Jonathan Carroll

#rstats primarily but learning #julia, #rust, and #python

:rstats: :julia: :rust: :python: :bash:

Happy to also talk #chess, #woodwork, #biotech, #physics, and lots more

Public Key

npub1kl6fnrhma07s6e67qecufyhhvyx8emckuase0x9u0emprjltzrrqeggf30

NIP-05 Address

jonocarroll_at_fosstodon.org@mostr.pub

Profile Code

nprofile1qqst0aye3ma7hlgdva0qvuwyjtmkzrruautwwcvhnz78uas3e043p3spz4mhxue69uhhyetvv9ujumt0wd68ytnsw43q6s0x96

Publishing to

relay.mostr.pub

Author Public Key

npub1kl6fnrhma07s6e67qecufyhhvyx8emckuase0x9u0emprjltzrrqeggf30

Show more details

Last Notes

2024-08-07 11:59:55

- reply

@npub1324…8qla worked like a charm, many thanks!!!

2024-08-07 04:30:58

- reply

My favourite part is the fact that the thing I submitted to the queue really is just a bash script that orchestrates parallel processing of more bash scripts that do some actual work.
I've learned a lot of cool bash tricks on this project.

2024-08-07 04:22:22

I'm fortunate in that I don't often get nervous, but I just submitted a script to a queue that should (all going well) process about 8PB of data 🙀
Bioinformatics is big.

2024-08-06 06:23:49

Is it maybe time that the `sets` functions were generic?
I don't know if this has been discussed (probably has) - its so hard to search the mailing lists.
#rstats
https://cdn.fosstodon.org/media_attachments/files/112/913/630/422/461/106/original/0fcecae89d9e4821.png

2024-08-05 00:19:12

- reply

@npub14d7…dm4s I think so, though anonymous functions rule out name lookups. It's just a higher cost than I expected.

2024-08-04 23:03:10

- reply

@npub10py…xxjm cheers, good to know - this was not one of the things on my list of optimisations, but will be now. Though I'd likely drop down to Rust the moment there was any suspicion that performance was at stake.

2024-08-04 12:36:21

- reply

@npub1vpp…mutu I suspect the overhead is well below the noise floor of an expensive function, for sure. I was just surprised it was so stark a difference on this simple example.

2024-08-04 12:35:00

- reply

@npub1vpp…mutu I'll leave the Rcpp vs Rust comparison for another time, but I feel much happier writing the Rust.

2024-08-04 09:48:46

My first example with enableJIT(0) still shows a 50% difference, so *some* of it was JIT, but not all of it.
https://cdn.fosstodon.org/media_attachments/files/112/903/117/793/853/210/original/b3bd02335eeef2ba.png

2024-08-04 07:31:33

R function call overhead: https://jcarroll.xyz/2024/08/04/r-function-call.html

2024-08-03 13:48:34

- reply

@npub1xt7…aztg much cleaner now, thank you!

2024-08-03 13:29:57

- reply

@npub1xt7…aztg the history doesn't have it, but it was even more convoluted originally.

2024-08-03 13:26:56

- reply

@npub1xt7…aztg mostly because I guessed wrong when I first implemented it and tried directly using the default method instead of unclassing - I have good tests, so I'll try refactoring. Cheers!

2024-08-03 02:58:48

I wanted to see what having iterable strings in R would feel like, and I actually love it!
Introducing {charcuterie}
Post: https://jcarroll.com.au/2024/08/03/charcuterie-what-if-strings-were-iterable-in-r/
Package: https://jonocarroll.github.io/charcuterie/index.html
#rstats
https://cdn.fosstodon.org/media_attachments/files/112/895/845/072/150/256/original/5876007ffcafc0b2.jpeg

2024-08-02 02:49:29

These posts are always thoroughly entertaining to read. I want to say "full of great points" but that makes it sound like it's just "things I happen to agree with" - they're full of well-reasoned points that we're all welcome to discuss and agree/disagree on.
"our best defense against stupidity is not authority, but diversity [...] a bunch of weirdos trying to figure things out together, and nobody gets to stomp on anyone else’s neck, even when they really want to.
https://www.experimental-history.com/p/good-ideas-dont-need-bayonets

