Backport PR #41990: CI: troubleshoot py310 build (#42019)

Co-authored-by: jbrockmendel <[email protected]>

Author: meeseeksmachine

pandas/_libs/algos.pyx

@@ -1030,9 +1030,9 @@ def rank_1d(
         if rank_t is object:
             nan_fill_val = Infinity()
         elif rank_t is int64_t:
-            nan_fill_val = np.iinfo(np.int64).max
+            nan_fill_val = util.INT64_MAX
         elif rank_t is uint64_t:
-            nan_fill_val = np.iinfo(np.uint64).max
+            nan_fill_val = util.UINT64_MAX
         else:
             nan_fill_val = np.inf
         order = (masked_vals, mask, labels)
@@ -1393,7 +1393,7 @@ def rank_2d(
 
             # int64 and datetimelike
             else:
-                nan_value = np.iinfo(np.int64).max
+                nan_value = util.INT64_MAX
 
         else:
             if rank_t is object:

pandas/_libs/lib.pyi

@@ -25,6 +25,9 @@ class NoDefault(Enum): ...
 
 no_default: NoDefault
 
+i8max: int
+u8max: int
+
 def item_from_zerodim(val: object) -> object: ...
 def infer_dtype(value: object, skipna: bool = True) -> str: ...
 def is_iterator(obj: object) -> bool: ...

pandas/_libs/lib.pyx

@@ -118,6 +118,10 @@ cdef:
 
     float64_t NaN = <float64_t>np.NaN
 
+# python-visible
+i8max = <int64_t>INT64_MAX
+u8max = <uint64_t>UINT64_MAX
+
 
 @cython.wraparound(False)
 @cython.boundscheck(False)

pandas/core/algorithms.py

@@ -1092,18 +1092,19 @@ def checked_add_with_arr(
     # it is negative, we then check whether its sum with the element in
     # 'arr' exceeds np.iinfo(np.int64).min. If so, we have an overflow
     # error as well.
+    i8max = lib.i8max
+    i8min = iNaT
+
     mask1 = b2 > 0
     mask2 = b2 < 0
 
     if not mask1.any():
-        to_raise = ((np.iinfo(np.int64).min - b2 > arr) & not_nan).any()
+        to_raise = ((i8min - b2 > arr) & not_nan).any()
     elif not mask2.any():
-        to_raise = ((np.iinfo(np.int64).max - b2 < arr) & not_nan).any()
+        to_raise = ((i8max - b2 < arr) & not_nan).any()
     else:
-        to_raise = (
-            (np.iinfo(np.int64).max - b2[mask1] < arr[mask1]) & not_nan[mask1]
-        ).any() or (
-            (np.iinfo(np.int64).min - b2[mask2] > arr[mask2]) & not_nan[mask2]
+        to_raise = ((i8max - b2[mask1] < arr[mask1]) & not_nan[mask1]).any() or (
+            (i8min - b2[mask2] > arr[mask2]) & not_nan[mask2]
         ).any()
 
     if to_raise:

pandas/core/arrays/_ranges.py

@@ -6,6 +6,7 @@
 
 import numpy as np
 
+from pandas._libs.lib import i8max
 from pandas._libs.tslibs import (
     BaseOffset,
     OutOfBoundsDatetime,
@@ -103,7 +104,7 @@ def _generate_range_overflow_safe(
     # GH#14187 raise instead of incorrectly wrapping around
     assert side in ["start", "end"]
 
-    i64max = np.uint64(np.iinfo(np.int64).max)
+    i64max = np.uint64(i8max)
     msg = f"Cannot generate range with {side}={endpoint} and periods={periods}"
 
     with np.errstate(over="raise"):
@@ -180,7 +181,7 @@ def _generate_range_overflow_safe_signed(
             # error: Incompatible types in assignment (expression has type
             # "unsignedinteger[_64Bit]", variable has type "signedinteger[_64Bit]")
             result = np.uint64(endpoint) + np.uint64(addend)  # type: ignore[assignment]
-            i64max = np.uint64(np.iinfo(np.int64).max)
+            i64max = np.uint64(i8max)
             assert result > i64max
             if result <= i64max + np.uint64(stride):
                 # error: Incompatible return value type (got "unsignedinteger", expected

pandas/core/nanops.py

@@ -205,7 +205,7 @@ def _get_fill_value(
     else:
         if fill_value_typ == "+inf":
             # need the max int here
-            return np.iinfo(np.int64).max
+            return lib.i8max
         else:
             return iNaT
 
@@ -376,7 +376,7 @@ def _wrap_results(result, dtype: np.dtype, fill_value=None):
                 result = np.nan
 
             # raise if we have a timedelta64[ns] which is too large
-            if np.fabs(result) > np.iinfo(np.int64).max:
+            if np.fabs(result) > lib.i8max:
                 raise ValueError("overflow in timedelta operation")
 
             result = Timedelta(result, unit="ns")
@@ -1758,7 +1758,7 @@ def na_accum_func(values: ArrayLike, accum_func, *, skipna: bool) -> ArrayLike:
         if accum_func == np.minimum.accumulate:
             # Note: the accum_func comparison fails as an "is" comparison
             y = values.view("i8")
-            y[mask] = np.iinfo(np.int64).max
+            y[mask] = lib.i8max
             changed = True
         else:
             y = values

pandas/core/sorting.py

@@ -40,8 +40,6 @@
     from pandas import MultiIndex
     from pandas.core.indexes.base import Index
 
-_INT64_MAX = np.iinfo(np.int64).max
-
 
 def get_indexer_indexer(
     target: Index,
@@ -133,7 +131,7 @@ def _int64_cut_off(shape) -> int:
         acc = 1
         for i, mul in enumerate(shape):
             acc *= int(mul)
-            if not acc < _INT64_MAX:
+            if not acc < lib.i8max:
                 return i
         return len(shape)
 
@@ -153,7 +151,7 @@ def maybe_lift(lab, size) -> tuple[np.ndarray, int]:
     labels = list(labels)
 
     # Iteratively process all the labels in chunks sized so less
-    # than _INT64_MAX unique int ids will be required for each chunk
+    # than lib.i8max unique int ids will be required for each chunk
     while True:
         # how many levels can be done without overflow:
         nlev = _int64_cut_off(lshape)
@@ -215,7 +213,7 @@ def is_int64_overflow_possible(shape) -> bool:
     for x in shape:
         the_prod *= int(x)
 
-    return the_prod >= _INT64_MAX
+    return the_prod >= lib.i8max
 
 
 def decons_group_index(comp_labels, shape):

pandas/core/util/hashing.py

@@ -14,6 +14,7 @@
 
 import numpy as np
 
+from pandas._libs import lib
 from pandas._libs.hashing import hash_object_array
 from pandas._typing import (
     ArrayLike,
@@ -244,7 +245,7 @@ def _hash_categorical(cat: Categorical, encoding: str, hash_key: str) -> np.ndar
         result = np.zeros(len(mask), dtype="uint64")
 
     if mask.any():
-        result[mask] = np.iinfo(np.uint64).max
+        result[mask] = lib.u8max
 
     return result
 

pandas/tests/scalar/timedelta/test_timedelta.py

@@ -4,6 +4,7 @@
 import numpy as np
 import pytest
 
+from pandas._libs import lib
 from pandas._libs.tslibs import (
     NaT,
     iNaT,
@@ -391,8 +392,7 @@ def test_round_implementation_bounds(self):
         "method", [Timedelta.round, Timedelta.floor, Timedelta.ceil]
     )
     def test_round_sanity(self, method, n, request):
-        iinfo = np.iinfo(np.int64)
-        val = np.random.randint(iinfo.min + 1, iinfo.max, dtype=np.int64)
+        val = np.random.randint(iNaT + 1, lib.i8max, dtype=np.int64)
         td = Timedelta(val)
 
         assert method(td, "ns") == td
@@ -552,8 +552,8 @@ def test_implementation_limits(self):
 
         # GH 12727
         # timedelta limits correspond to int64 boundaries
-        assert min_td.value == np.iinfo(np.int64).min + 1
-        assert max_td.value == np.iinfo(np.int64).max
+        assert min_td.value == iNaT + 1
+        assert max_td.value == lib.i8max
 
         # Beyond lower limit, a NAT before the Overflow
         assert (min_td - Timedelta(1, "ns")) is NaT

pandas/tests/scalar/timestamp/test_unary_ops.py

@@ -6,11 +6,13 @@
 import pytz
 from pytz import utc
 
+from pandas._libs import lib
 from pandas._libs.tslibs import (
     NaT,
     Timedelta,
     Timestamp,
     conversion,
+    iNaT,
     to_offset,
 )
 from pandas._libs.tslibs.period import INVALID_FREQ_ERR_MSG
@@ -279,8 +281,7 @@ def test_round_implementation_bounds(self):
         "method", [Timestamp.round, Timestamp.floor, Timestamp.ceil]
     )
     def test_round_sanity(self, method, n):
-        iinfo = np.iinfo(np.int64)
-        val = np.random.randint(iinfo.min + 1, iinfo.max, dtype=np.int64)
+        val = np.random.randint(iNaT + 1, lib.i8max, dtype=np.int64)
         ts = Timestamp(val)
 
         def checker(res, ts, nanos):