This channel is intended for people just starting with the Raku Programming Language (raku.org). Logs are available at irclogs.raku.org/raku-beginner/live.html
Set by lizmat on 8 June 2022.
rcmlz That was a good reading, thanks Liz for the good stuff you create. 06:14
librasteve I definitely second that, it makes me think that raku should be renamed tortoise 08:11
I recall raiph wrote something on how everything in raku is built on something else (that can be used by coders if they are unhappy with the standard abstractions) words like Int.^HOW.^HOW.^HOW come to mind … wonder if lizmat you wrote anything on this? 08:14
lizmat I haven't written anything about that (yet) 09:26
rcmlz Is there already a publication emphazising/comparing best practice on using lazy-stuff over non-lazy-things? Maybe this would help people like me leaping forward - like „learning apply in R“ did. 09:35
Maybe with some latency/throughput comparison between booth aporoaches as „Code-You-Should-Try-Yourself“ 09:37
I remember trying the code from here: github.com/Raku/marketing/blob/9ba...053519.pdf 09:42
Just to find out that it is not faster as promised on the poster 09:43
➜ ~ time raku -e "(^∞).grep(.is-prime)[10000].say" 104743 raku -e "(^∞).grep(.is-prime)[10000].say" 0,28s user 0,04s system 115% cpu 0,274 total ➜ ~ time raku -e "(^∞).hyper.grep(.is-prime)[10000].say" 104743 raku -e "(^∞).hyper.grep(.is-prime)[10000].say" 0,47s user 0,08s system 158% cpu 0,350 total
So for me it would help to have code snippets to execute to understand and test the iterator/laziness-things. 09:44
sort of see the benefits in terms of: "seeing is believing" 09:48
lizmat since the time of that example, two things have happened: 10:09
the determination of .is-prime has become *much* faster, so the 10000th prime number almost takes no effort 10:10
m: (^∞).grep(*.is-prime)[500000].say; say now - INIT now 10:15
camelia 7368791
2.796253531
lizmat now because of the fact that .is-prime is much faster, the default batch-size for .hyper is waaay too small 10:16
m: (^∞).hyper.grep(*.is-prime)[500000].say; say now - INIT now
camelia 7368791
4.734258993
lizmat you the overhead of hyper kicking in
m: (^∞).hyper(batch => 2048).grep(*.is-prime)[500000].say; say now - INIT now
camelia 7368791
0.712647335
lizmat with a greater batch size, you *can* see that hyper makes a difference
sadly, the design of .hyper does not allow the batch size to be adapted "on the fly" 10:18
that's what I tried to address with raku.land/zef:lizmat/ParaSeq
however, it looks like that module will need some TLC and possible some re-design at some stage as well 10:30
rcmlz Thank you for the explanation. I enjoy using Raku in its current state a lot, so no stress on that. And Raku execution speed was so far no issue for my problems. 11:57
You mean the "red flags" from the actions? 11:58
lizmat yes... the red flags are because some of the tests fail... and they fail intermittently, pointing at some kind of race condition 12:01
which I haven't been able to fix yet :-(
also, it looks like the overhead in ParaSeq is too high, need to think of a better way to pass performance info between the dispatcher and the threads 12:02
rcmlz I asked ChatGPT what programming languages actually have "batch size of parallel executed tasks adapted on the fly". It came up with Python, Julia, Scala, Rust, Go, C++ and Raku: 7. Raku Raku supports parallel constructs (start, await) and hyperoperators (>>, <<). While not as sophisticated as dedicated frameworks, custom task schedulers can adapt workloads. Would you like to see a Raku example for such 12:08
adaptive batching?
The example it gave me was not adjusting the batch size of parallel execution, but of single-thread execution ... 12:09
antononcube Wolfram Language / Mathematica has parallel batch size calculations “a on the fly.” (Of course!) 12:12
rcmlz If it is not in ChatGPT - it does not exist! 12:13
You know, in the old days "you had to be on the first page on google", but that might change ;-) LOL 12:14
PS: It probalbly does not realy matter if Python or any of the listed languages really have that capability - e.g. to my understanding removing the GIL in Python just started recently - but "information" from sources like ChatGPT might influence a lot of "less technical" people as falsification is a lot of effort that is probably avoided by those people. 12:20
antononcube Chat-GPT - Shmack-Shmi-PT -- using it is so 2023 or 2024. In 2025 what matters is what DeepSeek says. 13:07
Which means, I guess, that I have to demonstrate soon how to use DeepSeek with Raku. 13:08
rcmlz LOL 13:38

Powered by Cro and the Raku® Programming Language.

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

Thank you!