1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
|
from __future__ import annotations
from functools import lru_cache
from importlib import import_module
from typing import TYPE_CHECKING, Any, overload
from narwhals._utils import Implementation, isinstance_or_issubclass
from narwhals.exceptions import UnsupportedDTypeError
if TYPE_CHECKING:
from types import ModuleType
import sqlframe.base.types as sqlframe_types
from sqlframe.base.column import Column
from sqlframe.base.session import _BaseSession as Session
from typing_extensions import TypeAlias
from narwhals._spark_like.dataframe import SparkLikeLazyFrame
from narwhals._spark_like.expr import SparkLikeExpr
from narwhals._utils import Version
from narwhals.dtypes import DType
from narwhals.typing import IntoDType
_NativeDType: TypeAlias = sqlframe_types.DataType
SparkSession = Session[Any, Any, Any, Any, Any, Any, Any]
UNITS_DICT = {
"y": "year",
"q": "quarter",
"mo": "month",
"d": "day",
"h": "hour",
"m": "minute",
"s": "second",
"ms": "millisecond",
"us": "microsecond",
"ns": "nanosecond",
}
# see https://spark.apache.org/docs/latest/sql-ref-datetime-pattern.html
# and https://docs.python.org/3/library/datetime.html#strftime-strptime-behavior
DATETIME_PATTERNS_MAPPING = {
"%Y": "yyyy", # Year with century (4 digits)
"%y": "yy", # Year without century (2 digits)
"%m": "MM", # Month (01-12)
"%d": "dd", # Day of the month (01-31)
"%H": "HH", # Hour (24-hour clock) (00-23)
"%I": "hh", # Hour (12-hour clock) (01-12)
"%M": "mm", # Minute (00-59)
"%S": "ss", # Second (00-59)
"%f": "S", # Microseconds -> Milliseconds
"%p": "a", # AM/PM
"%a": "E", # Abbreviated weekday name
"%A": "E", # Full weekday name
"%j": "D", # Day of the year
"%z": "Z", # Timezone offset
"%s": "X", # Unix timestamp
}
# NOTE: don't lru_cache this as `ModuleType` isn't hashable
def native_to_narwhals_dtype( # noqa: C901, PLR0912
dtype: _NativeDType, version: Version, spark_types: ModuleType, session: SparkSession
) -> DType:
dtypes = version.dtypes
if TYPE_CHECKING:
native = sqlframe_types
else:
native = spark_types
if isinstance(dtype, native.DoubleType):
return dtypes.Float64()
if isinstance(dtype, native.FloatType):
return dtypes.Float32()
if isinstance(dtype, native.LongType):
return dtypes.Int64()
if isinstance(dtype, native.IntegerType):
return dtypes.Int32()
if isinstance(dtype, native.ShortType):
return dtypes.Int16()
if isinstance(dtype, native.ByteType):
return dtypes.Int8()
if isinstance(dtype, (native.StringType, native.VarcharType, native.CharType)):
return dtypes.String()
if isinstance(dtype, native.BooleanType):
return dtypes.Boolean()
if isinstance(dtype, native.DateType):
return dtypes.Date()
if isinstance(dtype, native.TimestampNTZType):
# TODO(marco): cover this
return dtypes.Datetime() # pragma: no cover
if isinstance(dtype, native.TimestampType):
return dtypes.Datetime(time_zone=fetch_session_time_zone(session))
if isinstance(dtype, native.DecimalType):
# TODO(marco): cover this
return dtypes.Decimal() # pragma: no cover
if isinstance(dtype, native.ArrayType):
return dtypes.List(
inner=native_to_narwhals_dtype(
dtype.elementType, version, spark_types, session
)
)
if isinstance(dtype, native.StructType):
return dtypes.Struct(
fields=[
dtypes.Field(
name=field.name,
dtype=native_to_narwhals_dtype(
field.dataType, version, spark_types, session
),
)
for field in dtype
]
)
if isinstance(dtype, native.BinaryType):
return dtypes.Binary()
return dtypes.Unknown() # pragma: no cover
@lru_cache(maxsize=4)
def fetch_session_time_zone(session: SparkSession) -> str:
