Henrik Andersson

The error is caused by the [`GeometryFM2D.contains`](https://github.com/DHI/mikeio/blob/745768c0359dd80a664e89fcdd85da02e19f14b7/mikeio/spatial/_FM_geometry.py#L834-L865) method, which incorrectly identifies points as outside the domain. Here is a breakdown of the boundary segments, where some of the exterior segments...

If I replace the current implementation with shapely it seems to work as expected. The reason why this approach is not the default, was that is can be quite slow...

> I was thinking yesterday about it, as in, why not use shapely or geopandas or similar (OsGeo or any of those GIS python packages) that already have some fancy...

This problem is not only relevant for track extraction. It is a problem for point extraction as well in some cases when using IDW interpolation. Below is an example of...

> But then why don't we: > > * create a failing test (we can use the files I sent and the example you just showed) > * change to...

I will fix it in #521 which is closely related to this.

One line in a notebook is failing ```python cmp.where(cmp.data.windsector=="W").plot.timeseries(); ``` I can't see that it is related to the changes in this PR, except for the new version of xarray.

Seems like there is some kind of agreement in the scientific Python community to support three year old Python versions https://scientific-python.org/specs/spec-0000/ Which would exclude 3.9 released in October 2020.

TODO: - [x] Re-enable legacy persistence method using zip file. - [x] convert persisted datasets to use groups.

Quick suggestion before 🎅🎄 I think you want to apply the unit conversion to all variables of a specfied **quantity**. Example of syntax: ```python from typing import Callable import modelskill...