Unexpected behavior for vertical datums that use 'depth' below a reference surface rather than 'height'

[Erik-Geo]

It is possible to convert 'surface' and 'end' in the header to depth below a reference level when changing to vertical datums like LAT NL depth (EPSG 9287) or MSL NL depth (EPSG 9288). This causes values in the end column to become higher than the surface values.

This breaks behaviour of (at least) the select_by_length method which now calculates negative lengths for all objects and also gives many unwanted validation errors if running through the header validation again.

dino = geost.read_borehole_table(path_to_dinodump)
dino.change_vertical_reference("LAT NL depth")
selection_result = dino.select_by_length(4)

selection_result will be empty!

Using negative lengths or changing back to NAP or any other vertical datum that uses 'height' rather than 'depth' are workarounds.

Suggestions:

• To fix select_by_length: (1) Use absolute values for difference between surface and end; (2) Convert user input to negative value if the vertical datum is of 'depth'-type
• For validation, pyproj.crs.crs.CRS has an axis_info property that includes information on Depth versus Height and positive up/down direction, which can potentially be used to select the appropriate validation schema.