CLN: assorted cleanups in cython

pandas/_libs/algos.pyx

@@ -1,6 +1,5 @@
-import cython
-from cython import Py_ssize_t
-
+cimport cython
+from cython cimport Py_ssize_t
 from libc.math cimport (
     fabs,
     sqrt,

pandas/_libs/groupby.pyx

@@ -1,7 +1,8 @@
-import cython
-from cython import Py_ssize_t
-
-from cython cimport floating
+cimport cython
+from cython cimport (
+    Py_ssize_t,
+    floating,
+)
 from libc.stdlib cimport (
     free,
     malloc,
@@ -268,7 +269,6 @@ def group_cumsum(
                         out[i, j] = na_val
                         continue
 
-
                 if isna_entry:
                     out[i, j] = na_val
                     if not skipna:
@@ -1220,7 +1220,7 @@ def group_nth(
                         if nobs[lab, j] == rank:
                             resx[lab, j] = val
 
-            # TODO: de-dup this whoel block with group_last?
+            # TODO: de-dup this whole block with group_last?
             for i in range(ncounts):
                 for j in range(K):
                     if nobs[i, j] < min_count:
@@ -1232,6 +1232,9 @@ def group_nth(
                         #  set a placeholder value in out[i, j].
                         if uses_mask:
                             result_mask[i, j] = True
+                            # set out[i, j] to 0 to be deterministic, as
+                            #  it was initialized with np.empty. Also ensures
+                            #  we can downcast out if appropriate.
                             out[i, j] = 0
                         elif numeric_object_t is float32_t or numeric_object_t is float64_t:
                             out[i, j] = NAN
@@ -1369,7 +1372,7 @@ cdef group_min_max(
     """
     cdef:
         Py_ssize_t i, j, N, K, lab, ngroups = len(counts)
-        numeric_t val, nan_val
+        numeric_t val
         ndarray[numeric_t, ndim=2] group_min_or_max
         int64_t[:, ::1] nobs
         bint uses_mask = mask is not None
@@ -1386,20 +1389,6 @@ cdef group_min_max(
     group_min_or_max = np.empty_like(out)
     group_min_or_max[:] = _get_min_or_max(<numeric_t>0, compute_max, is_datetimelike)
 
-    # NB: We do not define nan_val because there is no such thing
-    # for uint64_t.  We carefully avoid having to reference it in this
-    # case.
-    if numeric_t is int64_t:
-        nan_val = NPY_NAT
-    elif numeric_t is int32_t:
-        nan_val = util.INT32_MIN
-    elif numeric_t is int16_t:
-        nan_val = util.INT16_MIN
-    elif numeric_t is int8_t:
-        nan_val = util.INT8_MIN
-    elif numeric_t is float64_t or numeric_t is float32_t:
-        nan_val = NAN
-
     N, K = (<object>values).shape
 
     with nogil:
@@ -1442,11 +1431,11 @@ cdef group_min_max(
                         #  we can downcast out if appropriate.
                         out[i, j] = 0
                     elif numeric_t is float32_t or numeric_t is float64_t:
-                        out[i, j] = nan_val
+                        out[i, j] = NAN
                     elif numeric_t is int64_t:
                         # Per above, this is a placeholder in
                         #  non-is_datetimelike cases.
-                        out[i, j] = nan_val
+                        out[i, j] = NPY_NAT
                     else:
                         # placeholder, see above
                         out[i, j] = 0

pandas/_libs/hashing.pyx

@@ -1,8 +1,7 @@
 # Translated from the reference implementation
 # at https://github.com/veorq/SipHash
 
-import cython
-
+cimport cython
 from libc.stdlib cimport (
     free,
     malloc,

pandas/_libs/internals.pyx

@@ -1,9 +1,8 @@
 from collections import defaultdict
 
-import cython
-from cython import Py_ssize_t
-
+cimport cython
 from cpython.slice cimport PySlice_GetIndicesEx
+from cython cimport Py_ssize_t
 
 
 cdef extern from "Python.h":

pandas/_libs/interval.pyx

@@ -11,6 +11,7 @@ from cpython.datetime cimport (
 
 import_datetime()
 
+cimport cython
 from cpython.object cimport (
     Py_EQ,
     Py_GE,
@@ -20,9 +21,8 @@ from cpython.object cimport (
     Py_NE,
     PyObject_RichCompare,
 )
+from cython cimport Py_ssize_t
 
-import cython
-from cython import Py_ssize_t
 import numpy as np
 
 cimport numpy as cnp

pandas/_libs/join.pyx

@@ -1,5 +1,5 @@
-import cython
-from cython import Py_ssize_t
+cimport cython
+from cython cimport Py_ssize_t
 import numpy as np
 
 cimport numpy as cnp
@@ -233,6 +233,8 @@ cdef void _get_result_indexer(intp_t[::1] sorter, intp_t[::1] indexer) nogil:
         indexer[:] = -1
 
 
+@cython.wraparound(False)
+@cython.boundscheck(False)
 def ffill_indexer(const intp_t[:] indexer) -> np.ndarray:
     cdef:
         Py_ssize_t i, n = len(indexer)

pandas/_libs/lib.pyx

@@ -3,9 +3,7 @@ from decimal import Decimal
 from enum import Enum
 import warnings
 
-import cython
-from cython import Py_ssize_t
-
+cimport cython
 from cpython.datetime cimport (
     PyDate_Check,
     PyDateTime_Check,
@@ -25,7 +23,10 @@ from cpython.tuple cimport (
     PyTuple_New,
     PyTuple_SET_ITEM,
 )
-from cython cimport floating
+from cython cimport (
+    Py_ssize_t,
+    floating,
+)
 
 import_datetime()
 

pandas/_libs/missing.pyx

@@ -2,8 +2,8 @@ from decimal import Decimal
 import numbers
 from sys import maxsize
 
-import cython
-from cython import Py_ssize_t
+cimport cython
+from cython cimport Py_ssize_t
 import numpy as np
 
 cimport numpy as cnp

pandas/_libs/ops.pyx

@@ -1,5 +1,6 @@
 import operator
 
+cimport cython
 from cpython.object cimport (
     Py_EQ,
     Py_GE,
@@ -9,9 +10,8 @@ from cpython.object cimport (
     Py_NE,
     PyObject_RichCompareBool,
 )
+from cython cimport Py_ssize_t
 
-import cython
-from cython import Py_ssize_t
 import numpy as np
 
 from numpy cimport (

pandas/_libs/parsers.pyx

@@ -12,16 +12,7 @@ import sys
 import time
 import warnings
 
-from libc.stdlib cimport free
-from libc.string cimport (
-    strcasecmp,
-    strlen,
-    strncpy,
-)
-
-import cython
-from cython import Py_ssize_t
-
+cimport cython
 from cpython.bytes cimport (
     PyBytes_AsString,
     PyBytes_FromString,
@@ -40,6 +31,13 @@ from cpython.unicode cimport (
     PyUnicode_Decode,
     PyUnicode_DecodeUTF8,
 )
+from cython cimport Py_ssize_t
+from libc.stdlib cimport free
+from libc.string cimport (
+    strcasecmp,
+    strlen,
+    strncpy,
+)
 
 
 cdef extern from "Python.h":

pandas/_libs/properties.pyx

@@ -1,10 +1,9 @@
-from cython import Py_ssize_t
-
 from cpython.dict cimport (
     PyDict_Contains,
     PyDict_GetItem,
     PyDict_SetItem,
 )
+from cython cimport Py_ssize_t
 
 
 cdef class CachedProperty:

pandas/_libs/reshape.pyx

@@ -1,6 +1,5 @@
-import cython
-from cython import Py_ssize_t
-
+cimport cython
+from cython cimport Py_ssize_t
 from numpy cimport (
     int64_t,
     ndarray,

pandas/_libs/sparse.pyx

@@ -1,4 +1,4 @@
-import cython
+cimport cython
 import numpy as np
 
 cimport numpy as cnp

pandas/_libs/tslib.pyx

@@ -1,7 +1,6 @@
 import warnings
 
-import cython
-
+cimport cython
 from cpython.datetime cimport (
     PyDate_Check,
     PyDateTime_Check,
@@ -28,14 +27,14 @@ cnp.import_array()
 import pytz
 
 from pandas._libs.tslibs.np_datetime cimport (
-    _string_to_dts,
     check_dts_bounds,
     dt64_to_dtstruct,
     dtstruct_to_dt64,
     get_datetime64_value,
     npy_datetimestruct,
     pydate_to_dt64,
     pydatetime_to_dt64,
+    string_to_dts,
 )
 from pandas._libs.util cimport (
     is_datetime64_object,
@@ -63,6 +62,7 @@ from pandas._libs.tslibs.timestamps cimport _Timestamp
 from pandas._libs.tslibs.timestamps import Timestamp
 
 # Note: this is the only non-tslibs intra-pandas dependency here
+
 from pandas._libs.missing cimport checknull_with_nat_and_na
 from pandas._libs.tslibs.tzconversion cimport tz_localize_to_utc_single
 
@@ -83,7 +83,7 @@ def _test_parse_iso8601(ts: str):
     elif ts == 'today':
         return Timestamp.now().normalize()
 
-    _string_to_dts(ts, &obj.dts, &out_local, &out_tzoffset, True)
+    string_to_dts(ts, &obj.dts, &out_local, &out_tzoffset, True)
     obj.value = dtstruct_to_dt64(&obj.dts)
     check_dts_bounds(&obj.dts)
     if out_local == 1:
@@ -515,7 +515,7 @@ cpdef array_to_datetime(
                         iresult[i] = NPY_NAT
                         continue
 
-                    string_to_dts_failed = _string_to_dts(
+                    string_to_dts_failed = string_to_dts(
                         val, &dts, &out_local,
                         &out_tzoffset, False
                     )

pandas/_libs/tslibs/ccalendar.pyx

@@ -3,8 +3,7 @@
 Cython implementations of functions resembling the stdlib calendar module
 """
 
-import cython
-
+cimport cython
 from numpy cimport (
     int32_t,
     int64_t,
@@ -51,8 +50,8 @@ weekday_to_int = {int_to_weekday[key]: key for key in int_to_weekday}
 DAY_SECONDS = 86400
 HOUR_SECONDS = 3600
 
-cdef int64_t DAY_NANOS = DAY_SECONDS * 1_000_000_000
-cdef int64_t HOUR_NANOS = HOUR_SECONDS * 1_000_000_000
+cdef const int64_t DAY_NANOS = DAY_SECONDS * 1_000_000_000
+cdef const int64_t HOUR_NANOS = HOUR_SECONDS * 1_000_000_000
 
 # ----------------------------------------------------------------------
 

pandas/_libs/tslibs/conversion.pyx

@@ -1,4 +1,5 @@
-import cython
+cimport cython
+
 import numpy as np
 
 cimport numpy as cnp
@@ -31,7 +32,6 @@ from pandas._libs.tslibs.base cimport ABCTimestamp
 from pandas._libs.tslibs.np_datetime cimport (
     NPY_DATETIMEUNIT,
     NPY_FR_ns,
-    _string_to_dts,
     astype_overflowsafe,
     check_dts_bounds,
     dt64_to_dtstruct,
@@ -43,6 +43,7 @@ from pandas._libs.tslibs.np_datetime cimport (
     npy_datetimestruct,
     pandas_datetime_to_datetimestruct,
     pydatetime_to_dt64,
+    string_to_dts,
 )
 
 from pandas._libs.tslibs.np_datetime import (
@@ -602,7 +603,7 @@ cdef _TSObject _convert_str_to_tsobject(object ts, tzinfo tz, str unit,
         dt = datetime.now(tz)
         # equiv: datetime.today().replace(tzinfo=tz)
     else:
-        string_to_dts_failed = _string_to_dts(
+        string_to_dts_failed = string_to_dts(
             ts, &dts, &out_local,
             &out_tzoffset, False
         )

pandas/_libs/tslibs/dtypes.pyx

@@ -317,6 +317,7 @@ cdef NPY_DATETIMEUNIT freq_group_code_to_npy_unit(int freq) nogil:
         return NPY_DATETIMEUNIT.NPY_FR_D
 
 
+# TODO: use in _matplotlib.converter?
 cdef int64_t periods_per_day(NPY_DATETIMEUNIT reso=NPY_DATETIMEUNIT.NPY_FR_ns) except? -1:
     """
     How many of the given time units fit into a single day?

pandas/_libs/tslibs/fields.pyx

@@ -5,8 +5,8 @@ objects and arrays
 from locale import LC_TIME
 
 from _strptime import LocaleTime
-import cython
-from cython import Py_ssize_t
+cimport cython
+from cython cimport Py_ssize_t
 import numpy as np
 
 cimport numpy as cnp