polars_python/interop/numpy/
to_numpy_series.rs

1use ndarray::IntoDimension;
2use num_traits::{Float, NumCast};
3use numpy::npyffi::flags;
4use numpy::{Element, PyArray1};
5use polars_core::prelude::*;
6use pyo3::exceptions::PyRuntimeError;
7use pyo3::prelude::*;
8use pyo3::{intern, IntoPyObjectExt};
9
10use super::to_numpy_df::df_to_numpy;
11use super::utils::{
12    create_borrowed_np_array, dtype_supports_view, polars_dtype_to_np_temporal_dtype,
13    reshape_numpy_array, series_contains_null,
14};
15use crate::conversion::chunked_array::{decimal_to_pyobject_iter, time_to_pyobject_iter};
16use crate::conversion::ObjectValue;
17use crate::series::PySeries;
18
19#[pymethods]
20impl PySeries {
21    /// Convert this Series to a NumPy ndarray.
22    ///
23    /// This method copies data only when necessary. Set `allow_copy` to raise an error if copy
24    /// is required. Set `writable` to make sure the resulting array is writable, possibly requiring
25    /// copying the data.
26    fn to_numpy(&self, py: Python, writable: bool, allow_copy: bool) -> PyResult<PyObject> {
27        series_to_numpy(py, &self.series, writable, allow_copy)
28    }
29
30    /// Create a view of the data as a NumPy ndarray.
31    ///
32    /// WARNING: The resulting view will show the underlying value for nulls,
33    /// which may be any value. The caller is responsible for handling nulls
34    /// appropriately.
35    fn to_numpy_view(&self, py: Python) -> Option<PyObject> {
36        let (view, _) = try_series_to_numpy_view(py, &self.series, true, false)?;
37        Some(view)
38    }
39}
40
41/// Convert a Series to a NumPy ndarray.
42pub(super) fn series_to_numpy(
43    py: Python,
44    s: &Series,
45    writable: bool,
46    allow_copy: bool,
47) -> PyResult<PyObject> {
48    if s.is_empty() {
49        // Take this path to ensure a writable array.
50        // This does not actually copy data for an empty Series.
51        return Ok(series_to_numpy_with_copy(py, s, true));
52    }
53    if let Some((mut arr, writable_flag)) = try_series_to_numpy_view(py, s, false, allow_copy) {
54        if writable && !writable_flag {
55            if !allow_copy {
56                return Err(PyRuntimeError::new_err(
57                    "copy not allowed: cannot create a writable array without copying data",
58                ));
59            }
60            arr = arr.call_method0(py, intern!(py, "copy"))?;
61        }
62        return Ok(arr);
63    }
64
65    if !allow_copy {
66        return Err(PyRuntimeError::new_err(
67            "copy not allowed: cannot convert to a NumPy array without copying data",
68        ));
69    }
70
71    Ok(series_to_numpy_with_copy(py, s, writable))
72}
73
74/// Create a NumPy view of the given Series.
75fn try_series_to_numpy_view(
76    py: Python,
77    s: &Series,
78    allow_nulls: bool,
79    allow_rechunk: bool,
80) -> Option<(PyObject, bool)> {
81    if !dtype_supports_view(s.dtype()) {
82        return None;
83    }
84    if !allow_nulls && series_contains_null(s) {
85        return None;
86    }
87    let (s_owned, writable_flag) = handle_chunks(py, s, allow_rechunk)?;
88
89    let array = series_to_numpy_view_recursive(py, s_owned, writable_flag);
90    Some((array, writable_flag))
91}
92
93/// Rechunk the Series if required.
94///
95/// NumPy arrays are always contiguous, so we may have to rechunk before creating a view.
96/// If we do so, we can flag the resulting array as writable.
97fn handle_chunks(py: Python, s: &Series, allow_rechunk: bool) -> Option<(Series, bool)> {
98    let is_chunked = s.n_chunks() > 1;
99    match (is_chunked, allow_rechunk) {
100        (true, false) => None,
101        (true, true) => Some((py.allow_threads(|| s.rechunk()), true)),
102        (false, _) => Some((s.clone(), false)),
103    }
104}
105
106/// Create a NumPy view of the given Series without checking for data types, chunks, or nulls.
107fn series_to_numpy_view_recursive(py: Python, s: Series, writable: bool) -> PyObject {
108    debug_assert!(s.n_chunks() == 1);
109    match s.dtype() {
110        dt if dt.is_primitive_numeric() => numeric_series_to_numpy_view(py, s, writable),
111        DataType::Datetime(_, _) | DataType::Duration(_) => {
112            temporal_series_to_numpy_view(py, s, writable)
113        },
114        DataType::Array(_, _) => array_series_to_numpy_view(py, &s, writable),
115        _ => panic!("invalid data type"),
116    }
117}
118/// Create a NumPy view of a numeric Series.
119fn numeric_series_to_numpy_view(py: Python, s: Series, writable: bool) -> PyObject {
120    let dims = [s.len()].into_dimension();
121    with_match_physical_numpy_polars_type!(s.dtype(), |$T| {
122        let np_dtype = <$T as PolarsNumericType>::Native::get_dtype(py);
123        let ca: &ChunkedArray<$T> = s.unpack::<$T>().unwrap();
124        let flags = if writable {
125            flags::NPY_ARRAY_FARRAY
126        } else {
127            flags::NPY_ARRAY_FARRAY_RO
128        };
129
130        let slice = ca.data_views().next().unwrap();
131
132        unsafe {
133            create_borrowed_np_array::<_>(
134                py,
135                np_dtype,
136                dims,
137                flags,
138                slice.as_ptr() as _,
139                PySeries::from(s).into_py_any(py).unwrap(), // Keep the Series memory alive.,
140            )
141        }
142    })
143}
144/// Create a NumPy view of a Datetime or Duration Series.
145fn temporal_series_to_numpy_view(py: Python, s: Series, writable: bool) -> PyObject {
146    let np_dtype = polars_dtype_to_np_temporal_dtype(py, s.dtype());
147
148    let phys = s.to_physical_repr();
149    let ca = phys.i64().unwrap();
150    let slice = ca.data_views().next().unwrap();
151    let dims = [s.len()].into_dimension();
152    let flags = if writable {
153        flags::NPY_ARRAY_FARRAY
154    } else {
155        flags::NPY_ARRAY_FARRAY_RO
156    };
157
158    unsafe {
159        create_borrowed_np_array::<_>(
160            py,
161            np_dtype,
162            dims,
163            flags,
164            slice.as_ptr() as _,
165            PySeries::from(s).into_py_any(py).unwrap(), // Keep the Series memory alive.,
166        )
167    }
168}
169/// Create a NumPy view of an Array Series.
170fn array_series_to_numpy_view(py: Python, s: &Series, writable: bool) -> PyObject {
171    let ca = s.array().unwrap();
172    let s_inner = ca.get_inner();
173    let np_array_flat = series_to_numpy_view_recursive(py, s_inner, writable);
174
175    // Reshape to the original shape.
176    let DataType::Array(_, width) = s.dtype() else {
177        unreachable!()
178    };
179    reshape_numpy_array(py, np_array_flat, ca.len(), *width).unwrap()
180}
181
182/// Convert a Series to a NumPy ndarray, copying data in the process.
183///
184/// This method will cast integers to floats so that `null = np.nan`.
