add new minmax function

doc/api/index.rst

@@ -91,6 +91,7 @@ Utilities
 .. autosummary::
    :toctree: generated/
 
+    minmax
     maxabs
     variance_to_weights
     grid_to_table

verde/__init__.py

@@ -36,7 +36,7 @@
 from .scipygridder import Cubic, Linear, ScipyGridder
 from .spline import Spline, SplineCV
 from .trend import Trend
-from .utils import grid_to_table, make_xarray_grid, maxabs, variance_to_weights
+from .utils import grid_to_table, make_xarray_grid, maxabs, minmax, variance_to_weights
 from .vector import Vector, VectorSpline2D
 
 

verde/tests/test_utils.py

@@ -27,6 +27,7 @@
     maxabs,
     meshgrid_from_1d,
     meshgrid_to_1d,
+    minmax,
     parse_engine,
     partition_by_sum,
     variance_to_weights,
@@ -354,6 +355,89 @@ def test_check_ndim_easting_northing():
         get_ndim_horizontal_coords(easting, northing)
 
 
+def test_minmax_nans():
+    """
+    Test minmax handles nans correctly
+    """
+    assert tuple(map(float, minmax((-1, 100, 1, 2, np.nan)))) == (-1, 100)
+    assert tuple(map(float, minmax((np.nan, -3.2, -1, -2, 3.1)))) == (-3.2, 3.1)
+    assert np.all(np.isnan(minmax((np.nan, -3, -1, 3), nan=False)))
+
+
+def test_minmax_percentile():
+    """
+    Test minmax with percentile option
+    """
+    data = np.arange(0, 101)
+
+    # test generic functionality
+    result = tuple(map(float, minmax(data, min_percentile=0, max_percentile=100)))
+    assert result == (0, 100)
+    result = tuple(map(float, minmax(data, min_percentile=0.0, max_percentile=100.0)))
+    assert result == (0, 100)
+    result = tuple(map(float, minmax(data, min_percentile=10, max_percentile=90)))
+    assert pytest.approx(result, 0.1) == (10, 90)
+    result = tuple(map(float, minmax(data, min_percentile=10.0, max_percentile=90.0)))
+    assert pytest.approx(result, 0.1) == (10, 90)
+
+    # test with nans
+    data_with_nans = np.append(data, np.nan)
+    result = tuple(
+        map(float, minmax(data_with_nans, min_percentile=0, max_percentile=100))
+    )
+    assert result == (0, 100)
+    result = tuple(
+        map(float, minmax(data_with_nans, min_percentile=10, max_percentile=90))
+    )
+    assert pytest.approx(result, 0.1) == (10, 90)
+    result = tuple(
+        map(
+            float,
+            minmax(data_with_nans, min_percentile=10, max_percentile=90, nan=True),
+        )
+    )
+    assert pytest.approx(result, 0.1) == (10, 90)
+    result = minmax(data_with_nans, min_percentile=10, max_percentile=90, nan=False)
+    assert np.all(np.isnan(result))
+
+    # test with varying array sizes
+    result = tuple(
+        map(
+            float,
+            minmax(
+                [0, 1, 2, 3, 4], [[-2, 2], [0, 5]], min_percentile=0, max_percentile=100
+            ),
+        )
+    )
+    assert result == (-2, 5)
+    result = tuple(
+        map(
+            float,
+            minmax(
+                [0, 1, 2, 3, 4], [[-2, 2], [0, 5]], min_percentile=1, max_percentile=99
+            ),
+        )
+    )
+    assert pytest.approx(result, 0.1) == (-1.84, 4.92)
+
+    # test invalid percentile types
+    with pytest.raises(TypeError):
+        minmax(data, min_percentile=None)
+    with pytest.raises(TypeError):
+        minmax(data, max_percentile=[90])
+
+    # test invalid percentile values
+    msg = "'min_percentile'"
+    with pytest.raises(ValueError, match=msg):
+        minmax(data, min_percentile=99, max_percentile=90)
+    msg = "Invalid value for 'min_percentile'"
+    with pytest.raises(ValueError, match=msg):
+        minmax(data, min_percentile=-10)
+    msg = "Invalid value for 'max_percentile'"
+    with pytest.raises(ValueError, match=msg):
+        minmax(data, max_percentile=110)
+
+
 def test_maxabs_nans():
     """
     Test maxabs handles nans correctly

verde/utils.py

@@ -263,6 +263,113 @@ def maxabs(*args, nan=True, percentile=100):
         return nppercentile(combined_array, percentile)
 