2024-08-01 07:07:15

The R/Pharma APAC track call for submissions is being extended - everyone in R and Pharma or adjacent (e.g. bioinformatics, other languages) is invited to submit a proposal for a talk.
More details: https://rinpharma.com/post/2024-07-17-apac-track/
#rstats #rinpharma @npub1tc6…e8rd

2024-07-31 23:52:26

Book review: Data Analysis with AI and R: https://jcarroll.xyz/2024/08/01/book-review-data.html

2024-07-31 12:51:45

- reply

@npub1rq8…wmnl I imagined a helper every(x) that returned "*/x"

2024-07-30 23:29:45

Lots of good points in this post by @npub1dy4…9aec about adding in flexibility
https://tidyfirst.substack.com/p/responsible-slack
I need to put
“Do not tighten the screw at this time.”
at the top of all my in-development project files.

2024-07-27 13:22:12

- reply

@npub108m…4af8 I just got a copy from my local library network in South Australia and am looking forward to reading it!

2024-07-19 03:28:05

- reply

I'd love to see a conf where all the talks were recorded ahead of time and each session was instead used to discuss it, instead of the 2 minutes allowed for whatever questions immediately come to mind. The rest of the time would be for starting and strengthening networks between people with similar interests, particularly the topics they are now more acquainted with.

2024-07-19 03:27:50

- reply

Especially in a world of unprecedented information accessibility, we spend way too long having each presenter talk at us - sometimes literally just reading slides that we could digest ahead of time and spend the conf discussing, debating, exploring...

2024-07-19 03:27:28

Gah, Humboldt had it figured out nearly two centuries ago - the point of a conf is the people there, not exposition.
"Rather than being talked at, he wanted the scientists to talk with each other. There were convivial meals and social outings [...] Humboldt encouraged scientists to gather in small groups and across disciplines. He connected the visiting scientists on a more personal level, ensuring that they forged friendships that would foster close networks."
https://www.conferencesthatwork.com/index.php/event-design/2023/11/alexander-von-humboldt

2024-07-18 11:41:39

Oh, that's ... an unfortunate name collision. https://github.com/liborty/rstats isn't the :rstats: #rstats I was thinking of. At least one of the other talks will be, though.
#SciCompInRust24

2024-07-17 08:52:14

- reply

@npub1unf…4qc4 Sweep, you say?
https://youtu.be/MqcinlOYIyc

2024-07-17 08:44:22

- reply

Oh, and it's just perfect for the comment section to be littered with anti-vax, anti-toothpaste, anti-moonlanding, etc... comments.

2024-07-17 08:38:50

- reply

One point that stuck with me from Loonshots - transistors didn't come about because someone wanted smartphones, they were incrementally better than the thing they replaced.
You don't necessarily need to be aiming for the moon, but being able to contribute that incremental improvement is essential to getting there.

2024-07-17 08:36:18

Adam Mastroianni has fast become one of my favourite writers but I don't think he's on Fedi.
His latest post https://www.experimental-history.com/p/whos-got-the-guts-to-go-to-the-moon while perhaps a little naively unfair to what progress has been made in cancer research and how complex that problem is, highlights the wastage of low-risk investment into ideas.
"[...] you’re not doing a moonshot. You’re just doing normal science in a futuristic font."
If you like that article, I strongly recommend Loonshots by Safi Bahcall.

2024-07-17 03:56:40

- reply

@npub1unf…4qc4 Rockafeller Skank vibe
https://youtu.be/FMrIy9zm7QY?t=124

2024-07-11 09:44:57

In the fandom extension of the Mario universe, Bowser can be transformed via the Super Crown power-up into Bowsette, a gender-swapped, anthropomorphised version of Bowser. The presence of a Baguette therefore implies the existence of an as yet undiscovered masculine Baguer. In this essay I will

2024-07-10 12:15:38

- reply

@npub1rq8…wmnl rweekly links to diffify e.g. https://diffify.com/R/rextendr/

2024-07-10 03:34:56

- reply

@npub1unf…4qc4 if nothing else suits, perhaps what's here can be of use https://github.com/Jobsecond/random-midi

2024-07-08 05:16:47

- reply

@npub1z9u…pehg yeah, I don't have that. The tool is for taking a file tree snapshot since for whatever reason that can't be done easily on the site or via the API.

