šŸ¦‹ 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.
01:31 kylese left, kylese joined 02:15 kylese left 02:16 kylese joined 02:27 fredJ joined 02:41 fredJ left 02:59 Sgeo_ joined 03:02 Sgeo left 04:27 xinming left 04:52 oodani left, oodani joined 05:01 jrjsmrtn left, jrjsmrtn_ joined 05:30 Sgeo_ left 06:14 jrjsmrtn_ left 06:15 jpn joined 06:36 jpn left 07:09 jpn joined 07:26 avuserow_ left 07:27 avuserow_ joined 07:49 wayland76 joined 07:55 sena_kun joined 08:40 thaewrapt left 08:51 thaewrapt joined 08:57 thaewrapt left 09:07 thaewrapt joined, jpn left 09:26 thaewrapt left 09:38 thaewrapt joined 09:43 thaewrapt left 09:54 thaewrapt joined 09:59 thaewrapt left 10:00 jpn joined, thaewrapt joined 10:05 thaewrapt left 10:16 thaewrapt joined 10:21 thaewrapt left 10:31 thaewrapt joined 10:36 thaewrapt left 10:47 thaewrapt joined 10:52 thaewrapt left
tbrowder antononcube: interesting. nw florida is compressed to 2 places out of the whole graph. what are the factors creating the path nodes? 10:58
iā€™m essentially at the pensacola node. used to live between pā€™cola and tallahassee 11:01
11:02 thaewrapt joined
tbrowder line between cdt an edt timezones runs between pā€™cola and tallahassee 11:02
antononcube @tbrowder Well, my problem is that I cannot figure out yet a good way to plot graphs with vertex coordinates. I can only plot graphs using 4-5 different forces. (Like spring or electric force.) 11:06
This graph (tree) is only between FL cities with more than 50,000 citizens. 11:07
11:10 thaewrapt left
tbrowder try distance x population? 11:11
and f(lat long) 11:12
traveling salesman? 11:16
11:20 thaewrapt joined 11:23 wlhn joined 11:57 sena_kun left 12:10 jpn left
antononcube Well, the graph in the image is "minimal spanning tree" based on distances between cities. It is "initialization heuristic" for traveling salesman. 12:25
Distances between vertexes can be used to influence the graph layout. So, yeah, I have to verify that can be specified. 12:26
@tbrowder I forgot about the automatic graph distances -- thanks for remining me. 12:28
12:33 jpn joined
tbrowder glad i could help, been fumbling here a bit w/ macos, but itā€™s working well. my QueryOs helps a lot in differentiating test needs between the 3 os types. dipping toes into choco, too 12:48
really wanna encourage 3-os testing and showing 3 badges to highlight what works to noobs. a gray badge to anyone is offputting imho 12:51
antononcube How about red badge? 12:52
And, yes, I try to but the badges for my "heavier" packages with lots of dependencies. 12:53
Just put badges in the README of "Graph" -- good idea. 13:08
And as usual "Win64" fails. 13:09
ugexe over the years i've changed my opinion on badges that they are pointless for good software. developers should not be releasing broken software to their main branch to begin with, and their software should work on all OS to begin with. If a developer writes software that explicitly doesn't work for e.g. windows they certainly should not want a red broken badge showing up, so the badge doesn't work as 13:18
a marker for what OS it is intended to work on either
so as a developer looking to use software if i ever see a red testing badge on a repo i'll probably not use it nor look into using it again in the near future, assuming that it is not a finished product 13:19
Further, github already shows red Xs and green checkmarks for your CI actions which are redundant against the badges
arguably badges can make your distribution worse. consider that for badges to work they have to link to an external resource. the code you release with fez comes with a readme linking to that resource, but sometime in the future that resource is deleted or the url changes etc. now your released readme has a perma-broken image in it 13:25
i agree noobs like badges for some reason though. i know i did, although I don't know why in the present 13:27
13:36 jpn left
antononcube For me, badges are mostly "marketing." And I agree -- broken software should not be released. But in GitHub it is fine to have half-baked commits, so for me badges there are fine. 13:38
13:53 jpn joined
tbrowder but marketing is important iif we want to grow number of raku users, and too many devs are windows ppl, so windows green test badges are important 14:19
14:26 soverysour joined, soverysour left, soverysour joined 14:34 soverysour left
tbrowder btw, just found why linux ā€œfindā€¦ā€ wasnā€™t working on mac. because of brew using sym links i had to add ā€œ-Lā€ at the beginning: ā€œfind -L /opt ā€¦ā€ and i found the files i installed andneeded on mac for testing. 14:42
works for brew-installed pkgs 14:46
antononcube āœļø 14:51
14:57 Sgeo joined 14:58 thaewrapt left 15:03 Sgeo left
tbrowder ugexe: where do you see the checkmarks for OS on github? 15:09
the default ones?
antononcube @tbrowder See the "Actions" tab in a relevant GitHub repository. For example: github.com/antononcube/Raku-Graph/actions 15:12
tbrowder ugexe: ah, ok, i see your point, but my point is a noob raku person is going to search on raku.land, and the links there lead to the repo where the badges can be seen without checking another botton 15:15
*button
15:16 soverysour joined
tbrowder i should always say early on in the readme to file an issue if it doesn't work on *your* os 15:16
and that's why i want to publish that recent push to the community modules 15:18
15:19 Chanakan left 15:21 Sgeo joined, soverysour left
tbrowder 'course now one can probably ask siri: "on github show me all raku repos that have green action status for windows" 15:21
ugexe i would argue it is not the users job to tell an author they commit broken code. the author can just push their changes up as PRs and wait for them to turn green. or if they insist on merging into their main branch to check the checkmarks
tbrowder in yr arg who is the 'author' and who is the 'user' in my case? 15:23
of the community modules
ugexe as for seeing them on raku.land, my point about external resources / links still stands - once those urls change / go away, those readmes are broken forever
if you are committing code you are the author in this scenario 15:24
15:24 soverysour joined
_grenzo To help new module authors wouldn't it be a good idea to provide a starting point/documentation for setting up github actions to test on linux, mac, and windows? 15:24
tbrowder i don't understand why you save the links might change 15:25
ugexe because i've seen it happen
tbrowder _grenzo: yes for sure, that's what i'm pushing for, but no traction
ugexe travis-ci deleting old builds so those links to specific commits no longer exist. renaming the repo to not contain Perl6 in the name which causes some CI platform to not properly generate badge images for the old name 15:26
i think most people are willing to test on three platforms, but i think maintaining 3 seperate CI build files to do that is offputting some. Myself and others probably prefer using a single CI build file and to use a build matrix, but then you can't have a seperate badge for each OS (which I personally prefer) 15:27
tbrowder but nick, what do you see casing it? for community modules in particular? repo rename? author change? (jj takes over my github acc't when i pass)...got to go for a while 15:28
*causing
ugexe i mentioned travis-ci deleting old builds specifically
we do not know what CI platforms will do in the future. but we have evidence that some act in ways that break these type of things 15:29
roguerakudev I think there might be some overzealous optimization happening in actions classes for grammars 15:31
ugexe imgur.com/CpspiyO
in this image you can see a red X next to the commit message "Add github action". you can then click that to see the modal showing the build status on various OS
i imagine the github API allows access to those, which means raku.land could leverage those build statuses instead of having to explicitly include external images in readme 15:33
roguerakudev consider this actions class and the associated error: pastebin.com/S4LeQgkJ
15:33 [Coke] left
same result with or without .clone, same result if I assign into the variable first then mutate in another statement 15:34
however, the same thing works fine if I use the grammar without the actions class, then pull the data out of $/ manually the REPL 15:35
in the REPL*
15:35 [Coke] joined, kst`` joined
this is on Rakudo 2024.05 15:38
for reference, this is the grammar: pastebin.com/e7pqfY8j 15:39
15:39 kst left
ugexe my $escaped = S:g/ \\ \' /'/ given $<contents>.Str.clone; 15:41
evalable6 Use of Nil in string context
in block <unit> at /tmp/V3iJ8LY0TJ line 1
ugexe what does this do?
you're doing an inline search and replace but assigning what exactly to $escaped?
my guess is the search and replace is occurring on $/, which you can probably test by changing the signature to ($/ is rw) 15:42
roguerakudev shouldn't it operate on $_ which is set by given? my $e = S:g/ \\ \' /'/ given "abc\'def\'ghi" works as expected 15:43
but anyway it's capital-S substitution, which shouldn't actually try to mutate $_ 15:45
or whatever it's operating on
15:45 jpn left
is rw yields a different error, "Parameter '$/' expects a writable container (variable) as an argument, but got 'ļ½¢'two three'ļ½£ contents => ļ½¢two threeļ½£' (Tokenizer) as a value without a container." 15:47
input being "one 'two three'"
ugexe in some regard i imagine $/ has to act like a regex version of $_ given in regex we can write e.g. .<foo> instead of $/.<foo> explicitly
i'm not arguing the behavior is correct mind you 15:49
roguerakudev yeah, I could see something like that being the issue
but splitting it into two lines yields the same error, e.g. my $escaped = $<contents>.Str; $escaped ~~ s:g/ \\ \' /'/; 15:50
where we're being really explicit about what should be acted on, namely $escaped 15:51
ugexe might be worth trying to create a one liner that reproduces this you can post in here so others can probe it and iterate quickly 16:07
lizmat yeah, golfing is good
roguerakudev I've been unable to reproduce it outside of an actions class for an actual grammar 16:08
ugexe so create a 1 line example that creates a grammar and action class, and then uses it to get that error 16:09
roguerakudev I'll see if the same happens with a plain method that takes $/
sub foo($/) { return S:g/ \\ \' /'/ given $<contents>; }; "'ab\'cd'" ~~ /\'$<contents>=.*\'/; foo $/; 16:11
this reproduces it - I have to use backtricks or Discord will gobble backslashes weirdly
ugexe m: grammar Foo { token TOP { $<contents>=(.) } }; class FooActions { method TOP($/) { my $e = S:g/1/x/ given $<contents>.Str } }; say Foo.parse("1", actions => FooActions) 16:12
camelia Cannot assign to a readonly variable or a value
in method TOP at <tmp> line 1
in regex TOP at <tmp> line 1
in block <unit> at <tmp> line 1
roguerakudev backticks* 16:13
ugexe m: grammar Foo { token TOP { $<contents>=(.) } }; class FooActions { method TOP($/ is rw) { my $e = S:g/1/x/ given $<contents>.Str } }; say Foo.parse("1", actions => FooActions)
camelia Parameter '$/' expects a writable container (variable) as an argument,
but got 'ļ½¢1ļ½£ contents => ļ½¢1ļ½£' (Foo) as a value without a container.
in method TOP at <tmp> line 1
in regex TOP at <tmp> line 1
in block <unit> at <tmp> line 1ā€¦
ugexe that one is with a $/ is rw
roguerakudev yeah, and that's the same error I got as well
ugexe m: grammar Foo { token TOP { $<contents>=(.) } }; class FooActions { method TOP($/ is copy) { my $e = S:g/1/x/ given $<contents>.Str } }; say Foo.parse("1", actions => FooActions) 16:14
camelia ļ½¢1ļ½£
contents => ļ½¢1ļ½£
ugexe that one is $/ is copy
16:28 soverysour left
roguerakudev Very strange... something about the substitution operation turns $/ from a Match into a List 16:32
sub foo($/ is copy) { say $/.WHAT; my $e = S:g/ \ \' /'/ given $<contents>; say $/.WHAT; make $e; }; "'ab'cd'" ~~ /'$<contents>=.*'/; foo $/;
dang it, backslashes got gobbled
m: sub foo($/ is copy) { say $/.WHAT; my $e = S:g/ \\ \' /'/ given $<contents>; say $/.WHAT; make $e; }; "'ab\'cd'" ~~ /'$<contents>=.*'/; foo $/; 16:33
evalable6 (exit code 1) (Any)
Use of uninitialized value ā€¦
Raku eval Exit code: 1 ===SORRY!=== Error while compiling /home/glot/main.raku Cannot use variable $e in declaration to initialize itself at /home/glot/main.raku:1 ------> / given $<contents>; say $/.WHAT; make $āe; }; "'ab\\'cd'" ~~ /'$<contents>=.*'/; expecting any of: method arguments term
evalable6 roguerakudev, Full output: gist.github.com/1bc9e53a4b389898f2...30bd040ec6
roguerakudev Well, that's not what I'm getting in my REPL
I get (Match)ā¤(List)ā¤The make function expects $/ to contain a Match, but it contains List 16:35
16:47 soverysour joined
tbrowder ugexe: ref yr ci comment: i don't think we'll have that problem with GitHub workflows, reading GitHub's docs (hard as it is), they seem to be aiming for regimented API's, but their docs are hard for me to read (too much segmented into tiny explanations). i do like a more linear cookbook recipe approach. ( i got a lot of mileage from the Perl Cookbook) . but i appreciate your views. you and lizmat and tonyo and many others are in 16:57
the major leagues, and the rest of us are lucky to have you on our favorite programming team šŸ˜„
17:10 swaggboi left 17:21 swaggboi joined 18:15 soverysour left 18:41 xinming joined 18:49 sena_kun joined 19:02 jpn joined 19:10 jpn left 19:20 jpn joined 19:29 jpn left 19:31 jpn joined 19:39 jpn left 20:07 dmvrtx left, dmvrtx joined, wayland76 left, dmvrtx left, dmvrtx joined 20:11 wayland76 joined 21:11 thaewrapt joined
tbrowder i have a font file that ā€˜encguessā€™ defines as UTF-32BE. but raku docs only show up to UTF16BE. is there any way to handle that with slurp? 21:56
22:04 sena_kun left 22:13 Maylay joined
tbrowder ah, iconvā€¦ 22:26

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!