3 nukes power plant stopped in France (water rejected was getting too hot).
This is your monthly reminder that nuke is not the one trick solution and lowering consumption is still king!
Remote
Olivier Leroy
@defuneste@fosstodon.org
mastodon 4.7.3I am a geographer interested in all things spatial and open science!
he/him/homard
1022 Followers
916 Following
50 Posts
Joined October 26, 2022
a coworker asked me how to style a basemap vector tile (you know one of the kind of issues they were developed...). I could not figure how to do that with their access to arcgis pro (agol is supposed to works easily, no idea).
It was one copy paste, 3 clicks in #QGIS and very transparent.
One day govt sector will wake up and rise!!!
#gischat
Replying to
@lina@vt.social
I think your are displaying the same behavior you are criticizing.
side note: pressure on AI is crazy that is the whole point.
@whitequark@social.treehouse.systems
it seems France want to move away from windows to linux. If they use their own distro how should it be named? (obv wrong answer only!!)
every time I used the compost to plant stuff I feel so smug! Hahah I hacked the system and did not pay for good soil!
random post appreciation about jq
I do not use it a lot, forgot half the syntax most of the time but it still do everything you can dream with json and fast!
well even if news are saying "junior" do not exist anymore, I still find useful to gave less experienced people specific tasks that need to be done in one file (since they usually struggle with git) with clear input/output. "Having AI" is not a justification to give (or try to give in my case 😜 ) less structure.
(it may sound obvious but sadly I feel it need to be repeated)
Replying to
@alisynthesis@io.waxandleather.com @MichaelTBacon@social.coop @jillL@theblower.au
Yes the article is misleading, growing stuff can be hard, require equipment, I also sometimes fail while replanting seedling.
It is also not like you will learn everything in one season.
On the other side, collective gardening exists since long time.
You can share land, equipment, knowledge, plants etc. My grand father was able to go on strike for a month thanks to that and those gardens. I know the article is not advocate that but we should.
Replying to
@eamon@social.coop
good piece!
it is not needed for most of their current use cases and yes learning how to use a computer well or basic statistics (I mean basic!) are still a better skills than prompting.
None of them require training big models/data centers.
The current media coverage in the US is insane!
Replying to
@jonocarroll good post (and I 100% agree)! I have not try duckdb but my memory using PG is that with dbt you can also set up a staging / per user /branch. You are not just promoting code but also data (kind of). In fact you need it for validation.
You can also do that with targets using multiple "stores". I think gittarget help here but I usually use my crappy function that switch store if not on main branch.
I find validation tests slightly easier with dbt but never tried data.validator.
If you want to see some very real world data "analytics" in "production" (quote are from video)
From the great Cleveland #Rstats users groups
Replying to
@mdsumner I do not know which one you are now! but join us!!!! one more drink! one more software!
Replying to
@b0rk Thanks for sharing! It was a great read.
Still a bit sad that the author does not link it to any "political" ideas or forms of organizations. In some (plenty) of ways github (and previous forms he mentioned) was a product of a decade that is gone.
Replying to
Replying to
@Gina@fosstodon.org well that got me the opportunity to boost your LI post to my huge network (😉 ).
i guess LI apply some Noblesse protocol:p
Replying to
@geospacedman the plebe want a blog post! (so you can advertise your cartography consultancy 😉 )
Replying to
@SReyCoyrehourcq@social.sciences.re I am not in that continent, albeit I will go to the useR conference (and I am also a noob guix user)
Replying to
@LeafyEricScott@hachyderm.io @cd_newton@hachyderm.io
still good to know..
I usually write duckdb sql and wraps it with either dbExecute or dbGetquery (from DBI) maybe that could work?
Replying to
@trevorflowers@hachyderm.io @gvwilson@mastodon.social
I am always, maybe over, concerned on relying on a "file" for that. You will end build your own fake "WAL", adding some gotcha columns for validation. It is nice to have SQLite deal with that.
I will go the csv/files way if I will be the main user or if the csv can be rebuild from scratch.
Replying to
@LeafyEricScott@hachyderm.io You do not want (can't?) specify a column type while reading the multiple csv? You can always read as text and cast it later (worst case scenario).
In the blind, I would use duckdb to read them with globing (defining the column type I expect) and specify what I do with error (see store_reject = true).
That will not solve your "sometimes 3 columns or not". I would use {targets} with a custom function that check and "clean" every csv and then append.
Replying to
@cwebber@social.coop you are right it is just that I feel, at least in the US, that our industry lost their political "leaders" and we need to rebuild from the ground.
Replying to
I do not know much but giving infinite money to some folks seems to cause issues..
