Merge pull request #87 from lukasValentin/dev

Ignore no-data values when scaling band data and allow band selection by color name for Sentinel-2

eodal/core/band.py

@@ -1370,11 +1370,13 @@ def get_meta(self, driver: Optional[str] = "gTiff", **kwargs) -> Dict[str, Any]:
             name of the ``rasterio`` driver. `gTiff` (GeoTiff) by default
         :param kwargs:
             additional keyword arguments to append to metadata dictionary
+            or to overwrite defaults such as the "compress" attribute.
         :returns:
             ``rasterio`` compatible metadata dictionary to be used for
             writing new raster datasets
         """
         meta = {}
+        # set defaults
         meta["height"] = self.nrows
         meta["width"] = self.ncols
         meta["crs"] = self.crs
@@ -1387,6 +1389,8 @@ def get_meta(self, driver: Optional[str] = "gTiff", **kwargs) -> Dict[str, Any]:
         # "compress" as suggested here:
         # https://github.com/rasterio/rasterio/discussions/2933#discussioncomment-7208578
         meta["compress"] = "DEFLATE"
+
+        # defaults can be overwritten using custom kwargs
         meta.update(kwargs)
 
         return meta
@@ -1517,7 +1521,7 @@ def hist(
 
     def plot(
         self,
-        colormap: Optional[str] = "gray",
+        colormap: Optional[str] = "viridis",
         discrete_values: Optional[bool] = False,
         user_defined_colors: Optional[ListedColormap] = None,
         user_defined_ticks: Optional[List[Union[str, int, float]]] = None,
@@ -1532,7 +1536,7 @@ def plot(
 
         :param colormap:
             String identifying one of matplotlib's colormaps.
-            The default will plot the band in gray values.
+            The default will plot the band using the viridis colormap.
         :param discrete_values:
             if True (Default) assumes that the band has continuous values
             (i.e., ordinary spectral data). If False assumes that the
@@ -2201,21 +2205,28 @@ def reduce(
     def scale_data(
         self,
         inplace: Optional[bool] = False,
-        pixel_values_to_ignore: Optional[List[Union[int, float]]] = None,
+        pixel_values_to_ignore: Optional[List[int | float]] = [],
     ):
         """
         Applies scale and offset factors to the data.
 
+        .. versionadded:: 0.2.3
+            No-data values are ignored when applying scale and offset.
+
         :param inplace:
             if False (default) returns a copy of the ``Band`` instance
             with the changes applied. If True overwrites the values
             in the current instance.
         :param pixel_values_to_ignore:
-            optional list of pixel values (e.g., nodata values) to ignore,
-            i.e., where scaling has no effect
+            optional list of pixel values to ignore, i.e., where scaling
+            has no effect. From version 0.2.3 onwards, no-data values
+            are *always* ignored.
         :returns:
             ``Band`` instance if `inplace` is False, None instead.
         """
+        # add no-data to the `pixel_values_to_ignore` list
+        pixel_values_to_ignore.append(self.nodata)
+
         scale, offset = self.scale, self.offset
         if self.is_masked_array:
             if pixel_values_to_ignore is None:

eodal/core/raster.py

@@ -660,7 +660,7 @@ def _bands_from_selection(
         Selects bands in a multi-band raster dataset based on a custom
         selection of band indices or band names.
 
-        .. versionadd:: 0.2.0
+        .. versionadded:: 0.2.0
             works also with a dictionary of hrefs returned from a
             STAC query
 

eodal/core/sensors/landsat.py

@@ -386,6 +386,7 @@ def from_usgs(
             RasterCollection containing the Landsat bands.
         """
         # check band selection and determine the platform and sensor
+        _band_selection = deepcopy(band_selection)
         band_df = cls._preprocess_band_selection(
             cls,
             in_dir=in_dir,

eodal/core/sensors/sentinel2.py

@@ -31,6 +31,7 @@
 import geopandas as gpd
 import rasterio as rio
 
+from copy import deepcopy
 from matplotlib.pyplot import Figure
 from matplotlib import colors
 from numbers import Number
@@ -157,7 +158,7 @@ def _get_band_files(
         if processing_level == ProcessingLevels.L1C:
             is_l2a = False
 
