Jiří Moravec

Computational Biology, Phylogenetics, Stats

Senior R&D Data Scientist at @Biolumic

[Jiří Moravec @jcmoravec]

@DanielW_Kiwi I always use slippers. Can't touch these beasts with my bare hands.

[Jiří Moravec @jcmoravec]

@tehjizza @Chris_Segroves @itchy_wheezy This is wrong. PCA will produce you orthogonal dimensions sorted according to their variability. But there is no guarantee that the most variable dimension is predictive.

[Jiří Moravec @jcmoravec]

Why the hell LIM in PN cost 500? In ChCh it is just 300!

[Jiří Moravec @jcmoravec]

I am getting lost in all this. Is this anti-lynching or pro-lynching messaging?

[Diego A. Ballestero @InsurgentMemory]

793 people who looked at a man who conducted a sustained harassment campaign against Jason Arday who is now dead — and said: "Yes, defend him."
We will never forget these names.
We will expose them. Not to silence them — but to name them.
We will not let the world forget.

[Jiří Moravec @jcmoravec]

@DanielW_Kiwi Same when working with data, the best time to solve data issues is in ingestion. I am often able to remove so many lines of confusing special case handling that needs to happen in multiple places by fixing things right there in place when loading data.

[Jiří Moravec @jcmoravec]

@pp0196 A fairly common reason for selling house in NZ is because an old person is moving to retirement home or they passed away. Usually people clean the house, sometimes they don't and then you are glad that you took the spare socks with you.

[Jiří Moravec @jcmoravec]

When people are organizing pre-sale house viewing, I would consider it normal to clean, or at least remove the dead people smell.

[Jiří Moravec @jcmoravec]

@MightyMogomra @DanielW_Kiwi @meowbooksj The Use of Weapons have such weird ending it soured the story to me. People don't like Consider Phlegmas, but I vastly prefer it to UoW. Player of Games is also rad.

Need to find time to read the rest.

[Jiří Moravec @jcmoravec]

Ok, this is really brainrot.

Who came with this?

https://t.co/oTCIDbX33c

Why breaking one of the core zen of Python (and programming in general) "Better explicit than implicit"?

Why linting out explicit and clear intent like that?

[Jiří Moravec @jcmoravec]

Just updated to Ruff v0.16, which enabled bunch of lints, like I001 (organising imports) by default.

This import is terrible, it breaks logical grouping of imports in favour sort-of of alphabetic sorting.

Imagine running tool like that automatically to "clean up" your code.

[Jiří Moravec @jcmoravec]

@TimTeaFan @brodriguesco Good.

[Jiří Moravec @jcmoravec]

@RitchieVink The work I need to do in Polars is enough graphical noise for me that I shove it into a function.

[Jiří Moravec @jcmoravec]

Why are simple operations in Polars so complicated?

Like common operation "Check if these columns have unknown values". You need to run pipe of 7 different functions and return weird data structure.

In R, you just sapply(df, anyNA) and get named vector.

[Jiří Moravec @jcmoravec]

@RitchieVink In R, the only intricate thing is sapply.

In Polars, I need to know that has_nulls() is method for series, not df, then I need to get only fist row into dict, which I need to then select with list comprehension (or set). More work across boundaries.

[Jiří Moravec @jcmoravec]

@RitchieVink And I think I know why the need to call for row is there. Polars try to be "always return DF". Which is nice. But in situations like "Get me columns that have NAs" it gets unnecessary complicated.

[Jiří Moravec @jcmoravec]

@RitchieVink Thats pipe. And you forgot .row(0, named=True)

[Jiří Moravec @jcmoravec]

@iskander In either case, both are building on something that had to be bolted on top of Python, and are not de-facto part of Python. Compare to R where you got vector as a basic primitive, DF is just list of vectors, so any operation (like sum, mean, etc.) can be efficient. Its natural.

[Jiří Moravec @jcmoravec]

@iskander I don't like Pandas because it is very leaky abstraction on top of Numpy. Some API returns view or modify on place, other make clones, and they still didn't figured out type-specific NAs, something S got 50 years ago.