changes to set_index to preserve dtype from field

Author: Trevor Bye

pandas/core/frame.py

@@ -4249,6 +4249,7 @@ def set_index(
             "one-dimensional arrays."
         )
 
+        current_dtype = None
         missing: List[Optional[Hashable]] = []
         for col in keys:
             if isinstance(
@@ -4262,6 +4263,9 @@ def set_index(
                 # everything else gets tried as a key; see GH 24969
                 try:
                     found = col in self.columns
+                    if found:
+                        # get current dtype to preserve through index creation
+                        current_dtype = self.dtypes.get(col).type
                 except TypeError:
                     raise TypeError(f"{err_msg}. Received column of type {type(col)}")
                 else:
@@ -4317,7 +4321,7 @@ def set_index(
                     f"received array of length {len(arrays[-1])}"
                 )
 
-        index = ensure_index_from_sequences(arrays, names)
+        index = ensure_index_from_sequences(arrays, names, current_dtype)
 
         if verify_integrity and not index.is_unique:
             duplicates = index[index.duplicated()].unique()

pandas/core/indexes/base.py

@@ -5238,7 +5238,7 @@ def shape(self):
 Index._add_comparison_methods()
 
 
-def ensure_index_from_sequences(sequences, names=None):
+def ensure_index_from_sequences(sequences, names=None, dtype=None):
     """
     Construct an index from sequences of data.
 
@@ -5249,6 +5249,7 @@ def ensure_index_from_sequences(sequences, names=None):
     ----------
     sequences : sequence of sequences
     names : sequence of str
+    dtype : NumPy dtype
 
     Returns
     -------
@@ -5274,7 +5275,7 @@ def ensure_index_from_sequences(sequences, names=None):
     if len(sequences) == 1:
         if names is not None:
             names = names[0]
-        return Index(sequences[0], name=names)
+        return Index(sequences[0], name=names, dtype=dtype)
     else:
         return MultiIndex.from_arrays(sequences, names=names)