 
+def minmax(*args, nan=True, min_percentile=0, max_percentile=100):
+    """
+    Calculate the minimum and maximum values of the given array(s).
+
+    Use this to set the limits of your colorbars for non-diverging data.
+
+    Parameters
+    ----------
+    args
+        One or more arrays. If more than one are given, a minimum and maximum
+        will be calculated across all arrays.
+    nan : bool, optional
+        If True, will use the ``nan`` version of numpy functions to ignore
+        NaNs.
+    min_percentile : float
+        Return the supplied percentile (0 to 100) instead of the minimum value
+        of the arrays. Defaults to 0, giving the minimum value.
+    max_percentile : float
+        Return the supplied percentile (0 to 100) instead of the maximum value
+        of the arrays. Defaults to 100, giving the maximum value.
+
+    Returns
+    -------
+    min, max : float
+        The minimum and maximum (or percentile) values across all arrays.
+
+    Examples
+    --------
+
+    >>> result = minmax((1, -10, 25, 2, 3))
+    >>> tuple(map(float, result))
+    (-10.0, 25.0)
+    >>> result = minmax(
+    ...     (1, -10.5, 25, 2), (0.1, 100, -500), (-200, -300, -0.1, -499)
+    ... )
+    >>> tuple(map(float, result))
+    (-500.0, 100.0)
+
+    If the array contains NaNs, we'll use the ``nan`` version of of the numpy
+    functions by default. You can turn this off through the *nan* argument.
+
+    >>> import numpy as np
+    >>> result = minmax((1, -10, 25, 2, 3, np.nan))
+    >>> tuple(map(float, result))
+    (-10.0, 25.0)
+    >>> result = minmax((1, -10, 25, 2, 3, np.nan), nan=False)
+    >>> tuple(map(float, result))
+    (nan, nan)
+
+    If a more robust statistic is desired, you can use ``min_percentile`` and
+    or ``max_percentile`` to get the values at given percentiles instead of
+    the minimum and maximum.
+
+    >>> import numpy as np
+    >>> result = minmax(
+    ...     (1, -10, 25, 2, 3), min_percentile=2, max_percentile=98
+    ... )
+    >>> tuple(map(float, result))
+    (-9.12, 23.24)
+    >>> result = minmax(
+    ...     (1, -10, 25, 2, 3), min_percentile=0, max_percentile=100
+    ... )
+    >>> tuple(map(float, result))
+    (-10.0, 25.0)
+
+    """
+    # ensure percentiles are numeric
+    min_percentile = float(min_percentile)
+    max_percentile = float(max_percentile)
+
+    arrays = [np.atleast_1d(i) for i in args]
+
+    # checks
+    if not min_percentile <= max_percentile:
+        raise ValueError(
+            f"'min_percentile' ({min_percentile}) must not be larger than 'max_percentile' ({max_percentile})."
+        )
+    if min_percentile < 0:
+        raise ValueError(
+            f"Invalid value for 'min_percentile' ({min_percentile}). Must be between 0 and 100."
+        )
+    if max_percentile > 100:
+        raise ValueError(
+            f"Invalid value for 'max_percentile' ({max_percentile}). Must be between 0 and 100."
+        )
+
+    # determine which functions to use
+    if nan:
+        npmin = np.nanmin
+        npmax = np.nanmax
+        nppercentile = np.nanpercentile
+    else:
+        npmin = np.min
+        npmax = np.max
+        nppercentile = np.percentile
+
+    if min_percentile == 0 and max_percentile == 0:
+        min_ = npmin([npmin(i) for i in arrays])
+        max_ = npmax([npmax(i) for i in arrays])
+        return min_, max_
+
+    # concatenate values of all arrays
+    combined_array = np.concatenate([a.ravel() for a in arrays])
+    min_, max_ = nppercentile(combined_array, [min_percentile, max_percentile])
+    return min_, max_
+
+
 def make_xarray_grid(
     coordinates,
     data,