roxmltree VS polars

This is exactly what I needed when implementing xml-mut :D I have used roxmltree instead and manipulated text directly. will try to rewrite it using Xot.

In regards to the benchmarks, It makes sense to measure serializing/deserializing for parser crates. but since we are talking about dom implementations, metrics like traversal/iteration speed or insert/modification performance would be useful. a good example is roxmltree crate (readonly xml dom) which benches traversal/iteration performance and shows that by only focusing on readonly usecases, it gains substantial performance gains.

For xml parsing, I find https://github.com/RazrFalcon/roxmltree as a really good crate. It’s fast, light, and well documented/maintained. I have so much respect for the maintainer’s approach to merging PRs and the way they consider what’s important for the crate

I understand. But I've also wrote enough parsers and performance sensitive code in Rust (ttf-parser, tiny-skia, roxmltree). And in my experience, unsafe is not needed in 99% of the cases. Even something as performance sensitive as tiny-skia is unsafe-free (with some nuances).

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: