Eddy tracking and python3.12

[ctroupin]

Hello Antoine,

I've recently restarted to use the package and surprisingly with python3.12 (in a virtual environment), I don't get the expected results: whatever domain I select, the code only detects one eddy.

For instance with the data provided with the code

g = RegularGridDataset(
    data.get_demo_path("dt_med_allsat_phy_l4_20160515_20190101.nc"),
    "longitude",
    "latitude",
)
g.add_uv("adt")
a, c = g.eddy_identification("adt", "u", "v", datetime.datetime(2016, 5, 20), 0.001, shape_error=85, pixel_limit=(1, 50000))

(note that I used a large value for shape_error and a wide interval for pixel_limit so I'm sure I find eddies) and this is what I got (the eddy is visible in green):

Then I re-installed everything from scratch, again in virtual environment, but with python3.10, and now the results are OK:

Question: is there a specific reason why it would fail with python3.12?
(and sorry if it's in the doc and I didn't see it)

[AntSimi]

It's a known problem #217 that i didn't take time to solve, matplotlib >= 3.8 have a new object for contour that are not correctly managed by py eddy tracker.

[AntSimi]

Recently i set matplotlib version in requirements.txt to avoid this problem, before to correctly manage the upgrade of matplotlib

[ctroupin]

ok thanks, if it's known then let's close, I'll be using another Python (and matplotlib) versions.