drop instead of select

prototypes/external_code.py

@@ -73,7 +73,6 @@ def preprocess_psm(
 
     # drop duplicates
     # use null count as information measure, least amount of nulls should be contained
-    base_columns = data.columns
     data = data.with_columns(pl.sum_horizontal(pl.all().is_null()).alias("null_count"))
     data = data.sort(PRIM_KEYS + ["Meldezeitpunkt_Historie", "null_count"], descending=False)
     filtered_data = pl.concat(
@@ -81,7 +80,7 @@ def preprocess_psm(
             filtered_data,
             data.filter(
                 ~pl.struct(PRIM_KEYS + ["Meldezeitpunkt_Historie"]).is_first_distinct()
-            ).select(base_columns),
+            ).drop("null_count"),
         ]
     )
     data = data.filter(pl.struct(PRIM_KEYS + ["Meldezeitpunkt_Historie"]).is_first_distinct())