2024-07-08 03:55:30

- reply

@npub1z9u…pehg I don't think I ever got that working, so I built a recursive tree listing tool that I've used quite a bit https://gist.github.com/jonocarroll/d5d315f1504d4cf8a06caf0bfdf58af5

2024-07-07 02:20:39

- reply

@npub1rn0…2anh maybe the {icecream} package https://jcarroll.com.au/2023/11/07/print-debugging-now-with-icecream/ can be helpful?

2024-07-06 09:42:55

- reply

@npub1wzf…fkwn that's a good idea, cheers!

2024-07-06 02:31:30

I broke my blog's RSS feed and didn't notice because nothing went wrong for me - if you do, how do you monitor your site's RSS feed to make sure it's working/valid?
The breakage itself was strange - it looks like a daily version of RStudio was overriding some {cli} options when knitting Rmds despite setting them manually inline so my most recent posts had ansi control characters in them which didn't pass a validator. It didn't seem to be because of a dependency, and updating RStudio fixed it.

2024-07-03 00:40:25

- reply

@npub1r2k…8k6e Settings -> Extensions -> RStudio Keymap: Enable

2024-07-02 12:53:51

- reply

@npub1s8j…qv29 I guess my point was it's buried a bit deeper - installing a language tool chain will have lots of eyes on it. An Action that checks some formatting maybe not so much.

2024-07-02 12:02:14

- reply

@npub12f4…c89f
https://cdn.fosstodon.org/media_attachments/files/112/716/788/072/576/373/original/d97dc67db2953ca8.jpg

2024-07-02 00:53:03

Tempted to fill in every market-segmenting-data field with the phrase "personal data and information" for any app whose changelog is only ever "bug fixes and performance improvements"

2024-06-27 01:11:38

- reply

@npub1cfp…xugs one already supported pattern I've seen for this is to name the 'internal' (but exported) function starting with a dot - it signals that it's internal, but it's available for third-party use, e.g. as .bye()

2024-06-25 02:24:08

- reply

@npub1vzq…lfmh hmmm... I didn't consider having nothing between two digits - I'll need to see how that changes the solutions. Cheers!

2024-06-24 06:32:17

- reply

@npub1z9u…pehg Amazon - not particularly cheap but hasn't disintegrated after being steamed regularly.

2024-06-24 06:04:07

- reply

@npub1z9u…pehg I've gone as far as getting a waterproof notepad in the shower - workplaces really should have debugging shower facilities, they're missing out.

2024-06-24 02:34:44

I've been working from home for many years now. I sometimes spend too long sitting at my desk, but other times I love getting up and doing other things while thinking about how to approach a problem or coming up with new ideas. It feels like pressing a 'Start Background Job' button on my thoughts. It's also one of the best ways to debug something.
https://cdn.fosstodon.org/media_attachments/files/112/669/249/372/266/340/original/7d71e379111fc3f9.png

2024-06-20 22:36:03

- reply

@npub1sy9…hnku type the numbers into a console and it might become clear

2024-06-20 22:35:26

- reply

@npub1r5v…4d7c see, if that was a strongly typed language, "yeah nah" would be a valid output when you tried to use nzchar() on a number.

2024-06-20 10:27:57

Producing HTML from #elm and #rstats as a way to learn #haskell https://jcarroll.com.au/2024/06/20/constructing-html-with-functions/
One day I'll stop adding more languages to my learning stack. That day is not today!

2024-06-20 09:29:56

Ooh, that's a fun (cursed) one... I know why, but daaaang. #rstats
https://cdn.fosstodon.org/media_attachments/files/112/648/233/350/727/545/original/50dbc01329abbe95.png

2024-06-18 11:21:05

- reply

@npub1tnx…zyf4 if I was doing it, a bit of butter then vegemite, but straight on there is good, too.

