Powered by Cro and the Raku® Programming Language.

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

Thank you!