Test version

requirements.txt

@@ -1,4 +1,4 @@
-matplotlib
+matplotlib<3.5
 netCDF4
 numba>=0.53
 numpy<1.21

src/py_eddy_tracker/dataset/grid.py

@@ -680,6 +680,7 @@ def eddy_identification(
             )
             z_min, z_max = z_min_p, z_max_p
 
+        logger.debug("Levels from %f to %f", z_min, z_max)
         levels = arange(z_min - z_min % step, z_max - z_max % step + 2 * step, step)
 
         # Get x and y values
@@ -1404,7 +1405,8 @@ def convolve_filter_with_dynamic_kernel(
             tmp_matrix = ma.zeros((2 * d_lon + data.shape[0], k_shape[1]))
             tmp_matrix.mask = ones(tmp_matrix.shape, dtype=bool)
             # Slice to apply on input data
-            sl_lat_data = slice(max(0, i - d_lat), min(i + d_lat, data.shape[1]))
+            # +1 for upper bound, to take in acount this column
+            sl_lat_data = slice(max(0, i - d_lat), min(i + d_lat + 1, data.shape[1]))
             # slice to apply on temporary matrix to store input data
             sl_lat_in = slice(
                 d_lat - (i - sl_lat_data.start), d_lat + (sl_lat_data.stop - i)

src/py_eddy_tracker/observations/network.py

@@ -6,6 +6,8 @@
 import time
 from glob import glob
 
+import netCDF4
+import zarr
 from numba import njit
 from numpy import (
     arange,
@@ -23,9 +25,6 @@
     zeros,
 )
 
-import netCDF4
-import zarr
-
 from ..dataset.grid import GridCollection
 from ..generic import build_index, wrap_longitude
 from ..poly import bbox_intersection, vertice_overlap
@@ -680,13 +679,7 @@ def display_timeline(
         """
         self.only_one_network()
         j = 0
-        line_kw = dict(
-            ls="-",
-            marker="+",
-            markersize=6,
-            zorder=1,
-            lw=3,
-        )
+        line_kw = dict(ls="-", marker="+", markersize=6, zorder=1, lw=3,)
         line_kw.update(kwargs)
         mappables = dict(lines=list())
 
@@ -919,10 +912,7 @@ def event_map(self, ax, **kwargs):
         """Add the merging and splitting events to a map"""
         j = 0
         mappables = dict()
-        symbol_kw = dict(
-            markersize=10,
-            color="k",
-        )
+        symbol_kw = dict(markersize=10, color="k",)
         symbol_kw.update(kwargs)
         symbol_kw_split = symbol_kw.copy()
         symbol_kw_split["markersize"] += 4
@@ -951,13 +941,7 @@ def event_map(self, ax, **kwargs):
         return mappables
 
     def scatter(
-        self,
-        ax,
-        name="time",
-        factor=1,
-        ref=None,
-        edgecolor_cycle=None,
-        **kwargs,
+        self, ax, name="time", factor=1, ref=None, edgecolor_cycle=None, **kwargs,
     ):
         """
         This function scatters the path of each network, with the merging and splitting events

tests/test_grid.py

@@ -1,21 +1,13 @@
 from matplotlib.path import Path
-from numpy import array, isnan, ma
+from numpy import arange, array, isnan, ma, nan, ones, zeros
 from pytest import approx
 
 from py_eddy_tracker.data import get_demo_path
 from py_eddy_tracker.dataset.grid import RegularGridDataset
 
 G = RegularGridDataset(get_demo_path("mask_1_60.nc"), "lon", "lat")
 X = 0.025
-contour = Path(
-    (
-        (-X, 0),
-        (X, 0),
-        (X, X),
-        (-X, X),
-        (-X, 0),
-    )
-)
+contour = Path(((-X, 0), (X, 0), (X, X), (-X, X), (-X, 0),))
 
 
 # contour
@@ -85,3 +77,36 @@ def test_interp():
     assert g.interp("z", x0, y0) == 1.5
     assert g.interp("z", x1, y1) == 2
     assert isnan(g.interp("z", x2, y2))
+
+
+def test_convolution():
+    """
+    Add some dummy check on convolution filter
+    """
+    # Fake grid
+    z = ma.array(
+        arange(12).reshape((-1, 1)) * arange(10).reshape((1, -1)),
+        mask=zeros((12, 10), dtype="bool"),
+        dtype="f4",
+    )
+    g = RegularGridDataset.with_array(
+        coordinates=("x", "y"),
+        datas=dict(z=z, x=arange(0, 6, 0.5), y=arange(0, 5, 0.5),),
+        centered=True,
+    )
+
+    def kernel_func(lat):
+        return ones((3, 3))
+
+    # After transpose we must get same result
+    d = g.convolve_filter_with_dynamic_kernel("z", kernel_func)
+    assert (d.T[:9, :9] == d[:9, :9]).all()
+    # We mask one value and check convolution result
+    z.mask[2, 2] = True
+    d = g.convolve_filter_with_dynamic_kernel("z", kernel_func)
+    assert d[1, 1] == z[:3, :3].sum() / 8
+    # Add nan and check only nearest value is contaminate
+    z[2, 2] = nan
+    d = g.convolve_filter_with_dynamic_kernel("z", kernel_func)
+    assert not isnan(d[0, 0])
+    assert isnan(d[1:4, 1:4]).all()