2024-06-18 00:52:44

- reply

@npub1avl…00r0 if the schema is SDL, this tool I wrote let's you walk through it building up a query https://github.com/HIBio/graphql.parse (I also have a fork in my own account)

2024-06-17 22:23:26

- reply

@npub1avl…00r0 you're not alone https://bessey.dev/blog/2024/05/24/why-im-over-graphql/

2024-06-16 01:50:44

- reply

I can't find the author on Fedi but there are also discussions on lobste.rs https://lobste.rs/s/qpzubc/don_t_refactor_like_uncle_bob_please and HN https://news.ycombinator.com/item?id=40661232 with some great alternative refactorings - worth a read

2024-06-16 01:49:36

I don't know which I like more - this excellent post about code refactoring and how it's perhaps not done all that consistently in Clean Code https://theaxolot.wordpress.com/2024/05/08/dont-refactor-like-uncle-bob-please/ or someone using the name 'Robert Martin' in the comments just blowing off any criticism with an "oh well".

2024-06-15 10:47:48

- reply

@npub1ade…tcdv I'm not sure if negation (vs subtraction) counts but neat all the same.

2024-06-15 07:56:20

Fancy a puzzle?
"With the numbers 123456789, make them add up to 100. They must stay in the same order but you can use addition, subtraction, multiplication, division, brackets etc. All numbers must be used exactly once."
https://jcarroll.com.au/2024/06/15/digits-dilemma/
#haskell #rstats #julialang

2024-06-07 12:48:13

- reply

@npub17x4…lfz2 I've hit this one myself a handful of times - I think it gets triggered when some "bad" data gets sent to be rendered as a table, but I could never quite pin down what precisely would cause it.

2024-06-06 05:32:16

- reply

@npub1rn0…2anh you're more on top of both languages than I am, but I was under the impression that the trait:S3 correspondence was already pretty strong https://fosstodon.org/@jonocarroll/111765571489488123

2024-06-06 01:53:14

Enums in R: https://jcarroll.xyz/2024/06/06/enums-in-r.html

2024-06-06 00:10:47

This is great - https://graphite.dev/blog/onboarding-roulette - make your own staff go through the pain points enough that they actually fix them, rather than waiting for customers to suffer through.
So many times I'll use a product and be left saying "do your own staff even look at this?"

2024-06-04 03:46:01

- reply

@npub16dy…jdnq went down differently to this one https://www.smbc-comics.com/comic/wolf
https://cdn.fosstodon.org/media_attachments/files/112/556/290/784/225/390/original/a506d8ac1bb45835.png

2024-06-03 09:20:36

- reply

@npub1gzh…gdr8 my favourite renaming of `git blame` is "whose line is it anyway?"

2024-06-01 12:49:17

- reply

@npub18ya…m4jm I needed to run `apt get update` then `apt install r-base` but otherwise started with a fresh install today. Best of luck.

2024-06-01 05:27:27

How do you do personal knowledge management?: https://jcarroll.xyz/2024/06/01/how-do-you.html

2024-06-01 02:21:52

- reply

@npub1z9u…pehg I second the porkbun recommendation. I also use VentraIP https://ventraip.com.au/domain-names/register/

2024-05-29 13:28:44

- reply

@npub1z0u…nf4t the power operator is one that perhaps does *too much* - I showed an example of it applying a function 3 times, but if you pass it negative 1... it uses the inverse of the function (!).

2024-05-29 05:18:32

I just *love* opportunities to learn like this https://jcarroll.com.au/2024/05/29/iterative-square-root/
Solving something teaches you so much more than just reading about it. "You learn with your hands, not your eyes" and all that.

2024-05-29 05:14:17

- reply

@npub1z0u…nf4t In #apl it's
sqrt←{⍵{(⍵+(⍺÷⍵))÷2}⍣=1}
https://jcarroll.com.au/2024/05/29/iterative-square-root/ (with an explanation about _why_ it works)

2024-05-27 00:03:08

- reply

@npub1tnx…zyf4 excellent - I've been waiting for that book to be stable before I read it, but that's definitely in line with the OOP approach from the Google post.

2024-05-26 23:29:14

I understand the appeal of encapsulation in this post https://testing.googleblog.com/2024/05/avoid-long-parameter-list.html but I'm wondering how people deal with this from a functional programming perspective?
Pass in a new (potentially classed) object? For #rstats, ggplot2::ggplot() comes to mind; we pass in an "aes" mapping rather than e.g. aes_x, aes_y arguments, but I've seen some lengthy parameter lists.
Not to say that FP forbids encapsulation at all, it's just something I've not seen discussed a a lot in that space.

