Merge pull request #57 from daberer/master

Add error maps

src/qa4sm_reader/globals.py

@@ -6,6 +6,7 @@
 import warnings
 
 import cartopy.crs as ccrs
+import matplotlib.colors as cl
 import matplotlib.pyplot as plt
 
 # PLOT DEFAULT SETTINGS
@@ -62,11 +63,44 @@
     "frm_class": ['frm_class'],
 }
 
+# === calculation errors (pytesmo) === #TODO: import from pytesmo
+status = {
+    -1: 'Other error',
+    0: 'Success',
+    1: 'Not enough data',
+    2: 'Metric calculation failed',
+    3: 'Temporal matching failed',
+    4: 'No overlap for temporal match',
+    5: 'Scaling failed',
+    6: 'Unexpected validation error',
+    7: 'Missing GPI data',
+    8: 'Data reading failed'
+}
+
+# helper dict to replace some error codes and have merged categories
+# (e.g.: No overlap for temporal match -> Temporal matching failed)
+status_replace = {
+    4: 3,
+    7: 1,
+}
+
 # === colormaps used for plotting metrics ===
 # Colormaps can be set for classes of similar metrics or individually for metrics.
 # Any colormap name can be used, that works with matplotlib.pyplot.cm.get_cmap('colormap')
 # more on colormaps: https://matplotlib.org/users/colormaps.html | https://morphocode.com/the-use-of-color-in-maps/
 
+
+def get_status_colors():
+    # function to get custom cmap for calculation errors
+    # limited to 14 different error entries to produce distinct colors
+    cmap = plt.cm.get_cmap('Set3', len(status) - 2)
+    colors = [cmap(i) for i in range(cmap.N)]
+    colors.insert(0, (0, 0.66666667, 0.89019608, 1.0))
+    colors.insert(0, (0.45882353, 0.08235294, 0.11764706, 1.0))
+    cmap = cl.ListedColormap(colors=colors)
+    return cmap
+
+
 _cclasses = {
     'div_better': plt.cm.get_cmap(
         'RdYlBu'
@@ -81,6 +115,8 @@
     ),  # sequential: increasing value bad (p_R, p_rho, rmsd, ubRMSD, RSS)
     'seq_better': plt.cm.get_cmap(
         'YlGn'),  # sequential: increasing value good (n_obs, STDerr)
+    'qua_neutr':
+    get_status_colors(),  # qualitative category with 2 forced colors
 }
 
 _colormaps = {  # from /qa4sm/validator/validation/graphics.py
@@ -102,6 +138,7 @@
     'snr': _cclasses['div_better'],
     'err_std': _cclasses['seq_worse'],
     'beta': _cclasses['div_neutr'],
+    'status': _cclasses['qua_neutr'],
 }
 
 # Colorbars for difference plots
@@ -132,9 +169,9 @@
     0: ['n_obs'],
     2: [
         'R', 'p_R', 'rho', 'p_rho', 'RMSD', 'BIAS', 'urmsd', 'mse', 'mse_corr',
-        'mse_bias', 'mse_var', 'RSS', 'tau', 'p_tau'
+        'mse_bias', 'mse_var', 'RSS', 'tau', 'p_tau', 'status'
     ],
-    3: ['snr', 'err_std', 'beta']
+    3: ['snr', 'err_std', 'beta', 'status']
 }
 
 # === variable template ===
@@ -196,6 +233,7 @@
     'snr': [None, None],
     'err_std': [None, None],
     'beta': [None, None],
+    'status': [-1, len(status)-2],
 }
 # mask values out of range
 _metric_mask_range = {
@@ -226,6 +264,7 @@
     'snr': ' in dB',
     'err_std': ' in {}',
     'beta': ' in {}',
+    'status': '',
 }
 
 
@@ -291,6 +330,7 @@ def get_metric_units(dataset, raise_error=False):
     'snr': 'Signal-to-noise ratio',
     'err_std': 'Error standard deviation',
     'beta': 'TC scaling coefficient',
+    'status': 'Validation errors'
 }
 
 # BACKUPS

src/qa4sm_reader/handlers.py

@@ -8,6 +8,7 @@
 
 class MixinVarmeta:
     """Mixin class to provide functions that are common to the MetricVariable and ConfidenceInterval subclasses"""
+
     @property
     def pretty_name(self):
         """Create a nice name for the variable"""
@@ -63,8 +64,13 @@ def get_varmeta(self) -> (tuple, tuple, tuple):
             ref_ds = self.Datasets.dataset_metadata(self.parts['ref_id'])
             mds = self.Datasets.dataset_metadata(self.parts['sat_id0'])
             dss = None
+
+            # if metric is status and globals.metric_groups is 3, add third dataset
+            if self.g == 3 and self.metric == 'status':
+                dss = self.Datasets.dataset_metadata(self.parts['sat_id1'])
+
             # if metric is TC, add third dataset
-            if self.g == 3:
+            elif self.g == 3:
                 mds = self.Datasets.dataset_metadata(self.parts['mds_id'])
                 dss = self.Datasets.dataset_metadata(self.parts['sat_id1'])
                 if dss == mds:
@@ -84,6 +90,7 @@ class QA4SMDatasets():
     1-based and 0-based index number of the datasets, respectively. For newer validations, these are always
     the same
     """
+
     def __init__(self, global_attrs):
         """
         Parameters
@@ -310,6 +317,7 @@ def dataset_metadata(self, id: int, element: str or list = None) -> tuple:
 
 class QA4SMVariable():
     """Super class for all variable types in the validations (MetricVariable, CI and Metadata)"""
+
     def __init__(self, varname, global_attrs, values=None):
         """
         Validation results for a validation metric and a combination of datasets.
@@ -372,6 +380,25 @@ def is_CI(self):
     def ismetric(self) -> bool:
         return self.g is not None
 
+    def _parse_wrap(self, pattern, g):
+        """Wrapper function that handles case of metric 'status' that occurs
+        in two globals.metric_groups (2,3). This is because a status array
+        can be the result of a validation between two or three datasets (tc)
+        """
+        # ignore this case - (status is also in globals.metric_groups 2 but
+        # should be treated as group 3)
+        if self.varname.startswith('status') and (self.varname.count('_and_')
+                                                  == 2) and g == 2:
+            return None
+        # parse self.varname when three datasets
+        elif self.varname.startswith('status') and (self.varname.count('_and_')
+                                                    == 2) and g == 3:
+            template = globals.var_name_ds_sep[3]
+            return parse(
+                '{}{}'.format(globals.var_name_metric_sep[2], template),
+                self.varname)
+        return parse(pattern, self.varname)
+
     def _parse_varname(self) -> (str, int, dict):
         """
         Parse the name to get the metric, group and variable data
@@ -393,7 +420,8 @@ def _parse_varname(self) -> (str, int, dict):
                 template = ''
             pattern = '{}{}'.format(globals.var_name_metric_sep[g], template)
             # parse infromation from pattern and name
-            parts = parse(pattern, self.varname)
+
+            parts = self._parse_wrap(pattern, g)
 
             if parts is not None and parts['metric'] in globals.metric_groups[
                     g]:
@@ -411,6 +439,7 @@ def _parse_varname(self) -> (str, int, dict):
 
 class MetricVariable(QA4SMVariable, MixinVarmeta):
     """Class that describes a metric variable, i.e. the metric for a specific set of Datasets"""
+
     def __init__(self, varname, global_attrs, values=None):
         super().__init__(varname, global_attrs, values)
 
@@ -420,6 +449,7 @@ def __init__(self, varname, global_attrs, values=None):
 
 class ConfidenceInterval(QA4SMVariable, MixinVarmeta):
     """Class for a MetricVariable representing confidence intervals"""
+
     def __init__(self, varname, global_attrs, values=None):
         super().__init__(varname, global_attrs, values)
 
@@ -431,6 +461,7 @@ def __init__(self, varname, global_attrs, values=None):
 
 class Metadata(QA4SMVariable):
     """Class for a MetricVariable representing metadata (only with ISMN as reference)"""
+
     def __init__(self, varname, global_attrs, values=None):
         super().__init__(varname, global_attrs, values)
 
@@ -451,6 +482,7 @@ def pretty_name(self) -> str:
 
 class QA4SMMetric():
     """Class for validation metric"""
+
     def __init__(self, name, variables_list=None):
 
         self.name = name
@@ -476,7 +508,8 @@ def _get_attribute(self, attr: str):
         """
         for n, Var in enumerate(self.variables):
             value = getattr(Var, attr)
-            if n != 0:
+            # special case for "status" attribute (self.g can be 2 or 3)
+            if n != 0 and not Var.varname.startswith('status'):
                 assert value == previous, "The attribute {} is not equal in all variables".format(
                     attr)
             previous = value

src/qa4sm_reader/plot_all.py

@@ -93,6 +93,10 @@ def plot_all(filepath: str,
                     warnings.warn(
                         "Too few points are available to generate metadata-based plots"
                     )
+                except AttributeError as e:
+                    warnings.warn(
+                        f"Error when trying to plot_all for triple collocation nc files. \nIssue: #59 {e}"
+                    )
 
         if save_csv:
             out_csv = plotter.save_stats()