ENH: Enable read_csv interpret 'Infinity' as floating point value (issue pandas-dev#10065)

Author: githeap

doc/source/whatsnew/v0.25.2.rst

@@ -61,7 +61,7 @@ Missing
 
 I/O
 ^^^
-
+- Improve infinity parsing. ``pd.read_csv()`` will now interpret ``Infinity``, ``+Infinity``, ``-Infinity`` as floating point values  (:issue:`10065`).
 -
 -
 -

pandas/_libs/parsers.pyx

@@ -1693,6 +1693,10 @@ cdef:
     char* cposinf = b'+inf'
     char* cneginf = b'-inf'
 
+    char* cinfty = b'Infinity'
+    char* cposinfty = b'+Infinity'
+    char* cneginfty = b'-Infinity'
+
 
 cdef _try_double(parser_t *parser, int64_t col,
                  int64_t line_start, int64_t line_end,
@@ -1772,9 +1776,12 @@ cdef inline int _try_double_nogil(parser_t *parser,
                 if error != 0 or p_end == word or p_end[0]:
                     error = 0
                     if (strcasecmp(word, cinf) == 0 or
-                            strcasecmp(word, cposinf) == 0):
+                        strcasecmp(word, cposinf) == 0 or
+                        strcasecmp(word, cinfty) == 0 or
+                        strcasecmp(word, cposinfty) == 0):
                         data[0] = INF
-                    elif strcasecmp(word, cneginf) == 0:
+                    elif (strcasecmp(word, cneginf) == 0 or 
+                          strcasecmp(word, cneginfty) == 0 ):
                         data[0] = NEGINF
                     else:
                         return 1
@@ -1793,9 +1800,12 @@ cdef inline int _try_double_nogil(parser_t *parser,
             if error != 0 or p_end == word or p_end[0]:
                 error = 0
                 if (strcasecmp(word, cinf) == 0 or
-                        strcasecmp(word, cposinf) == 0):
+                        strcasecmp(word, cposinf) == 0 or
+                        strcasecmp(word, cinfty) == 0 or
+                        strcasecmp(word, cposinfty) == 0):
                     data[0] = INF
-                elif strcasecmp(word, cneginf) == 0:
+                elif (strcasecmp(word, cneginf) == 0 or 
+                          strcasecmp(word, cneginfty) == 0):
                     data[0] = NEGINF
                 else:
                     return 1

pandas/tests/io/parser/test_common.py

@@ -1865,6 +1865,23 @@ def test_inf_parsing(all_parsers, na_filter):
     tm.assert_frame_equal(result, expected)
 
 
+@pytest.mark.parametrize("na_filter", [True, False])
+def test_infinity_parsing(all_parsers, na_filter):
+    parser = all_parsers
+    data = """\
+,A
+a,Infinity
+b,-Infinity
+c,+Infinity
+"""
+    expected = DataFrame(
+        {"A": [float("inf"), float("-inf"), float("inf")]},
+        index=["a", "b", "c"],
+    )
+    result = parser.read_csv(StringIO(data), index_col=0, na_filter=na_filter)
+    tm.assert_frame_equal(result, expected)
+
+
 @pytest.mark.parametrize("nrows", [0, 1, 2, 3, 4, 5])
 def test_raise_on_no_columns(all_parsers, nrows):
     parser = all_parsers