2024-05-21 02:03:23

Someone's built a way to call #rstats from #uiua (a stack-based glyph language in the style of APL/BQN) so now we can have powerful statistics functionality while working super efficiently with arrays! They've also wrapped simple base plotting!!!
https://cdn.fosstodon.org/media_attachments/files/112/476/607/461/700/193/original/ae9a108c094c7f6c.png

2024-05-18 13:59:44

I feel like I've missed the boat on most of the Discords I've tried to join. Work and personal slacks (17 of which I seem to be logged in to right now) have maybe a couple of posts per day at most, but Discords all seem to have a devout group of dedicated posters (plus hundreds of lurkers) forming a tight-knit community with dozens of posts *per channel* that feels imposing to chime in on. Just catching up with what happened overnight once seems like a lot to take in.
Again, I may just be old.

2024-05-18 13:19:56

- reply

@npub1z9u…pehg this is borderline extreme Australian but I ordered 2 potato fitters (ugh, scallops for the other states) and got 2... that's not how it works - I order 2, you're supposed to give me 3.

2024-05-17 13:46:56

- reply

Being able to pause reading and ask "tell me more about [this]" or "how does [this] work?" greatly adds value to the experience of reading non-fiction. Heck, I even asked for some examples of fiction where [this] was explored and got what look to be reasonable suggestions. Keyword search just can't do that.

2024-05-17 13:43:48

Setting aside all the valid objections around the training data for a moment, I have found significant value in asking questions to an LLM as I read a fairly dense science book. When I find something I'm curious about, or have further questions on - I can ask, just as I would someone potentially knowledgeable on the subject. It may be wrong, but so might the person, so I'll verify, but digging through keyword searches for the answer would probably lead me astray, too.

2024-05-16 12:18:22

- reply

@npub1ckn…nde5 a cool project (presumably, i'm out of my depth to judge) but I'm more interested in the name choice, given the adjacency to https://r-project.org/

2024-05-16 10:11:46

- reply

@npub1gzh…gdr8 my to-read list grows larger by the day... this quote piques my interest, but would you recommend it?

2024-05-16 07:02:45

I'm disappointed (maybe relieved) that all the current chat models I tried were able to pass the Idiocracy IQ test question, despite this https://toot.cat/@devopscats/112445057997076822
https://youtu.be/0p43MhMB7Kg

2024-05-11 12:06:18

- reply

Ah, remembering all too late that this new phone has a 200M pixel mode
https://cdn.fosstodon.org/media_attachments/files/112/422/366/077/710/264/original/3d0178ebb2486ab3.jpg

2024-05-11 11:47:35

As far north as #Adelaide South Australia. Whoa!
https://cdn.fosstodon.org/media_attachments/files/112/422/259/141/375/843/original/61e163fffc439665.jpg

2024-05-06 00:30:21

The web is not dying
https://manuelmoreale.com/the-web-is-not-dying
> The web is not dying. The web is huge. The web is ever-expanding. The fact that the web is just the same 5 big websites is a fucking lie. It’s like saying the restaurant industry is the same 5 fast food chains. It is not. It’s up to you to decide to stop visiting those 5 sites and stop ingesting their fast food content.

2024-05-05 23:48:31

On content creation motivation: https://jcarroll.xyz/2024/05/06/on-content-creation.html

2024-05-04 12:47:49

Things That Make Us Smart by Don Norman (fourth book in the series starting with The Design of Everyday Things, also a great read) comes out swinging in the preface.
So many things made difficult for us so that they're easier for the machines is definitely a sign that we're on the wrong path.
https://www.goodreads.com/book/show/16868.Things_That_Make_Us_Smart
https://cdn.fosstodon.org/media_attachments/files/112/382/863/036/099/379/original/b8a22352133fc7c7.png

2024-04-27 09:30:17

- reply

@npub1fs2…sltu /kæʒ/
https://en.m.wiktionary.org/wiki/cazh

2024-04-23 22:45:21

- reply

@npub18cr…3r4s gotta have them at some point, and I really don't have many, but "early morning meeting cancelled while you were asleep" is probably near the top of my 'most hated things' list.