-        # check if SCL should e read (L2A)
+        # check if SCL should be read (L2A)
         if is_l2a and read_scl:
             scl_in_selection = "scl" in band_selection or "SCL" in band_selection
             if not scl_in_selection:
@@ -209,6 +210,20 @@ def _process_band_selection(
             for band in bands_to_exclude:
                 band_selection.remove(band)
 
+        # check if band or color names are passed
+        color_names = set(band_selection).issubset(s2_band_mapping.values())
+        band_names = set(band_selection).issubset(s2_band_mapping.keys())
+
+        if not color_names and not band_names:
+            raise BandNotFoundError(
+                f'Invalid selection of bands: {band_selection}')
+
+        # internally, we use band names only. So we have to map the color names
+        # back to their band names
+        if color_names:
+            band_selection = [
+                k for k, v in s2_band_mapping.items() if v in band_selection]
+
         # determine which spatial resolutions are selected and check processing level
         band_df_safe = self._get_band_files(
             in_dir=in_dir, band_selection=band_selection, read_scl=read_scl
@@ -224,7 +239,7 @@ def _process_band_selection(
                     if bands_not_found[0] == "SCL":
                         return band_df_safe
                 raise BandNotFoundError(
-                    f"Couldnot find bands {bands_not_found} " "provided in selection"
+                    f"Could not find bands {bands_not_found} " "provided in selection"
                 )
         return band_df_safe
 
@@ -280,8 +295,9 @@ def from_safe(
             `Sentinel2` instance with S2 bands loaded
         """
         # load 10 and 20 bands by default
+        _band_selection = deepcopy(band_selection)
         band_df_safe = cls._process_band_selection(
-            cls, in_dir=in_dir, band_selection=band_selection, read_scl=read_scl
+            cls, in_dir=in_dir, band_selection=_band_selection, read_scl=read_scl
         )
 
         # check the clipping extent of the raster with the lowest (coarsest) spatial

tests/core/test_sentinel2.py

@@ -12,7 +12,7 @@
 from eodal.config import get_settings
 from eodal.core.sensors import Sentinel2
 from eodal.core.band import Band
-from eodal.utils.exceptions import BandNotFoundError, InputError
+from eodal.utils.exceptions import BandNotFoundError
 
 settings = get_settings()
 settings.USE_STAC = False
@@ -251,7 +251,7 @@ def test_ignore_scl(datadir, get_s2_safe_l2a, get_polygons_2):
 
 
 def test_band_selections(datadir, get_s2_safe_l2a, get_polygons_2):
-    """testing invalid band selections"""
+    """testing valid and invalid band selections"""
 
     in_dir = get_s2_safe_l2a()
     in_file_aoi = get_polygons_2()
@@ -265,6 +265,20 @@ def test_band_selections(datadir, get_s2_safe_l2a, get_polygons_2):
             band_selection=['B02', 'B13']
         )
 
+    # test with color names instead of band names
+    band_selection = ['red', 'green', 'blue']
+    ds = Sentinel2.from_safe(
+        in_dir=in_dir,
+        band_selection=band_selection,
+        read_scl=False,
+        apply_scaling=False
+    )
+    assert ds.band_names == ['B02', 'B03', 'B04']
+    assert set(ds.band_aliases) == set(band_selection)
+    assert (ds['B02'] == ds['blue']).values.all()
+    assert (ds['B03'] == ds['green']).values.all()
+    assert (ds['B04'] == ds['red']).values.all()
+
 
 @pytest.mark.skip(reason='too heavy test for Github workflows')
 def test_read_from_safe_l2a(datadir, get_s2_safe_l2a):