Merge pull request #78 from awst-austria/fix_metaplots

Fix bugs with metadata plots

src/qa4sm_reader/plotter.py

@@ -108,7 +108,7 @@ def _box_caption(Var,
         ds_parts = []
         id, meta = mds_meta
         if tc:
-            id, meta = other_meta
+            id, meta = other_meta   # show name of the OTHER dataset
         if short_caption:
             ds_parts.append(
                 f"{id}-{meta['pretty_name']} ({meta['pretty_version']})")
@@ -117,10 +117,8 @@ def _box_caption(Var,
                 id, meta['pretty_name'], meta['pretty_version'],
                 meta['pretty_variable'], meta['mu']))
         capt = '\n and \n'.join(ds_parts)
-
         if tc:
             capt = 'Other Data:\n' + capt
-
         return capt
 
     @staticmethod
@@ -513,7 +511,7 @@ def boxplot_basic(self,
         if not values:
             return None
         # put all Variables in the same dataframe
-        values = pd.concat(values)
+        values = pd.concat(values, axis=1)
         # create plot
         fig, ax = self._boxplot_definition(metric=metric,
                                            df=values,
@@ -909,6 +907,18 @@ def meta_single(self,
         mu = globals._metric_description[metric].format(
             globals.get_metric_units(unit_ref))
 
+        # For tca with > 3 datasets, it can be that the same metric comes
+        # from different combinations of datasets, in that case take average
+        # the different versions
+        for col in np.unique(values.columns):
+            loc = np.where(values.columns == col)[0]
+            if len(loc) == 1:
+                continue
+            else:
+                colmean = values.pop(col).mean(axis=1, skipna=True)
+                values[col] = colmean
+
+
         out = plm.boxplot_metadata(df=values,
                                    metadata_values=meta_values,
                                    ax_label=Var.Metric.pretty_name + mu,

src/qa4sm_reader/plotting_methods.py

@@ -1020,7 +1020,7 @@ def bin_discrete(
         group.columns = ["values", meta_key]
         group["Dataset"] = col
         groups.append(group)
-    grouped = pd.concat(groups)
+    grouped = pd.concat(groups, axis=0)
     formatted = []
     for meta, meta_df in grouped.groupby(meta_key).__iter__():
         if meta_df["values"].count() < min_size:
@@ -1238,17 +1238,20 @@ def _dict2df(to_plot_dict: dict, meta_key: str) -> pd.DataFrame:
             values_ds["Dataset"] = ds
             values_ds[meta_key] = "\n[".join(range.split(" ["))
             range_grouped.append(values_ds)
-        range_grouped = pd.concat(range_grouped)
+        range_grouped = pd.concat(range_grouped, axis=0)
         to_plot_df.append(range_grouped)
-    to_plot_df = pd.concat(to_plot_df)
+    to_plot_df = pd.concat(to_plot_df, axis=0)
 
     return to_plot_df
 
 
 def add_cat_info(to_plot: pd.DataFrame, metadata_name: str) -> pd.DataFrame:
     """Add info (N, median value) to metadata category labels"""
-    groups = to_plot.groupby(metadata_name)["values"]
-    counts = groups.count()
+    groups = to_plot.groupby(metadata_name)["values"]#
+    counts = {}
+    for name, group in groups:
+        counts[name] = group[~group.index.duplicated(keep='first')].index.size
+
     to_plot[metadata_name] = to_plot[metadata_name].apply(
         lambda x: x + "\nN: {}".format(counts[x]))
 
@@ -1296,7 +1299,7 @@ def bplot_catplot(to_plot,
         x=x,
         y=y,
         hue="Dataset",
-        data=to_plot,
+        data=to_plot.set_index(np.arange(to_plot.index.size)),
         palette="Set2",
         ax=axis,
         showfliers=False,