185fn series_to_numpy_with_copy(py: Python, s: &Series, writable: bool) -> PyObject {
186    use DataType::*;
187    match s.dtype() {
188        Int8 => numeric_series_to_numpy::<Int8Type, f32>(py, s),
189        Int16 => numeric_series_to_numpy::<Int16Type, f32>(py, s),
190        Int32 => numeric_series_to_numpy::<Int32Type, f64>(py, s),
191        Int64 => numeric_series_to_numpy::<Int64Type, f64>(py, s),
192        Int128 => {
193            let s = s.cast(&DataType::Float64).unwrap();
194            series_to_numpy(py, &s, writable, true).unwrap()
195        },
196        UInt8 => numeric_series_to_numpy::<UInt8Type, f32>(py, s),
197        UInt16 => numeric_series_to_numpy::<UInt16Type, f32>(py, s),
198        UInt32 => numeric_series_to_numpy::<UInt32Type, f64>(py, s),
199        UInt64 => numeric_series_to_numpy::<UInt64Type, f64>(py, s),
200        Float32 => numeric_series_to_numpy::<Float32Type, f32>(py, s),
201        Float64 => numeric_series_to_numpy::<Float64Type, f64>(py, s),
202        Boolean => boolean_series_to_numpy(py, s),
203        Date => date_series_to_numpy(py, s),
204        Datetime(tu, _) => {
205            use numpy::datetime::{units, Datetime};
206            match tu {
207                TimeUnit::Milliseconds => {
208                    temporal_series_to_numpy::<Datetime<units::Milliseconds>>(py, s)
209                },
210                TimeUnit::Microseconds => {
211                    temporal_series_to_numpy::<Datetime<units::Microseconds>>(py, s)
212                },
213                TimeUnit::Nanoseconds => {
214                    temporal_series_to_numpy::<Datetime<units::Nanoseconds>>(py, s)
215                },
216            }
217        },
218        Duration(tu) => {
219            use numpy::datetime::{units, Timedelta};
220            match tu {
221                TimeUnit::Milliseconds => {
222                    temporal_series_to_numpy::<Timedelta<units::Milliseconds>>(py, s)
223                },
224                TimeUnit::Microseconds => {
225                    temporal_series_to_numpy::<Timedelta<units::Microseconds>>(py, s)
226                },
227                TimeUnit::Nanoseconds => {
228                    temporal_series_to_numpy::<Timedelta<units::Nanoseconds>>(py, s)
229                },
230            }
231        },
232        Time => {
233            let ca = s.time().unwrap();
234            let values = time_to_pyobject_iter(ca).map(|v| v.into_py_any(py).unwrap());
235            PyArray1::from_iter(py, values).into_py_any(py).unwrap()
236        },
237        String => {
238            let ca = s.str().unwrap();
239            let values = ca.iter().map(|s| s.into_py_any(py).unwrap());
240            PyArray1::from_iter(py, values).into_py_any(py).unwrap()
241        },
242        Binary => {
243            let ca = s.binary().unwrap();
244            let values = ca.iter().map(|s| s.into_py_any(py).unwrap());
245            PyArray1::from_iter(py, values).into_py_any(py).unwrap()
246        },
247        Categorical(_, _) | Enum(_, _) => {
248            let ca = s.categorical().unwrap();
249            let values = ca.iter_str().map(|s| s.into_py_any(py).unwrap());
250            PyArray1::from_iter(py, values).into_py_any(py).unwrap()
251        },
252        Decimal(_, _) => {
253            let ca = s.decimal().unwrap();
254            let values = decimal_to_pyobject_iter(py, ca)
255                .unwrap()
256                .map(|v| v.into_py_any(py).unwrap());
257            PyArray1::from_iter(py, values).into_py_any(py).unwrap()
258        },
259        List(_) => list_series_to_numpy(py, s, writable),
260        Array(_, _) => array_series_to_numpy(py, s, writable),
261        Struct(_) => {
262            let ca = s.struct_().unwrap();
263            let df = ca.clone().unnest();
264            df_to_numpy(py, &df, IndexOrder::Fortran, writable, true).unwrap()
265        },
266        #[cfg(feature = "object")]
267        Object(_, _) => {
268            let ca = s
269                .as_any()
270                .downcast_ref::<ObjectChunked<ObjectValue>>()
271                .unwrap();
272            let values = ca.iter().map(|v| v.into_py_any(py).unwrap());
273            PyArray1::from_iter(py, values).into_py_any(py).unwrap()
274        },
275        Null => {
276            let n = s.len();
277            let values = std::iter::repeat(f32::NAN).take(n);
278            PyArray1::from_iter(py, values).into_py_any(py).unwrap()
279        },
280        Unknown(_) | BinaryOffset => unreachable!(),
281    }
282}
283
284/// Convert numeric types to f32 or f64 with NaN representing a null value.
285fn numeric_series_to_numpy<T, U>(py: Python, s: &Series) -> PyObject
286where
287    T: PolarsNumericType,
288    T::Native: numpy::Element,
289    U: Float + numpy::Element,
290{
291    let ca: &ChunkedArray<T> = s.as_ref().as_ref();
292    if s.null_count() == 0 {
293        let values = ca.into_no_null_iter();
294        PyArray1::<T::Native>::from_iter(py, values)
295            .into_py_any(py)
296            .unwrap()
297    } else {
298        let mapper = |opt_v: Option<T::Native>| match opt_v {
299            Some(v) => NumCast::from(v).unwrap(),
300            None => U::nan(),
301        };
302        let values = ca.iter().map(mapper);
303        PyArray1::from_iter(py, values).into_py_any(py).unwrap()
304    }
305}
306/// Convert booleans to u8 if no nulls are present, otherwise convert to objects.
307fn boolean_series_to_numpy(py: Python, s: &Series) -> PyObject {
308    let ca = s.bool().unwrap();
309    if s.null_count() == 0 {
310        let values = ca.into_no_null_iter();
311        PyArray1::<bool>::from_iter(py, values)
312            .into_py_any(py)
313            .unwrap()
314    } else {
315        let values = ca.iter().map(|opt_v| opt_v.into_py_any(py).unwrap());
316        PyArray1::from_iter(py, values).into_py_any(py).unwrap()
317    }
318}
319/// Convert dates directly to i64 with i64::MIN representing a null value.
320fn date_series_to_numpy(py: Python, s: &Series) -> PyObject {
321    use numpy::datetime::{units, Datetime};
322
323    let s_phys = s.to_physical_repr();
324    let ca = s_phys.i32().unwrap();
325
326    if s.null_count() == 0 {
327        let mapper = |v: i32| (v as i64).into();
328        let values = ca.into_no_null_iter().map(mapper);
329        PyArray1::<Datetime<units::Days>>::from_iter(py, values)
330            .into_py_any(py)
331            .unwrap()
332    } else {
333        let mapper = |opt_v: Option<i32>| {
334            match opt_v {
335                Some(v) => v as i64,
336                None => i64::MIN,
337            }
338            .into()
339        };
340        let values = ca.iter().map(mapper);
341        PyArray1::<Datetime<units::Days>>::from_iter(py, values)
342            .into_py_any(py)
343            .unwrap()
344    }
345}
346/// Convert datetimes and durations with i64::MIN representing a null value.
347fn temporal_series_to_numpy<T>(py: Python, s: &Series) -> PyObject
348where
349    T: From<i64> + numpy::Element,
350{
351    let s_phys = s.to_physical_repr();
352    let ca = s_phys.i64().unwrap();
353    let values = ca.iter().map(|v| v.unwrap_or(i64::MIN).into());
354    PyArray1::<T>::from_iter(py, values)
355        .into_py_any(py)
356        .unwrap()
357}
358fn list_series_to_numpy(py: Python, s: &Series, writable: bool) -> PyObject {
359    let ca = s.list().unwrap();
360
361    let iter = ca.amortized_iter().map(|opt_s| match opt_s {
362        None => py.None(),
363        Some(s) => series_to_numpy(py, s.as_ref(), writable, true).unwrap(),
364    });
365    PyArray1::from_iter(py, iter).into_py_any(py).unwrap()
366}
367/// Convert arrays by flattening first, converting the flat Series, and then reshaping.
368fn array_series_to_numpy(py: Python, s: &Series, writable: bool) -> PyObject {
369    let ca = s.array().unwrap();
370    let s_inner = ca.get_inner();
371    let np_array_flat = series_to_numpy_with_copy(py, &s_inner, writable);
372
373    // Reshape to the original shape.
374    let DataType::Array(_, width) = s.dtype() else {
375        unreachable!()
376    };
377    reshape_numpy_array(py, np_array_flat, ca.len(), *width).unwrap()
378}
polars_python/interop/numpy/to_numpy_series.rs

polars_python/interop/numpy/
to_numpy_series.rs
