Several fixes

A container branch for the fixes of several issues I encountered while rolling out the current SaQC on the SEEFO pipeline.

Not sure, if this needs a real review, but as some of the changes touched code I am not completely familiar with, maybe you, @palmb and @luenensc want to have a look anyhow?

saqc/core/history.py

@@ -78,7 +78,7 @@ class History:
             hist = hist.copy()
             mask = mask.copy()
 
-        self.hist = hist
+        self.hist = hist.astype("category", copy=copy)
         self.mask = mask
 
     @property
@@ -166,7 +166,7 @@ class History:
             touched = s.notna()
             self.mask.iloc[touched, :pos] = False
 
-        self.hist[pos] = s
+        self.hist[pos] = s.astype("category")
 
         return self
 
@@ -263,7 +263,7 @@ class History:
         # clear the current mask
         self.mask.loc[(~value_mask & value_hist.notna()).any(axis="columns")] = False
 
-        self.hist.loc[:, columns] = value_hist.copy()
+        self.hist.loc[:, columns] = value_hist.astype("category")
         self.mask.loc[:, columns] = value_mask.copy()
         return self
 
@@ -309,7 +309,7 @@ class History:
         # this may leave us in an unstable state, because
         # the last column maybe is not entirely True, but
         # the following append, will fix this
-        self.hist = self.hist.iloc[:, :-n]
+        self.hist = self.hist.iloc[:, :-n].astype("category")
         self.mask = self.mask.iloc[:, :-n]
 
         self.append(s)
@@ -323,7 +323,7 @@ class History:
         -------
         pd.Series: maximum values
         """
-        return self.hist[self.mask].idxmax(axis=1)
+        return self.hist[self.mask].astype(float).idxmax(axis=1)
 
     def max(self) -> pd.Series:
         """
@@ -371,11 +371,18 @@ class History:
         -------
         History
         """
-        self.hist = self.hist.reindex(index=index, copy=False, fill_value=np.nan)
-        self.mask = self.mask.reindex(index=index, copy=False, fill_value=False)
+        hist = self.hist.astype(float).reindex(
+            index=index, copy=False, fill_value=np.nan
+        )
+        mask = self.mask.astype(bool).reindex(index=index, copy=False, fill_value=False)
+
         # Note: all following code must handle empty frames
-        self.hist.iloc[:, -1:] = self.hist.iloc[:, -1:].fillna(fill_value_last)
-        self.mask.iloc[:, -1:] = True
+        hist.iloc[:, -1:] = hist.iloc[:, -1:].fillna(fill_value_last)
+        mask.iloc[:, -1:] = True
+
+        self.mask = mask.astype(bool)
+        self.hist.hist = hist.astype("category")
+
         return self
 
     def __copy__(self, deep: bool = True):
@@ -455,8 +462,10 @@ class History:
                 f"'hist' must be of type pd.DataFrame, but {type(obj).__name__} was given"
             )
 
-        if (obj.dtypes != float).any():
-            raise ValueError("dtype of all columns in hist must be float")
+        if obj.dtypes.isin([float, pd.Categorical]).any() is False:
+            raise ValueError(
+                "dtype of all columns in hist must be float or categorical"
+            )
 
         if not obj.empty and (
             not obj.columns.equals(pd.Index(range(len(obj.columns))))
@@ -478,8 +487,8 @@ class History:
                 f"value must be of type pd.Series, but {type(obj).__name__} was given"
             )
 
-        if not obj.dtype == float:
-            raise ValueError("dtype must be float")
+        if not ((obj.dtype == float) or isinstance(obj.dtype, pd.CategoricalDtype)):
+            raise ValueError("dtype must be float or categorical")
 
         return obj
 
@@ -533,12 +542,17 @@ def applyFunctionOnHistory(
     new_history = History()
 
     if func_handle_df:
-        history.hist = hist_func(history.hist, **hist_kws)
+        # we need to pass the data as floats as functions may fail with Categorical
+        history.hist = hist_func(history.hist.astype(float), **hist_kws).astype(
+            "category"
+        )
         history.mask = hist_func(history.mask, **mask_kws)
 
     else:
         for pos in history.columns:
-            new_history.hist[pos] = hist_func(history.hist[pos], **hist_kws)
+            new_history.hist[pos] = hist_func(
+                history.hist[pos].astype(float), **hist_kws
+            ).astype("category")
             new_history.mask[pos] = mask_func(history.mask[pos], **mask_kws)
 
     # handle unstable state
@@ -548,7 +562,7 @@ def applyFunctionOnHistory(
         if isinstance(last_column, str) and last_column == "dummy":
             last_column = pd.Series(UNTOUCHED, index=new_history.index, dtype=float)
 
-        new_history.append(last_column, force=True)
+        new_history.append(last_column.astype("category"), force=True)
 
     # assure a boolean mask and UNFLAGGED column
     new_history.mask = new_history.mask.fillna(True).astype(bool)