2024-04-23 22:12:07

The pros and cons of working remotely from a significantly different timezone to the main team:
Pro:
- chunks of work to complete without interruption
- flexibility about precisely when during the day the work gets done
- better documentation
- async communication
- fewer meetings
Con
- a 7:30am (your time) meeting being cancelled while you're asleep which means you still need to get up during school holidays only to see that it's cancelled.
<grumble>

2024-04-22 03:57:20

- reply

@npub1xf0…fw0n I'd say wifi unless you think you'll need more than 100mb/s or have issues with walls.
I have a crimp and connection test kit if you'd like to borrow it.

2024-04-14 02:37:07

I've become obsessed with #Obsidian for notetaking and am now cataloguing everything I think of.
To be honest, it's very comforting knowing that my notes, thoughts, and reminders are stored somewhere as a 'second brain' in the Tiago Forte sense.
This clip came to mind and I just had to make the image https://youtu.be/aACQjYEmOEo
https://cdn.fosstodon.org/media_attachments/files/112/267/225/650/255/607/original/49080ee5a0dec0e5.jpg
https://cdn.fosstodon.org/media_attachments/files/112/267/239/637/460/503/original/5118d0210995849d.png

2024-04-11 02:57:10

- reply

@npub1tnx…zyf4 in that case, spread some on a dry weet-bix sometime.

2024-04-11 02:47:40

- reply

@npub1tnx…zyf4 hopefully someone can give you a proper introduction. Toast, butter, black gold; find the ratios that work for you.

2024-04-06 10:08:37

- reply

@npub10y4…099z httptest2::without_internet() ?
https://search.r-project.org/CRAN/refmans/httptest2/html/without_internet.html

2024-04-06 00:10:43

- reply

@npub1y78…qvty woo! Back when bands actually came to Adelaide.

2024-04-03 05:29:26

There's lots of good points in this post from @npub1740…h2zd
https://registerspill.thorstenball.com/p/the-basics
but this one in particular stands out to me:
"Understand what problem you’re solving. Knowing why you’re doing something is a requirement to knowing whether you’re actually solving the problem."

2024-04-02 23:04:58

- reply

@npub1wky…6c6c "publish or peri-peri" was right there

2024-04-02 05:58:19

- reply

@npub1ser…ztzz {ntfy} ( in R: https://github.com/jonocarroll/ntfy and many other ways ) or magic wormhole or create a draft email in gMail with the text then open it on the other device.

2024-04-01 08:03:50

This is driving me crazy. My daughter has M365 Word for school and wants to insert an image off to the side of an assignment. Every time we click 'Wrap -> Behind Text' we get the movement controls but it immediately falls back to 'in line with text' and messes up the content.
I'm not finding much help online... Is this just not supported in the web/M365 version???

2024-03-30 02:39:55

If I had a dollar for every time I'd seen the crew of the Firefly used in programming language syntax documentation ... I'd have two dollars. Which isn't a lot, but it's weird that it happened twice. Right?
https://elmprogramming.com/list.html
https://docs.swift.org/swift-book/documentation/the-swift-programming-language/guidedtour/
https://cdn.fosstodon.org/media_attachments/files/112/182/308/283/841/022/original/ec6fa1fd8109c191.png
https://cdn.fosstodon.org/media_attachments/files/112/182/308/305/387/671/original/6aa52f876b58063f.png

2024-03-28 10:53:31

- reply

@npub1nll…xx3v I once had a corporate spyware tool that got greedy with memory so I had a cron job run every 5 mins to `kill -9` it ... depending on the use-case, that works.

2024-03-26 21:38:25

Seen in the wild: a 'written by human, not by AI' badge from https://notbyai.fyi/
My immediate thoughts:
1) maybe I'll add these to my own posts (100% human-produced content)
2) how long before AI stuff has this slapped on it regardless?
https://cdn.fosstodon.org/media_attachments/files/112/164/145/698/351/822/original/0fe16a2d25ad7065.png