Avoid false positives in _cast_to_numeric

Previously, _cast_to_numeric would raise a TypeError if a column was
missing. We now check that a column exists before trying to cast it.

This does result in some redundant checking, as some functions call both
_require_columns and _cast_to_numeric. We may want to consider
refactoring to combine all of this functionality into a single DataFrame
validation step.

Signed-off-by: John Pennycook <[email protected]>

p3analysis/_utils.py

@@ -23,6 +23,7 @@ def _cast_to_numeric(df, columns):
     """
     Check that the named columns are numeric, and cast them.
     """
+    _require_columns(df, columns)
     result = df.copy(deep=True)
     for column in columns:
         try: