David Neuzerling
@md.neuzerl.ing
2.7K followers 450 following 330 posts
Data scientist and R engineer. Also Python and Julia. Accidental econometrician. I mostly post about R, sometimes Georgism/YIMBYism and Warhammer. Melbourne, Australia.
Posts Media Videos Starter Packs
md.neuzerl.ing
Maybe they really like writing regex and hate that the computer does it for you now?
Reposted by David Neuzerling
clapifyoulikeme.favrd.social
Aw yeah that’s what charts were designed for
Lateralized Sleep Positions in Domestic Cats bar graph. 192 cats curve left. 109 curve right.
md.neuzerl.ing
Experimenting with a custom made static site generator. Super simple stuff: a build script that uses pandoc to convert markdown to HTML and then spatchcock it into a template file. Surprisingly successful so far
md.neuzerl.ing
At this point I’ve just made peace with the fact that divs cannot be centred and I’m much happier for it. Blessed is the mind too small for CSS
md.neuzerl.ing
LLMs have got me doubting my sanity. Y’all call spronions “spronions”, don’t you?
md.neuzerl.ing
I don’t blame the individual. Notebooks incentivise this.

By default there’s no where to put throwaway pieces of code, like a REPL. That stuff has to go somewhere so it ends up in the notebook.

It’s temporary code, and nothing is as permanent as temporary code.
md.neuzerl.ing
I am simple guy my brain is small. I am not smart enough to run a notebook.

It’s always stuff like “okay run cell 1 and then cell 3 but don’t run cell 2 because that’s just some throwaway code I wrote and oh you don’t have conduvpip installed so this won’t work”
md.neuzerl.ing
The worst thing about notebooks isn’t that they’re an incentive to write horrendously unstructured code, but that they displace proper IDEs. Lots of data science tooling is built with a mindset of “data scientists only use Jupyter”.

Plz I just want directories and plain-text files
md.neuzerl.ing
I believe kids should be protected from having to deal with merge conflicts until they are of a suitable age (at least 500 years old)
Reposted by David Neuzerling
jonobri.com
No, Australia does not need new cities.

My new essay sets the record straight: we have a lot of cities, but we aren’t using them as well as we could be. To make our cities more successful, we have to open up a lot more land for commercial uses to enable agglomeration.
md.neuzerl.ing
Disagree. I think we could build lots of big cities around Melbourne. Like Carlton, South Yarra, Richmond…

Seriously though, great article.
md.neuzerl.ing
My little victory is that I’ve introduced the word “spruik” to the Americans and some of them have started to use it
md.neuzerl.ing
Once again explaining to my overseas colleagues that we have public holidays for football and for celebrating animal abuse. One day they’re going to call me out on this
md.neuzerl.ing
I use `next` and `break` in my for loops in R, and I’ll never stop
md.neuzerl.ing
Every programmer dreams about this day and it’s finally happening to me: one of my bullshit abstractions made a future change easier than it would have been otherwise
md.neuzerl.ing
Why is it that universities don’t want to offer their Master of Economics course online, but are willing to offer a degree in clinical ultrasound online? How does that work?
md.neuzerl.ing
That sounds like some bullshit. I’m so sorry
md.neuzerl.ing
Why do you think this NIMBY considers London, New York and Los Angeles to be good, but Hong Kong, Bangkok, and Mexico City are bad? What's the difference? What could be the distinguishing factors there? What a mystery!
md.neuzerl.ing
The neighbourhood kids have invented a game called “Wizards vs. Snipers”. The Wizards have summoned zombies, which is a rookie error because the Snipers are “casting headshot”
md.neuzerl.ing
That was it, yep. Lazy imports.
md.neuzerl.ing
I could never get uv working, unfortunately.

I found a way to locally import pandas or polars based on the class/module of the input, which stops me from forcing both on the user. It’s hacky, but it works!
md.neuzerl.ing
I have a function that is optimised to work across big lists, and I want to make it easier for pandas/polars people to use it directly instead of using apply/map.

I find the syntax for this in pandas/polars to be bad, especially pandas, so I made it so you can just do my_function(series)
md.neuzerl.ing
When you try to construct a new series in narwhals it expects that you specify the backend, which defeats the purpose
md.neuzerl.ing
That’s exactly the approach I used. Local imports to avoid depending on two rather heavy modules