Advertisement · 728 × 90

Posts by Tin Tvrtković

github.com/Tinche/quatt...

Why did it make me so long to do this? Video games probably

3 weeks ago 0 0 0 0

Played a little Death Rally (classic) last night. One of my favorite games from ~30 years ago. It holds up, still a fun game.

Although the experience is a little different playing it on a 38" widescreen vs a 14" Adi ProVista 😬

4 weeks ago 0 0 0 0

We live in a modern art installation, at least that part's kinda cool

1 month ago 0 0 0 0
Preview
GitHub - Tinche/pytest-typing Contribute to Tinche/pytest-typing development by creating an account on GitHub.

Feeling cute, might publish later: github.com/Tinche/pytes...

1 month ago 0 0 0 0

You're right in that, theoretically, we can bend the type system any way we want.

My argument boils down to "if it's hard to explain, it's a bad idea". 'Here's a function that actually defines a class that's actually half-protocol' falls in that category, to me.

1 month ago 1 0 1 0

Agreed. The proposal prominently mentions structural typing but produces classes (which are nominal types). So typing-wise they'd probably have to behave somewhat like protocols, making them difficult to explain and teach. Right? Whereas typeddicts are already structural, but can't be (yet) frozen.

1 month ago 0 1 1 0

Literate testing is great, yes, but mostly I got bored of writing my typing tests in yaml files

The first version will support Mypy and ty ;)

1 month ago 1 0 0 0

@alexwaygood.bsky.social I liked your "typing tests in markdown code blocks" idea so much I stole it into a pytest plugin I'm planning to publish soon. That's ok with you folks right? I mention that it was inspired by you in the README.

1 month ago 1 0 1 0

Reading the latest extended incident report from Honeycomb makes me want to never run Kafka in production.

Which is an approach that's served me fine up to this point in my career, why not in the future as well? 🤞

1 month ago 0 0 0 0

Ghostty is written in Zig and it really shows, unfortunately

1 month ago 0 0 0 0
Advertisement

The Steering Council has formally accepted the TypeForm PEP. This is big news for cattrs but I'll hold off on implementing it until Mypy enables support for it without a flag.

2 months ago 0 1 0 0

yeah reflinking is probably the way to go here

2 months ago 1 0 0 0

Apparently on Linux uv defaults to hardlinking files from its cache when installing stuff into a virtualenv. That kind of sucks for people who sometimes edit third-party libraries to stick prints or breakpoints in there, like me.

2 months ago 3 0 3 0

Touche

2 months ago 0 0 0 0

Yes from a practical perspective (sure, it needs to import the code) but also from a logical perspective.

If you have pytest unit tests, pytest is a (dev) dependency of your project. If you format your files, Black or ruff is a dependency of your project. Same for Mypy. They should all be listed.

2 months ago 1 0 0 0

It has been [0] days since some monkeypatching has caused me to waste over an hour of debugging.

2 months ago 1 0 0 0

Since upgrading my OnePlus Open to Android 16, my home screen reset itself to blank twice.

Super #firstworldproblems, but it feels like someone rearranged the furniture in my house. Trying to reconstruct app position by muscle memory right now (again).

2 months ago 2 1 0 0

They wanted to simplify by not having exceptions and just returning errors from functions but got bored before they added sum types

So instead functions return tuples with errors and exceptions still exist but they're called panics

2 months ago 0 0 0 0

Doing chess lessons on Duolingo.

"Pin my queen with your bishop" sounds way more fun out of context

2 months ago 1 0 0 0

In your particular case, the type system doesn't model hashability. Lists are mutable and so not hashable.

Tuples are immutable, but only hashable if their contents are hashable. So if you try hashing something like `([],)`, you'll again get an error.

3 months ago 1 0 0 0
Advertisement

You don't have to go that far to get an exception to be raised in properly typed code:

```py
x = [1]
x[2]
```

The point of type checking isn't to ensure no exceptions, but to apply an additional, stricter model of execution to your code. This model sometimes includes exceptions.

3 months ago 1 0 1 0

Ya

3 months ago 1 0 1 0

The law

3 months ago 0 0 1 0

This might get really thorny, yeah. Let's say Claude knows a ton of Textual (which is true in my experience) having ingested a ton of Textual, is Claude a derived work of Textual? If Claude spits out a tool written in Textual for me, is that tool a derived work of Textual?

🤷🤷🤷

3 months ago 0 0 1 0

You find a project you like, but you can't use it due to the license (maybe it's GPL or AGPL). What's stopping you feeding it to an agent and having it port it to another language, and then potentially back again?

Bam, no more copyright.

This is what AI companies are doing in the first place, yes?

3 months ago 0 0 1 0

Serious question: what happens to copyright when this is done? I'm guessing you lose any claims to the new code?

3 months ago 0 0 2 0
a guy wearing a “i ❤️ ny” cap reading the qur’an in one hand and the communist manifesto in the other

a guy wearing a “i ❤️ ny” cap reading the qur’an in one hand and the communist manifesto in the other

day one in mamdani’s new york

3 months ago 21256 3690 264 158

Yeah wtf? War is war, get him

3 months ago 0 0 0 0

Dipshit Derek king of season 5

3 months ago 0 0 0 0
Advertisement

This is no longer the case since AI can do all of this for you, zero effort required. In fact it's much worse since the sycophantism inherent to the process will actually push this further in the wrong direction.

3 months ago 0 0 1 0