Replying to
@jonocarroll @noamross @tylermorganwall yes it is easier in a culture of scripts/modules than in a more package one. Shiny is also similar (and would count "before" quarto).
Replying to
@jonocarroll @noamross @tylermorganwall
Now that I think also, I would say that for the user dbt is "independent": you just need to learn the framework.
{targets} is less opinionated and rely on other parts of the R ecosystem. I am still using it inside an R package (and always unsure if that is "right"), use testthat for unit tests (but also to do some "data structure" check) and rely also on R package for documentation... It would be beneficial to "framework" it a bit.
Replying to
@jonocarroll @noamross @tylermorganwall 💯 I definitely did stuff with targets that were bordeline using PG has back end and using time stamp and/or number or row return for targets to keep track.
I think the test set podcast make a good case of explaining why it make sense in a corporate setting: SQL people, need of contract between team, etc (but fail to see that targets was here :P)
Replying to
@noamross @jonocarroll @tylermorganwall I tested and it with PG mostly and it cached data. This is also one issue with it: not everything is transparent. I think it does different stuff depending on the engine used. This is cool since it is an abstraction layer but sometimes this is also a default.
Your sql files are in particular directory and dbt call them for you. I do not think the framework is that flexible. I could be wrong since I did not try all the jinja magic directly.
Replying to
@jonocarroll
not @tylermorganwall but the scale is more about team members / organization structure and tools you need to connect than volume of data.
dbt with duckdb will be different than dbt with spark. I ran half TB data with targets and duckdb without (big) issue (but it did not need to run every few minutes).
Replying to
@sarahalang@fedihum.org I am too afraid to look at what could be male professor definition of "ambition" (outside of it being a total bs excuse).
Mastodon leftist give me the power to argue with some republicans that financing private schools (that cost between 13k to 25k / year before our "freedom of choice discount") mostly help rich families!
Edit: opening an other beer maybe that will gave me some patience.
Replying to
@gvwilson@mastodon.social and to be clear I advocate C nearly all the time, use something very similar (R wrapping duckdb targeting parquet in object storage), even gave a presentation for R users on it but my understanding of your workflow is that it will rely on the file system/operating system.
Here I think SQLite is safer (could be wrong).
Replying to
@mattiabunel@framapiaf.org @SReyCoyrehourcq@social.sciences.re @civodul@toot.aquilenet.fr
😬 (I surrendered and went for litedown lol)
Replying to
Replying to
@LeafyEricScott@hachyderm.io if you do not need column 3, I would cut in shell it before reading the csv.
#podcast commute! Super great episode of test set with Joe Cheng:
https://posit.co/thetestset/episode/curiosity-duty-and-existential-dread-with-joe-cheng
Grumpy take: this is not only managers or "lead" that are "gapfilling"😉
Really appreciate that Joe is sharing his ethical concerns about LLM and his tough process (I might disagree but it is refreshing that is not dodging it).
To me it is not weird that Hadley trust more Claude in base R than himself. Would he have build a dialect if he did not?
this episode is in my test set podium.
#rstats
Replying to
Replying to
@jimgar@hachyderm.io @mdsumner@rstats.me @eliocamp@mastodon.social
I think the mental model of doing opérations in order is very strong but the hidden idea that an order can mean a relation (or not) is what make me struggle with for loop. (while are easier conceptually but slightly harder to implement)
"prompt engineering is a skill that employees in your agency will need to train"
Replying to
Replying to
@copdeb@metalhead.club @mhoye@cosocial.ca going slow is definitely better here than running.
Replying to
@cwebber@social.coop
I have an hard processing all of that. I fully agree with you on the structural aspect of it and not the worker fault.
The line gets blurry when we talk about "influencers" (are they workers? Like some president are we all influencers gambling now?), "our industry leaders" that should know how to test a product before injecting it everywhere. Even if we have a system on one end it seems some are dodging any kind of responsibility and putting it on "us".
Replying to
@tarsius@fosstodon.org for folks like me who hate recurring donations can you turn the option to donate once? (if that button exists in liberapay)?
Replying to
@eliocamp@mastodon.social @henrikbengtsson@mastodon.social @datastory@mstdn.ca
the sad point is income is usually a good proxy for urban heat ...
Replying to
@geospacedman @stevenfeldman size of image? ie not content but more "do you want to load that?" 🤷♂️
Replying to
@gvwilson@mastodon.social A would have been my go, agree on b (at that point just go a).
C is the most controversial. I would say give it a quick try / test. The polars/parquet feel more adapted for object storage and less mutable (If I can say that) workflow.
do you expect complex query, can some examples works?