Fix shading (#123)

* Updated environment, quicker setup

* Fixed bug + added docs for shading plot

* fix formatting

* Fix bug for spread defaults

* fix formatting

* Update whats-new.rst

docs/environment.yml

@@ -1,4 +1,6 @@
 name: xarrayutils-docs
+channels:
+  - conda-forge
 dependencies:
   - python=3.8
   - numpydoc
@@ -13,16 +15,16 @@ dependencies:
   - dask
   - numpy
   - xarray
+  - scipy
+  - nc-time-axis
+  - astropy
+  - intake-esm
+  - gcsfs
+  - nbsphinx
+  - jupyter_client
+  - sphinxcontrib-srclinks
+  - pandoc
+  - recommonmark
   - pip
   - pip:
-    - scipy
-    - nc-time-axis
-    - astropy
-    - intake-esm
-    - gcsfs
-    - nbsphinx
-    - jupyter_client
-    - sphinx_pangeo_theme
-    - sphinxcontrib-srclinks
-    - pandoc
-    - recommonmark
+     - sphinx_pangeo_theme

docs/whats-new.rst

@@ -10,6 +10,10 @@ New Features
 ~~~~~~~~~~~~
 - `sing_agreement` now supports the option to not count nans along the given dimension (:pull:`118`). By `Julius Busecke <https://github.com/jbusecke>`_
 
+Bugfixes
+~~~~~~~~
+- Fixed bug in `shaded_line_plot`, where std bounds were displayed incorrectly (:issue:`74`, :pull:`123`). By `Julius Busecke <https://github.com/jbusecke>`_
+
 .. _whats-new.1.0.0:
 
 v1.0.0 (2021/6/22)

xarrayutils/plotting.py

@@ -1,3 +1,4 @@
+from tkinter import Y
 import numpy as np
 import xarray as xr
 import warnings
@@ -233,7 +234,7 @@ def shaded_line_plot(
     dim,
     ax=None,
     horizontal=True,
-    spreads=[1, 3],
+    spreads=None,
     alphas=[0.25, 0.4],
     spread_style="std",
     line_kwargs=dict(),
@@ -317,8 +318,12 @@ def shaded_line_plot(
     # define the line plot values
     if spread_style == "std":
         y = da.mean(dim)
+        if spreads is None:
+            spreads = [1, 3]
     elif spread_style in ["quantile", "percentile"]:
         y = da.quantile(0.5, dim)
+        if spreads is None:
+            spreads = [0.5, 0.8]
     else:
         raise ValueError(
             f"Got unknown option ['{spread_style}'] for  `spread_style`. Supported options are : ['std', 'quantile']"
@@ -340,16 +345,19 @@ def shaded_line_plot(
     fill_defaults.update(fill_kwargs)
     ff = []
 
-    for spread, alpha in zip(
-        (spreads), (alphas)
-    ):  # np.flip(this ensures that the shadings are drawn from outer to inner otherwise they blend too much into each other
+    spreads = list(np.flip(spreads))
+    alphas = list(np.flip(alphas))
+    # np.flip(this ensures that the shadings are drawn from outer to inner otherwise they blend too much into each other
+
+    for spread, alpha in zip(spreads, alphas):
         f_kwargs = {k: v for k, v in fill_defaults.items()}
         f_kwargs["alpha"] = alpha
 
         if spread_style == "std":
             y_std = da.std(dim)  # i could probably precompute that.
-            y_lower = y - (y_std / (2 * spread))
-            y_upper = y + (y_std / (2 * spread))
+            y_spread = y_std * spread
+            y_lower = y - (y_spread / 2)
+            y_upper = y + (y_spread / 2)
 
         elif spread_style in ["quantile", "percentile"]:
             y_lower = da.quantile(0.5 - (spread / 2), dim)

xarrayutils/test/test_plotting.py

@@ -39,10 +39,11 @@ def test_shaded_line_plot(dim, horizontal, spreads, alphas):
     ll, ff = shaded_line_plot(
         da, dim, spreads=spreads, alphas=alphas, horizontal=horizontal
     )
+
     assert isinstance(ll[0], matplotlib.lines.Line2D)
     assert len(ff) == min(len(spreads), len(alphas))
 
-    for f, a in zip(ff, alphas):
+    for f, a in zip(ff, np.flip(alphas)):
         assert isinstance(f, matplotlib.collections.PolyCollection)
         assert f.get_alpha() == a
 

xarrayutils/utils.py

@@ -73,7 +73,7 @@ def xr_linregress(x, y, dim="time"):
     cov = ((x - xmean) * (y - ymean)).sum(dim) / (n)
     cor = cov / (xstd * ystd)
 
-    slope = cov / (xstd ** 2)
+    slope = cov / (xstd**2)
     intercept = ymean - xmean * slope
 
     df = n - 2

xarrayutils/xmitgcm_utils.py

@@ -68,7 +68,7 @@ def laplacian(grid, data):
 
 def gradient_sq_amplitude(grid, data):
     gradx, grady = gradient(grid, data, interpolate=True)
-    return gradx ** 2 + grady ** 2
+    return gradx**2 + grady**2
 
 
 # Silly functions