datafusion_python/common/
data_type.rs

1// Licensed to the Apache Software Foundation (ASF) under one
2// or more contributor license agreements.  See the NOTICE file
3// distributed with this work for additional information
4// regarding copyright ownership.  The ASF licenses this file
5// to you under the Apache License, Version 2.0 (the
6// "License"); you may not use this file except in compliance
7// with the License.  You may obtain a copy of the License at
8//
9//   http://www.apache.org/licenses/LICENSE-2.0
10//
11// Unless required by applicable law or agreed to in writing,
12// software distributed under the License is distributed on an
13// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
14// KIND, either express or implied.  See the License for the
15// specific language governing permissions and limitations
16// under the License.
17
18use datafusion::arrow::array::Array;
19use datafusion::arrow::datatypes::{DataType, IntervalUnit, TimeUnit};
20use datafusion::common::ScalarValue;
21use datafusion::logical_expr::expr::NullTreatment as DFNullTreatment;
22use pyo3::exceptions::{PyNotImplementedError, PyValueError};
23use pyo3::prelude::*;
24
25#[derive(Debug, Clone, PartialEq, Eq, Hash, PartialOrd)]
26pub struct PyScalarValue(pub ScalarValue);
27
28impl From<ScalarValue> for PyScalarValue {
29    fn from(value: ScalarValue) -> Self {
30        Self(value)
31    }
32}
33impl From<PyScalarValue> for ScalarValue {
34    fn from(value: PyScalarValue) -> Self {
35        value.0
36    }
37}
38
39#[derive(Debug, Clone, PartialEq, Eq, Hash, PartialOrd, Ord)]
40#[pyclass(frozen, eq, eq_int, name = "RexType", module = "datafusion.common")]
41pub enum RexType {
42    Alias,
43    Literal,
44    Call,
45    Reference,
46    ScalarSubquery,
47    Other,
48}
49
50/// These bindings are tying together several disparate systems.
51/// You have SQL types for the SQL strings and RDBMS systems itself.
52/// Rust types for the DataFusion code
53/// Arrow types which represents the underlying arrow format
54/// Python types which represent the type in Python
55/// It is important to keep all of those types in a single
56/// and manageable location. Therefore this structure exists
57/// to map those types and provide a simple place for developers
58/// to map types from one system to another.
59// TODO: This looks like this needs pyo3 tracking so leaving unfrozen for now
60#[derive(Debug, Clone)]
61#[pyclass(name = "DataTypeMap", module = "datafusion.common", subclass)]
62pub struct DataTypeMap {
63    #[pyo3(get, set)]
64    pub arrow_type: PyDataType,
65    #[pyo3(get, set)]
66    pub python_type: PythonType,
67    #[pyo3(get, set)]
68    pub sql_type: SqlType,
69}
70
71impl DataTypeMap {
72    fn new(arrow_type: DataType, python_type: PythonType, sql_type: SqlType) -> Self {
73        DataTypeMap {
74            arrow_type: PyDataType {
75                data_type: arrow_type,
76            },
77            python_type,
78            sql_type,
79        }
80    }
81
82    pub fn map_from_arrow_type(arrow_type: &DataType) -> Result<DataTypeMap, PyErr> {
83        match arrow_type {
84            DataType::Null => Ok(DataTypeMap::new(
85                DataType::Null,
86                PythonType::None,
87                SqlType::NULL,
88            )),
89            DataType::Boolean => Ok(DataTypeMap::new(
90                DataType::Boolean,
91                PythonType::Bool,
92                SqlType::BOOLEAN,
93            )),
94            DataType::Int8 => Ok(DataTypeMap::new(
95                DataType::Int8,
96                PythonType::Int,
97                SqlType::TINYINT,
98            )),
99            DataType::Int16 => Ok(DataTypeMap::new(
100                DataType::Int16,
101                PythonType::Int,
102                SqlType::SMALLINT,
103            )),
104            DataType::Int32 => Ok(DataTypeMap::new(
105                DataType::Int32,
106                PythonType::Int,
107                SqlType::INTEGER,
108            )),
109            DataType::Int64 => Ok(DataTypeMap::new(
110                DataType::Int64,
111                PythonType::Int,
112                SqlType::BIGINT,
113            )),
114            DataType::UInt8 => Ok(DataTypeMap::new(
115                DataType::UInt8,
116                PythonType::Int,
117                SqlType::TINYINT,
118            )),
119            DataType::UInt16 => Ok(DataTypeMap::new(
120                DataType::UInt16,
121                PythonType::Int,
122                SqlType::SMALLINT,
123            )),
124            DataType::UInt32 => Ok(DataTypeMap::new(
125                DataType::UInt32,
126                PythonType::Int,
127                SqlType::INTEGER,
128            )),
129            DataType::UInt64 => Ok(DataTypeMap::new(
130                DataType::UInt64,
131                PythonType::Int,
132                SqlType::BIGINT,
133            )),
134            DataType::Float16 => Ok(DataTypeMap::new(
135                DataType::Float16,
136                PythonType::Float,
137                SqlType::FLOAT,
138            )),
139            DataType::Float32 => Ok(DataTypeMap::new(
140                DataType::Float32,
141                PythonType::Float,
142                SqlType::FLOAT,
143            )),
144            DataType::Float64 => Ok(DataTypeMap::new(
145                DataType::Float64,
146                PythonType::Float,
147                SqlType::FLOAT,
148            )),
149            DataType::Timestamp(unit, tz) => Ok(DataTypeMap::new(
150                DataType::Timestamp(*unit, tz.clone()),
151                PythonType::Datetime,
152                SqlType::DATE,
153            )),
154            DataType::Date32 => Ok(DataTypeMap::new(
155                DataType::Date32,
156                PythonType::Datetime,
157                SqlType::DATE,
158            )),
159            DataType::Date64 => Ok(DataTypeMap::new(
160                DataType::Date64,
161                PythonType::Datetime,
162                SqlType::DATE,
163            )),
164            DataType::Time32(unit) => Ok(DataTypeMap::new(
165                DataType::Time32(*unit),
166                PythonType::Datetime,
167                SqlType::DATE,
168            )),
169            DataType::Time64(unit) => Ok(DataTypeMap::new(
170                DataType::Time64(*unit),
171                PythonType::Datetime,
172                SqlType::DATE,
173            )),
174            DataType::Duration(_) => Err(PyNotImplementedError::new_err(format!("{arrow_type:?}"))),
175            DataType::Interval(interval_unit) => Ok(DataTypeMap::new(
176                DataType::Interval(*interval_unit),
177                PythonType::Datetime,
178                match interval_unit {
179                    IntervalUnit::DayTime => SqlType::INTERVAL_DAY,
180                    IntervalUnit::MonthDayNano => SqlType::INTERVAL_MONTH,
181                    IntervalUnit::YearMonth => SqlType::INTERVAL_YEAR_MONTH,
182                },
183            )),
184            DataType::Binary => Ok(DataTypeMap::new(
185                DataType::Binary,
186                PythonType::Bytes,
187                SqlType::BINARY,
188            )),
189            DataType::FixedSizeBinary(_) => {
190                Err(PyNotImplementedError::new_err(format!("{arrow_type:?}")))
191            }
192            DataType::LargeBinary => Ok(DataTypeMap::new(
193                DataType::LargeBinary,
194                PythonType::Bytes,
195                SqlType::BINARY,
196            )),
197            DataType::Utf8 => Ok(DataTypeMap::new(
198                DataType::Utf8,
199                PythonType::Str,
200                SqlType::VARCHAR,
201            )),
202            DataType::LargeUtf8 => Ok(DataTypeMap::new(
203                DataType::LargeUtf8,
204                PythonType::Str,
205                SqlType::VARCHAR,
206            )),
207            DataType::List(_) => Err(PyNotImplementedError::new_err(format!("{arrow_type:?}"))),
208            DataType::FixedSizeList(_, _) => {
209                Err(PyNotImplementedError::new_err(format!("{arrow_type:?}")))
210            }
211            DataType::LargeList(_) => {
212                Err(PyNotImplementedError::new_err(format!("{arrow_type:?}")))
213            }
214            DataType::Struct(_) => Err(PyNotImplementedError::new_err(format!("{arrow_type:?}"))),
215            DataType::Union(_, _) => Err(PyNotImplementedError::new_err(format!("{arrow_type:?}"))),
216            DataType::Dictionary(_, _) => {
217                Err(PyNotImplementedError::new_err(format!("{arrow_type:?}")))
218            }
219            DataType::Decimal32(precision, scale) => Ok(DataTypeMap::new(
220                DataType::Decimal32(*precision, *scale),
221                PythonType::Float,
222                SqlType::DECIMAL,
223            )),
224            DataType::Decimal64(precision, scale) => Ok(DataTypeMap::new(
225                DataType::Decimal64(*precision, *scale),
226                PythonType::Float,
227                SqlType::DECIMAL,
228            )),
229            DataType::Decimal128(precision, scale) => Ok(DataTypeMap::new(
230                DataType::Decimal128(*precision, *scale),
231                PythonType::Float,
232                SqlType::DECIMAL,
233            )),
234            DataType::Decimal256(precision, scale) => Ok(DataTypeMap::new(
235                DataType::Decimal256(*precision, *scale),
236                PythonType::Float,
237                SqlType::DECIMAL,
238            )),
239            DataType::Map(_, _) => Err(PyNotImplementedError::new_err(format!("{arrow_type:?}"))),
240            DataType::RunEndEncoded(_, _) => {
241                Err(PyNotImplementedError::new_err(format!("{arrow_type:?}")))
242            }
243            DataType::BinaryView => Err(PyNotImplementedError::new_err(format!("{arrow_type:?}"))),
244            DataType::Utf8View => Err(PyNotImplementedError::new_err(format!("{arrow_type:?}"))),
245            DataType::ListView(_) => Err(PyNotImplementedError::new_err(format!("{arrow_type:?}"))),
246            DataType::LargeListView(_) => {
247                Err(PyNotImplementedError::new_err(format!("{arrow_type:?}")))
248            }
249        }
250    }
251
252    /// Generate the `DataTypeMap` from a `ScalarValue` instance
253    pub fn map_from_scalar_value(scalar_val: &ScalarValue) -> Result<DataTypeMap, PyErr> {
254        DataTypeMap::map_from_arrow_type(&DataTypeMap::map_from_scalar_to_arrow(scalar_val)?)
255    }
256
257    /// Maps a `ScalarValue` to an Arrow `DataType`
258    pub fn map_from_scalar_to_arrow(scalar_val: &ScalarValue) -> Result<DataType, PyErr> {
259        match scalar_val {
260            ScalarValue::Boolean(_) => Ok(DataType::Boolean),
261            ScalarValue::Float16(_) => Ok(DataType::Float16),
262            ScalarValue::Float32(_) => Ok(DataType::Float32),
263            ScalarValue::Float64(_) => Ok(DataType::Float64),
264            ScalarValue::Decimal32(_, precision, scale) => {
265                Ok(DataType::Decimal32(*precision, *scale))
266            }
267            ScalarValue::Decimal64(_, precision, scale) => {
268                Ok(DataType::Decimal64(*precision, *scale))
269            }
270            ScalarValue::Decimal128(_, precision, scale) => {
271                Ok(DataType::Decimal128(*precision, *scale))
272            }
273            ScalarValue::Decimal256(_, precision, scale) => {
274                Ok(DataType::Decimal256(*precision, *scale))
275            }
276            ScalarValue::Dictionary(data_type, scalar_type) => {
277                // Call this function again to map the dictionary scalar_value to an Arrow type
278                Ok(DataType::Dictionary(
279                    Box::new(*data_type.clone()),
280                    Box::new(DataTypeMap::map_from_scalar_to_arrow(scalar_type)?),
281                ))
282            }
283            ScalarValue::Int8(_) => Ok(DataType::Int8),
284            ScalarValue::Int16(_) => Ok(DataType::Int16),
285            ScalarValue::Int32(_) => Ok(DataType::Int32),
286            ScalarValue::Int64(_) => Ok(DataType::Int64),
287            ScalarValue::UInt8(_) => Ok(DataType::UInt8),
288            ScalarValue::UInt16(_) => Ok(DataType::UInt16),
289            ScalarValue::UInt32(_) => Ok(DataType::UInt32),
290            ScalarValue::UInt64(_) => Ok(DataType::UInt64),
291            ScalarValue::Utf8(_) => Ok(DataType::Utf8),
292            ScalarValue::LargeUtf8(_) => Ok(DataType::LargeUtf8),
293            ScalarValue::Binary(_) => Ok(DataType::Binary),
294            ScalarValue::LargeBinary(_) => Ok(DataType::LargeBinary),
295            ScalarValue::Date32(_) => Ok(DataType::Date32),
296            ScalarValue::Date64(_) => Ok(DataType::Date64),
297            ScalarValue::Time32Second(_) => Ok(DataType::Time32(TimeUnit::Second)),
298            ScalarValue::Time32Millisecond(_) => Ok(DataType::Time32(TimeUnit::Millisecond)),
299            ScalarValue::Time64Microsecond(_) => Ok(DataType::Time64(TimeUnit::Microsecond)),
300            ScalarValue::Time64Nanosecond(_) => Ok(DataType::Time64(TimeUnit::Nanosecond)),
301            ScalarValue::Null => Ok(DataType::Null),
302            ScalarValue::TimestampSecond(_, tz) => {
303                Ok(DataType::Timestamp(TimeUnit::Second, tz.to_owned()))
304            }
305            ScalarValue::TimestampMillisecond(_, tz) => {
306                Ok(DataType::Timestamp(TimeUnit::Millisecond, tz.to_owned()))
307            }
308            ScalarValue::TimestampMicrosecond(_, tz) => {
309                Ok(DataType::Timestamp(TimeUnit::Microsecond, tz.to_owned()))
310            }
311            ScalarValue::TimestampNanosecond(_, tz) => {
312                Ok(DataType::Timestamp(TimeUnit::Nanosecond, tz.to_owned()))
313            }
314            ScalarValue::IntervalYearMonth(..) => Ok(DataType::Interval(IntervalUnit::YearMonth)),
315            ScalarValue::IntervalDayTime(..) => Ok(DataType::Interval(IntervalUnit::DayTime)),
316            ScalarValue::IntervalMonthDayNano(..) => {
317                Ok(DataType::Interval(IntervalUnit::MonthDayNano))
318            }
319            ScalarValue::List(arr) => Ok(arr.data_type().to_owned()),
320            ScalarValue::Struct(_fields) => Err(PyNotImplementedError::new_err(
321                "ScalarValue::Struct".to_string(),
322            )),
323            ScalarValue::FixedSizeBinary(size, _) => Ok(DataType::FixedSizeBinary(*size)),
324            ScalarValue::FixedSizeList(_array_ref) => {
325                // The FieldRef was removed from ScalarValue::FixedSizeList in
326                // https://github.com/apache/arrow-datafusion/pull/8221, so we can no
327                // longer convert back to a DataType here
328                Err(PyNotImplementedError::new_err(
329                    "ScalarValue::FixedSizeList".to_string(),
330                ))
331            }
332            ScalarValue::LargeList(_) => Err(PyNotImplementedError::new_err(
333                "ScalarValue::LargeList".to_string(),
334            )),
335            ScalarValue::DurationSecond(_) => Ok(DataType::Duration(TimeUnit::Second)),
336            ScalarValue::DurationMillisecond(_) => Ok(DataType::Duration(TimeUnit::Millisecond)),
337            ScalarValue::DurationMicrosecond(_) => Ok(DataType::Duration(TimeUnit::Microsecond)),
338            ScalarValue::DurationNanosecond(_) => Ok(DataType::Duration(TimeUnit::Nanosecond)),
339            ScalarValue::Union(_, _, _) => Err(PyNotImplementedError::new_err(
340                "ScalarValue::LargeList".to_string(),
341            )),
342            ScalarValue::Utf8View(_) => Ok(DataType::Utf8View),
343            ScalarValue::BinaryView(_) => Ok(DataType::BinaryView),
344            ScalarValue::Map(_) => Err(PyNotImplementedError::new_err(
345                "ScalarValue::Map".to_string(),
346            )),
347        }
348    }
349}
350
351#[pymethods]
352impl DataTypeMap {
353    #[new]
354    pub fn py_new(arrow_type: PyDataType, python_type: PythonType, sql_type: SqlType) -> Self {
355        DataTypeMap {
356            arrow_type,
357            python_type,
358            sql_type,
359        }
360    }
361
362    #[staticmethod]
363    #[pyo3(name = "from_parquet_type_str")]
364    /// When using pyarrow.parquet.read_metadata().schema.column(x).physical_type you are presented
365    /// with a String type for schema rather than an object type. Here we make a best effort
366    /// to convert that to a physical type.
367    pub fn py_map_from_parquet_type_str(parquet_str_type: String) -> PyResult<DataTypeMap> {
368        let arrow_dtype = match parquet_str_type.to_lowercase().as_str() {
369            "boolean" => Ok(DataType::Boolean),
370            "int32" => Ok(DataType::Int32),
371            "int64" => Ok(DataType::Int64),
372            "int96" => {
373                // Int96 is an old parquet datatype that is now deprecated. We convert to nanosecond timestamp
374                Ok(DataType::Timestamp(TimeUnit::Nanosecond, None))
375            }
376            "float" => Ok(DataType::Float32),
377            "double" => Ok(DataType::Float64),
378            "byte_array" => Ok(DataType::Utf8),
379            _ => Err(PyValueError::new_err(format!(
380                "Unable to determine Arrow Data Type from Parquet String type: {parquet_str_type:?}"
381            ))),
382        };
383        DataTypeMap::map_from_arrow_type(&arrow_dtype?)
384    }
385
386    #[staticmethod]
387    #[pyo3(name = "arrow")]
388    pub fn py_map_from_arrow_type(arrow_type: &PyDataType) -> PyResult<DataTypeMap> {
389        DataTypeMap::map_from_arrow_type(&arrow_type.data_type)
390    }
391
392    #[staticmethod]
393    #[pyo3(name = "arrow_str")]
394    pub fn py_map_from_arrow_type_str(arrow_type_str: String) -> PyResult<DataTypeMap> {
395        let data_type = PyDataType::py_map_from_arrow_type_str(arrow_type_str);
396        DataTypeMap::map_from_arrow_type(&data_type?.data_type)
397    }
398
399    #[staticmethod]
400    #[pyo3(name = "sql")]
401    pub fn py_map_from_sql_type(sql_type: &SqlType) -> PyResult<DataTypeMap> {
402        match sql_type {
403            SqlType::ANY => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
404            SqlType::ARRAY => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
405            SqlType::BIGINT => Ok(DataTypeMap::new(
406                DataType::Int64,
407                PythonType::Int,
408                SqlType::BIGINT,
409            )),
410            SqlType::BINARY => Ok(DataTypeMap::new(
411                DataType::Binary,
412                PythonType::Bytes,
413                SqlType::BINARY,
414            )),
415            SqlType::BOOLEAN => Ok(DataTypeMap::new(
416                DataType::Boolean,
417                PythonType::Bool,
418                SqlType::BOOLEAN,
419            )),
420            SqlType::CHAR => Ok(DataTypeMap::new(
421                DataType::UInt8,
422                PythonType::Int,
423                SqlType::CHAR,
424            )),
425            SqlType::COLUMN_LIST => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
426            SqlType::CURSOR => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
427            SqlType::DATE => Ok(DataTypeMap::new(
428                DataType::Date64,
429                PythonType::Datetime,
430                SqlType::DATE,
431            )),
432            SqlType::DECIMAL => Ok(DataTypeMap::new(
433                DataType::Decimal128(1, 1),
434                PythonType::Float,
435                SqlType::DECIMAL,
436            )),
437            SqlType::DISTINCT => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
438            SqlType::DOUBLE => Ok(DataTypeMap::new(
439                DataType::Decimal256(1, 1),
440                PythonType::Float,
441                SqlType::DOUBLE,
442            )),
443            SqlType::DYNAMIC_STAR => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
444            SqlType::FLOAT => Ok(DataTypeMap::new(
445                DataType::Decimal128(1, 1),
446                PythonType::Float,
447                SqlType::FLOAT,
448            )),
449            SqlType::GEOMETRY => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
450            SqlType::INTEGER => Ok(DataTypeMap::new(
451                DataType::Int8,
452                PythonType::Int,
453                SqlType::INTEGER,
454            )),
455            SqlType::INTERVAL => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
456            SqlType::INTERVAL_DAY => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
457            SqlType::INTERVAL_DAY_HOUR => {
458                Err(PyNotImplementedError::new_err(format!("{sql_type:?}")))
459            }
460            SqlType::INTERVAL_DAY_MINUTE => {
461                Err(PyNotImplementedError::new_err(format!("{sql_type:?}")))
462            }
463            SqlType::INTERVAL_DAY_SECOND => {
464                Err(PyNotImplementedError::new_err(format!("{sql_type:?}")))
465            }
466            SqlType::INTERVAL_HOUR => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
467            SqlType::INTERVAL_HOUR_MINUTE => {
468                Err(PyNotImplementedError::new_err(format!("{sql_type:?}")))
469            }
470            SqlType::INTERVAL_HOUR_SECOND => {
471                Err(PyNotImplementedError::new_err(format!("{sql_type:?}")))
472            }
473            SqlType::INTERVAL_MINUTE => {
474                Err(PyNotImplementedError::new_err(format!("{sql_type:?}")))
475            }
476            SqlType::INTERVAL_MINUTE_SECOND => {
477                Err(PyNotImplementedError::new_err(format!("{sql_type:?}")))
478            }
479            SqlType::INTERVAL_MONTH => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
480            SqlType::INTERVAL_SECOND => {
481                Err(PyNotImplementedError::new_err(format!("{sql_type:?}")))
482            }
483            SqlType::INTERVAL_YEAR => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
484            SqlType::INTERVAL_YEAR_MONTH => {
485                Err(PyNotImplementedError::new_err(format!("{sql_type:?}")))
486            }
487            SqlType::MAP => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
488            SqlType::MULTISET => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
489            SqlType::NULL => Ok(DataTypeMap::new(
490                DataType::Null,
491                PythonType::None,
492                SqlType::NULL,
493            )),
494            SqlType::OTHER => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
495            SqlType::REAL => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
496            SqlType::ROW => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
497            SqlType::SARG => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
498            SqlType::SMALLINT => Ok(DataTypeMap::new(
499                DataType::Int16,
500                PythonType::Int,
501                SqlType::SMALLINT,
502            )),
503            SqlType::STRUCTURED => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
504            SqlType::SYMBOL => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
505            SqlType::TIME => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
506            SqlType::TIME_WITH_LOCAL_TIME_ZONE => {
507                Err(PyNotImplementedError::new_err(format!("{sql_type:?}")))
508            }
509            SqlType::TIMESTAMP => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
510            SqlType::TIMESTAMP_WITH_LOCAL_TIME_ZONE => {
511                Err(PyNotImplementedError::new_err(format!("{sql_type:?}")))
512            }
513            SqlType::TINYINT => Ok(DataTypeMap::new(
514                DataType::Int8,
515                PythonType::Int,
516                SqlType::TINYINT,
517            )),
518            SqlType::UNKNOWN => Err(PyNotImplementedError::new_err(format!("{sql_type:?}"))),
519            SqlType::VARBINARY => Ok(DataTypeMap::new(
520                DataType::LargeBinary,
521                PythonType::Bytes,
522                SqlType::VARBINARY,
523            )),
524            SqlType::VARCHAR => Ok(DataTypeMap::new(
525                DataType::Utf8,
526                PythonType::Str,
527                SqlType::VARCHAR,
528            )),
529        }
530    }
531
532    /// Unfortunately PyO3 does not allow for us to expose the DataType as an enum since
533    /// we cannot directly annotate the Enum instance of dependency code. Therefore, here
534    /// we provide an enum to mimic it.
535    #[pyo3(name = "friendly_arrow_type_name")]
536    pub fn friendly_arrow_type_name(&self) -> PyResult<&str> {
537        Ok(match &self.arrow_type.data_type {
538            DataType::Null => "Null",
539            DataType::Boolean => "Boolean",
540            DataType::Int8 => "Int8",
541            DataType::Int16 => "Int16",
542            DataType::Int32 => "Int32",
543            DataType::Int64 => "Int64",
544            DataType::UInt8 => "UInt8",
545            DataType::UInt16 => "UInt16",
546            DataType::UInt32 => "UInt32",
547            DataType::UInt64 => "UInt64",
548            DataType::Float16 => "Float16",
549            DataType::Float32 => "Float32",
550            DataType::Float64 => "Float64",
551            DataType::Timestamp(_, _) => "Timestamp",
552            DataType::Date32 => "Date32",
553            DataType::Date64 => "Date64",
554            DataType::Time32(_) => "Time32",
555            DataType::Time64(_) => "Time64",
556            DataType::Duration(_) => "Duration",
557            DataType::Interval(_) => "Interval",
558            DataType::Binary => "Binary",
559            DataType::FixedSizeBinary(_) => "FixedSizeBinary",
560            DataType::LargeBinary => "LargeBinary",
561            DataType::Utf8 => "Utf8",
562            DataType::LargeUtf8 => "LargeUtf8",
563            DataType::List(_) => "List",
564            DataType::FixedSizeList(_, _) => "FixedSizeList",
565            DataType::LargeList(_) => "LargeList",
566            DataType::Struct(_) => "Struct",
567            DataType::Union(_, _) => "Union",
568            DataType::Dictionary(_, _) => "Dictionary",
569            DataType::Decimal32(_, _) => "Decimal32",
570            DataType::Decimal64(_, _) => "Decimal64",
571            DataType::Decimal128(_, _) => "Decimal128",
572            DataType::Decimal256(_, _) => "Decimal256",
573            DataType::Map(_, _) => "Map",
574            DataType::RunEndEncoded(_, _) => "RunEndEncoded",
575            DataType::BinaryView => "BinaryView",
576            DataType::Utf8View => "Utf8View",
577            DataType::ListView(_) => "ListView",
578            DataType::LargeListView(_) => "LargeListView",
579        })
580    }
581}
582
583/// PyO3 requires that objects passed between Rust and Python implement the trait `PyClass`
584/// Since `DataType` exists in another package we cannot make that happen here so we wrap
585/// `DataType` as `PyDataType` This exists solely to satisfy those constraints.
586#[derive(Debug, Clone, PartialEq, Eq, Hash, PartialOrd, Ord)]
587#[pyclass(frozen, name = "DataType", module = "datafusion.common")]
588pub struct PyDataType {
589    pub data_type: DataType,
590}
591
592impl PyDataType {
593    /// There are situations when obtaining dtypes on the Python side where the Arrow type
594    /// is presented as a String rather than an actual DataType. This function is used to
595    /// convert that String to a DataType for the Python side to use.
596    pub fn py_map_from_arrow_type_str(arrow_str_type: String) -> PyResult<PyDataType> {
597        // Certain string types contain "metadata" that should be trimmed here. Ex: "datetime64[ns, Europe/Berlin]"
598        let arrow_str_type = match arrow_str_type.find('[') {
599            Some(index) => arrow_str_type[0..index].to_string(),
600            None => arrow_str_type, // Return early if ',' is not found.
601        };
602
603        let arrow_dtype = match arrow_str_type.to_lowercase().as_str() {
604            "bool" => Ok(DataType::Boolean),
605            "boolean" => Ok(DataType::Boolean),
606            "uint8" => Ok(DataType::UInt8),
607            "uint16" => Ok(DataType::UInt16),
608            "uint32" => Ok(DataType::UInt32),
609            "uint64" => Ok(DataType::UInt64),
610            "int8" => Ok(DataType::Int8),
611            "int16" => Ok(DataType::Int16),
612            "int32" => Ok(DataType::Int32),
613            "int64" => Ok(DataType::Int64),
614            "float" => Ok(DataType::Float32),
615            "double" => Ok(DataType::Float64),
616            "float16" => Ok(DataType::Float16),
617            "float32" => Ok(DataType::Float32),
618            "float64" => Ok(DataType::Float64),
619            "datetime64" => Ok(DataType::Date64),
620            "object" => Ok(DataType::Utf8),
621            _ => Err(PyValueError::new_err(format!(
622                "Unable to determine Arrow Data Type from Arrow String type: {arrow_str_type:?}"
623            ))),
624        };
625        Ok(PyDataType {
626            data_type: arrow_dtype?,
627        })
628    }
629}
630
631impl From<PyDataType> for DataType {
632    fn from(data_type: PyDataType) -> DataType {
633        data_type.data_type
634    }
635}
636
637impl From<DataType> for PyDataType {
638    fn from(data_type: DataType) -> PyDataType {
639        PyDataType { data_type }
640    }
641}
642
643/// Represents the possible Python types that can be mapped to the SQL types
644#[derive(Debug, Clone, PartialEq, Eq, Hash, PartialOrd, Ord)]
645#[pyclass(frozen, eq, eq_int, name = "PythonType", module = "datafusion.common")]
646pub enum PythonType {
647    Array,
648    Bool,
649    Bytes,
650    Datetime,
651    Float,
652    Int,
653    List,
654    None,
655    Object,
656    Str,
657}
658
659/// Represents the types that are possible for DataFusion to parse
660/// from a SQL query. Aka "SqlType" and are valid values for
661/// ANSI SQL
662#[allow(non_camel_case_types)]
663#[allow(clippy::upper_case_acronyms)]
664#[derive(Debug, Clone, PartialEq, Eq, Hash, PartialOrd, Ord)]
665#[pyclass(frozen, eq, eq_int, name = "SqlType", module = "datafusion.common")]
666pub enum SqlType {
667    ANY,
668    ARRAY,
669    BIGINT,
670    BINARY,
671    BOOLEAN,
672    CHAR,
673    COLUMN_LIST,
674    CURSOR,
675    DATE,
676    DECIMAL,
677    DISTINCT,
678    DOUBLE,
679    DYNAMIC_STAR,
680    FLOAT,
681    GEOMETRY,
682    INTEGER,
683    INTERVAL,
684    INTERVAL_DAY,
685    INTERVAL_DAY_HOUR,
686    INTERVAL_DAY_MINUTE,
687    INTERVAL_DAY_SECOND,
688    INTERVAL_HOUR,
689    INTERVAL_HOUR_MINUTE,
690    INTERVAL_HOUR_SECOND,
691    INTERVAL_MINUTE,
692    INTERVAL_MINUTE_SECOND,
693    INTERVAL_MONTH,
694    INTERVAL_SECOND,
695    INTERVAL_YEAR,
696    INTERVAL_YEAR_MONTH,
697    MAP,
698    MULTISET,
699    NULL,
700    OTHER,
701    REAL,
702    ROW,
703    SARG,
704    SMALLINT,
705    STRUCTURED,
706    SYMBOL,
707    TIME,
708    TIME_WITH_LOCAL_TIME_ZONE,
709    TIMESTAMP,
710    TIMESTAMP_WITH_LOCAL_TIME_ZONE,
711    TINYINT,
712    UNKNOWN,
713    VARBINARY,
714    VARCHAR,
715}
716
717/// Specifies Ignore / Respect NULL within window functions.
718/// For example
719/// `FIRST_VALUE(column2) IGNORE NULLS OVER (PARTITION BY column1)`
720#[allow(non_camel_case_types)]
721#[allow(clippy::upper_case_acronyms)]
722#[derive(Debug, Clone, PartialEq, Eq, Hash, PartialOrd, Ord)]
723#[pyclass(
724    frozen,
725    eq,
726    eq_int,
727    name = "NullTreatment",
728    module = "datafusion.common"
729)]
730pub enum NullTreatment {
731    IGNORE_NULLS,
732    RESPECT_NULLS,
733}
734
735impl From<NullTreatment> for DFNullTreatment {
736    fn from(null_treatment: NullTreatment) -> DFNullTreatment {
737        match null_treatment {
738            NullTreatment::IGNORE_NULLS => DFNullTreatment::IgnoreNulls,
739            NullTreatment::RESPECT_NULLS => DFNullTreatment::RespectNulls,
740        }
741    }
742}
743
744impl From<DFNullTreatment> for NullTreatment {
745    fn from(null_treatment: DFNullTreatment) -> NullTreatment {
746        match null_treatment {
747            DFNullTreatment::IgnoreNulls => NullTreatment::IGNORE_NULLS,
748            DFNullTreatment::RespectNulls => NullTreatment::RESPECT_NULLS,
749        }
750    }
751}
datafusion_python/common/data_type.rs

datafusion_python/common/
data_type.rs
