Follow-Up Translations

This merge requests incorporates the following changes/additions:

• Allow the use of translator specific keywords in function calls, e.g. comment and cause for the DmpTranslator
• Generate valid Historys from given input flags. The user flags usually lack information (i.e. concrete parameter values, order of function calls, ...) to infer a real history. I loosed the definition of the word valid, to 'rerunning the history inferred from a DataFrame of user flags on the same input data, needs to yield the same user flags again'
• Make the different translators available through the CLI (i.e. the exe)

saqc/core/translator/dmptranslator.py

@@ -60,7 +60,22 @@ class DmpTranslator(Translator):
         ----
         Could (and maybe should) be implemented as a method of `CallGraph`
         """
-        return [SaQCFunction(name="")] + [f for l, f in call_stack if l.field == field]
+        out = [SaQCFunction(name="")]
+
+        for sel, func in call_stack:
+            if sel.field == field:
+                out.append(func)
+                # NOTE:
+                # This is an intermediary hack, to work around
+                # the problem, that field names are mutable and
+                # used as an mapping between `History` and
+                # `call_stack`. There are better ideas, to solve
+                # this (i.e. global function pointer) but for the
+                # moment this has to do the trick
+                if func.name == "tools.rename":
+                    field = func.keywords.get("new_name") or func.args[3]
+
+        return out
 
     def forward(self, flags: pd.DataFrame) -> Tuple[Flags, MaterializedGraph]:
         """
@@ -144,6 +159,11 @@ class DmpTranslator(Translator):
             flag_call_history = self._getFieldFunctions(field, call_graph)
             flag_pos = flags.history[field].idxmax()
             comments, causes = [], []
+            # NOTE:
+            # Strangely enough, this loop withstood all my efforts
+            # to speed it up through vectorization - the simple
+            # loop always outperformed even careful `pd.DataFrame.apply`
+            # versions. The latest try is left as a comment below.
             for p in flag_pos:
                 func = flag_call_history[p]
                 cause = func.keywords.get("cause", self.ARGUMENTS["cause"])
@@ -165,3 +185,26 @@ class DmpTranslator(Translator):
             }
             out[field] = pd.DataFrame(var_flags)
         return pd.concat(out, axis="columns")
+
+        # for field in tflags.columns:
+        #     call_history = []
+        #     for func in self._getFieldFunctions(field, call_graph):
+        #         func_info = {
+        #             "cause": func.keywords.get("cause", self.ARGUMENTS["comment"]),
+        #             "comment": json.dumps({
+        #                 "test": func.name,
+        #                 "comment": func.keywords.get("comment", self.ARGUMENTS["comment"]),
+        #             })
+        #          }
+        #         call_history.append(func_info)
+
+        #     functions = pd.DataFrame(call_history)
+        #     flag_pos = flags.history[field].idxmax()
+
+        #     var_flags = {
+        #         "quality_flag": tflags[field].reset_index(drop=True),
+        #         "quality_comment": functions.loc[flag_pos, "comment"].reset_index(drop=True),
+        #         "quality_cause": functions.loc[flag_pos, "cause"].reset_index(drop=True),
+        #     }
+        #     out[field] = pd.DataFrame(var_flags, index=flag_pos.index)
+        # return pd.concat(out, axis="columns")