allsorts VS polars

But again you don't need to reinvent text layout from scratch (that's my discussion, BTW) if you use the platform's native text drawing facilities, like DirectWrite on Windows, or CoreText on macOS.

All these layers of Cows create an extremely complex lifetime tree and it's actually not possible to factor out the complex parsing steps into functions because they depend on a lot of stuff. (trying to return an OutlineBuilder from a function would dangling reference about 3 different temporaries, for example. Obtaining the temporaries is part of the parsing step)

machine learning, neural networks, image processing, cryptography (though it is getting better), font shaping/rendering (though it is getting better), CPU/software rendering (though it is getting better)

It can be used for the things you suggest but our driving use case is font parsing. The goal is to use it to generate the font parsing code in Allsorts, which is ultimately used in Prince.

Use allsorts or rustbuzz for text shaping

Loader -> Allsorts

It's something I'd like to improve on in future -- are there any pure-Rust libraries you'd recommend for shaping? Allsorts looks quite promising.

This is because 0.1 is in actuality the floating point value value 0.1000000000000000055511151231257827021181583404541015625, and thus 1 divided by it is ever so slightly smaller than 10. Nevertheless, fpround(1 / fpround(1 / 10)) = 10 exactly.

I found out about this recently because in Polars I defined a // b for floats to be (a / b).floor(), which does return 10 for this computation. Since Python's correctly-rounded division is rather expensive, I chose to stick to this (more context: https://github.com/pola-rs/polars/issues/14596#issuecomment-...).

https://github.com/pola-rs/polars/releases/tag/py-0.19.0

That turned out to be related to pola-rs/polars#11912, and this linked comment provided a deceptively simple solution - use PARSE_DECLTYPES when creating the connection: