move int8/16/32 type conversions to index_class_helper.pxi.in

Author: topper-123

pandas/_libs/index.pyx

@@ -5,9 +5,10 @@ import cython
 
 import numpy as np
 cimport numpy as cnp
-from numpy cimport (ndarray, float64_t, int32_t,
-                    int8_t, int16_t, int32_t, int64_t,
-                    uint8_t, uint64_t,
+from numpy cimport (ndarray,
+                    float64_t, float32_t,
+                    int64_t,int32_t, int16_t, int8_t,
+                    uint64_t, uint32_t, uint16_t, uint8_t,
                     intp_t,
                     # Note: NPY_DATETIME, NPY_TIMEDELTA are only available
                     # for cimport in cython>=0.27.3
@@ -244,8 +245,6 @@ cdef class IndexEngine:
         if not self.is_mapping_populated:
 
             values = self._get_index_values()
-            if values.dtype in {'int8', 'int16', 'int32'}:
-                values = algos.ensure_int64(values)
             self.mapping = self._make_hash_table(len(values))
             self._call_map_locations(values)
 

pandas/_libs/index_class_helper.pxi.in

@@ -12,11 +12,15 @@ WARNING: DO NOT edit .pxi FILE directly, .pxi is generated from .pxi.in
 
 # name, dtype, ctype
 dtypes = [('Float64', 'float64', 'float64_t'),
+          ('Float32', 'float32', 'float32_t'),
           ('Int64', 'int64', 'int64_t'),
           ('Int32', 'int32', 'int32_t'),
           ('Int16', 'int16', 'int16_t'),
           ('Int8', 'int8', 'int8_t'),
           ('UInt64', 'uint64', 'uint64_t'),
+          ('UInt32', 'uint32', 'uint32_t'),
+          ('UInt16', 'uint16', 'uint16_t'),
+          ('UInt8', 'uint8', 'uint8_t'),
           ('Object', 'object', 'object'),
           ]
 }}
@@ -41,11 +45,24 @@ cdef class {{name}}Engine(IndexEngine):
         {{if name == 'Object'}}
         return _hash.PyObjectHashTable(n)
         {{elif name in {'Int8', 'Int16', 'Int32'} }}
+        # {{name}}HashTable is not available, so we use Int64HashTable
         return _hash.Int64HashTable(n)
+        {{elif name in {'UInt8', 'UInt16', 'UInt32'} }}
+        # {{name}}HashTable is not available, so we use UInt64HashTable
+        return _hash.UInt64HashTable(n)
+        {{elif name in {'Float32'} }}
+        # {{name}}HashTable is not available, so we use Float64HashTable
+        return _hash.Float64HashTable(n)
         {{else}}
         return _hash.{{name}}HashTable(n)
         {{endif}}
 
+    {{if name in {'Int8', 'Int16', 'Int32'} }}
+    cpdef _call_map_locations(self, values):
+        # self.mapping is of type Int64HashTable, so convert dtype of values
+        self.mapping.map_locations(algos.ensure_int64(values))
+    {{endif}}
+
     {{if name != 'Float64' and name != 'Object'}}
     cdef _check_type(self, object val):
         hash(val)

pandas/core/indexes/category.py

@@ -86,8 +86,10 @@ class CategoricalIndex(Index, accessor.PandasDelegate):
 
     @property
     def _engine_type(self):
-        type_name = self.codes.dtype.name.capitalize()
-        return getattr(libindex, "{}Engine".format(type_name))
+        # self.codes can have dtype int8, int16, int 32 or int64, so we need
+        # to return the corresponding engine type (libindex.Int8Engine, etc.).
+        engine_name = "{}Engine".format(self.codes.dtype.name.capitalize())
+        return getattr(libindex, engine_name)
     _attributes = ['name']
 
     def __new__(cls, data=None, categories=None, ordered=None, dtype=None,