BUG: Don't segfault to_numeric when input is empty (#16305)

Closes gh-16302.
(cherry picked from commit 81aa70c)

Author: gfyoung

doc/source/whatsnew/v0.20.2.txt

@@ -39,7 +39,7 @@ Bug Fixes
 Conversion
 ^^^^^^^^^^
 
-
+- Bug in ``pd.to_numeric()`` in which empty data inputs were causing Python to crash (:issue:`16302`)
 
 
 Indexing
@@ -51,7 +51,7 @@ Indexing
 I/O
 ^^^
 
-- Bug that would force importing of the clipboard routines unecessarily, potentially causing an import error on startup (:issue:`16288`)
+- Bug that would force importing of the clipboard routines unnecessarily, potentially causing an import error on startup (:issue:`16288`)
 
 
 Plotting

pandas/_libs/src/inference.pyx

@@ -947,8 +947,13 @@ def maybe_convert_numeric(ndarray[object] values, set na_values,
     -------
     numeric_array : array of converted object values to numerical ones
     """
+
+    if len(values) == 0:
+        return np.array([], dtype='i8')
+
     # fastpath for ints - try to convert all based on first value
     cdef object val = values[0]
+
     if util.is_integer_object(val):
         try:
             maybe_ints = values.astype('i8')

pandas/tests/tools/test_numeric.py

@@ -11,6 +11,21 @@
 
 class TestToNumeric(object):
 
+    def test_empty(self):
+        # see gh-16302
+        s = pd.Series([], dtype=object)
+
+        res = to_numeric(s)
+        expected = pd.Series([], dtype=np.int64)
+
+        tm.assert_series_equal(res, expected)
+
+        # Original issue example
+        res = to_numeric(s, errors='coerce', downcast='integer')
+        expected = pd.Series([], dtype=np.int8)
+
+        tm.assert_series_equal(res, expected)
+
     def test_series(self):
         s = pd.Series(['1', '-3.14', '7'])
         res = to_numeric(s)