# Timezone can't be changed in PySpark session, so this can be cached.
try:
return session.conf.get("spark.sql.session.timeZone") # type: ignore[attr-defined]
except Exception: # noqa: BLE001
# https://github.com/eakmanrq/sqlframe/issues/406
return "<unknown>"
def narwhals_to_native_dtype( # noqa: C901, PLR0912
dtype: IntoDType, version: Version, spark_types: ModuleType
) -> _NativeDType:
dtypes = version.dtypes
if TYPE_CHECKING:
native = sqlframe_types
else:
native = spark_types
if isinstance_or_issubclass(dtype, dtypes.Float64):
return native.DoubleType()
if isinstance_or_issubclass(dtype, dtypes.Float32):
return native.FloatType()
if isinstance_or_issubclass(dtype, dtypes.Int64):
return native.LongType()
if isinstance_or_issubclass(dtype, dtypes.Int32):
return native.IntegerType()
if isinstance_or_issubclass(dtype, dtypes.Int16):
return native.ShortType()
if isinstance_or_issubclass(dtype, dtypes.Int8):
return native.ByteType()
if isinstance_or_issubclass(dtype, dtypes.String):
return native.StringType()
if isinstance_or_issubclass(dtype, dtypes.Boolean):
return native.BooleanType()
if isinstance_or_issubclass(dtype, dtypes.Date):
return native.DateType()
if isinstance_or_issubclass(dtype, dtypes.Datetime):
dt_time_zone = dtype.time_zone
if dt_time_zone is None:
return native.TimestampNTZType()
if dt_time_zone != "UTC": # pragma: no cover
msg = f"Only UTC time zone is supported for PySpark, got: {dt_time_zone}"
raise ValueError(msg)
return native.TimestampType()
if isinstance_or_issubclass(dtype, (dtypes.List, dtypes.Array)):
return native.ArrayType(
elementType=narwhals_to_native_dtype(
dtype.inner, version=version, spark_types=native
)
)
if isinstance_or_issubclass(dtype, dtypes.Struct): # pragma: no cover
return native.StructType(
fields=[
native.StructField(
name=field.name,
dataType=narwhals_to_native_dtype(
field.dtype, version=version, spark_types=native
),
)
for field in dtype.fields
]
)
if isinstance_or_issubclass(dtype, dtypes.Binary):
return native.BinaryType()
if isinstance_or_issubclass(
dtype,
(
dtypes.UInt64,
dtypes.UInt32,
dtypes.UInt16,
dtypes.UInt8,
dtypes.Enum,
dtypes.Categorical,
dtypes.Time,
),
): # pragma: no cover
msg = "Unsigned integer, Enum, Categorical and Time types are not supported by spark-like backend"
raise UnsupportedDTypeError(msg)
msg = f"Unknown dtype: {dtype}" # pragma: no cover
raise AssertionError(msg)
def evaluate_exprs(
df: SparkLikeLazyFrame, /, *exprs: SparkLikeExpr
) -> list[tuple[str, Column]]:
native_results: list[tuple[str, Column]] = []
for expr in exprs:
native_series_list = expr._call(df)
output_names = expr._evaluate_output_names(df)
if expr._alias_output_names is not None:
output_names = expr._alias_output_names(output_names)
if len(output_names) != len(native_series_list): # pragma: no cover
msg = f"Internal error: got output names {output_names}, but only got {len(native_series_list)} results"
raise AssertionError(msg)
native_results.extend(zip(output_names, native_series_list))
return native_results
def import_functions(implementation: Implementation, /) -> ModuleType:
if implementation is Implementation.PYSPARK:
from pyspark.sql import functions
return functions
if implementation is Implementation.PYSPARK_CONNECT:
from pyspark.sql.connect import functions
return functions
from sqlframe.base.session import _BaseSession
return import_module(f"sqlframe.{_BaseSession().execution_dialect_name}.functions")
def import_native_dtypes(implementation: Implementation, /) -> ModuleType:
if implementation is Implementation.PYSPARK:
from pyspark.sql import types
return types
if implementation is Implementation.PYSPARK_CONNECT:
from pyspark.sql.connect import types
return types
from sqlframe.base.session import _BaseSession
return import_module(f"sqlframe.{_BaseSession().execution_dialect_name}.types")
def import_window(implementation: Implementation, /) -> type[Any]:
if implementation is Implementation.PYSPARK:
from pyspark.sql import Window
return Window
if implementation is Implementation.PYSPARK_CONNECT:
from pyspark.sql.connect.window import Window
return Window
from sqlframe.base.session import _BaseSession
return import_module(
f"sqlframe.{_BaseSession().execution_dialect_name}.window"
).Window
@overload
def strptime_to_pyspark_format(format: None) -> None: ...
@overload
def strptime_to_pyspark_format(format: str) -> str: ...
def strptime_to_pyspark_format(format: str | None) -> str | None:
"""Converts a Python strptime datetime format string to a PySpark datetime format string."""
if format is None: # pragma: no cover
return None
# Replace Python format specifiers with PySpark specifiers
pyspark_format = format
for py_format, spark_format in DATETIME_PATTERNS_MAPPING.items():
pyspark_format = pyspark_format.replace(py_format, spark_format)
return pyspark_format.replace("T", " ")
|
