1 | use crate::err::{PyErr, PyResult}; |
2 | use crate::{ffi, AsPyPointer, Py, PyAny, Python}; |
3 | use std::os::raw::c_char; |
4 | use std::slice; |
5 | |
6 | /// Represents a Python `bytearray`. |
7 | #[repr (transparent)] |
8 | pub struct PyByteArray(PyAny); |
9 | |
10 | pyobject_native_type_core!(PyByteArray, pyobject_native_static_type_object!(ffi::PyByteArray_Type), #checkfunction=ffi::PyByteArray_Check); |
11 | |
12 | impl PyByteArray { |
13 | /// Creates a new Python bytearray object. |
14 | /// |
15 | /// The byte string is initialized by copying the data from the `&[u8]`. |
16 | pub fn new<'p>(py: Python<'p>, src: &[u8]) -> &'p PyByteArray { |
17 | let ptr = src.as_ptr() as *const c_char; |
18 | let len = src.len() as ffi::Py_ssize_t; |
19 | unsafe { py.from_owned_ptr::<PyByteArray>(ffi::PyByteArray_FromStringAndSize(ptr, len)) } |
20 | } |
21 | |
22 | /// Creates a new Python `bytearray` object with an `init` closure to write its contents. |
23 | /// Before calling `init` the bytearray is zero-initialised. |
24 | /// * If Python raises a MemoryError on the allocation, `new_with` will return |
25 | /// it inside `Err`. |
26 | /// * If `init` returns `Err(e)`, `new_with` will return `Err(e)`. |
27 | /// * If `init` returns `Ok(())`, `new_with` will return `Ok(&PyByteArray)`. |
28 | /// |
29 | /// # Examples |
30 | /// |
31 | /// ``` |
32 | /// use pyo3::{prelude::*, types::PyByteArray}; |
33 | /// |
34 | /// # fn main() -> PyResult<()> { |
35 | /// Python::with_gil(|py| -> PyResult<()> { |
36 | /// let py_bytearray = PyByteArray::new_with(py, 10, |bytes: &mut [u8]| { |
37 | /// bytes.copy_from_slice(b"Hello Rust" ); |
38 | /// Ok(()) |
39 | /// })?; |
40 | /// let bytearray: &[u8] = unsafe { py_bytearray.as_bytes() }; |
41 | /// assert_eq!(bytearray, b"Hello Rust" ); |
42 | /// Ok(()) |
43 | /// }) |
44 | /// # } |
45 | /// ``` |
46 | pub fn new_with<F>(py: Python<'_>, len: usize, init: F) -> PyResult<&PyByteArray> |
47 | where |
48 | F: FnOnce(&mut [u8]) -> PyResult<()>, |
49 | { |
50 | unsafe { |
51 | let pyptr = |
52 | ffi::PyByteArray_FromStringAndSize(std::ptr::null(), len as ffi::Py_ssize_t); |
53 | // Check for an allocation error and return it |
54 | let pypybytearray: Py<PyByteArray> = Py::from_owned_ptr_or_err(py, pyptr)?; |
55 | let buffer: *mut u8 = ffi::PyByteArray_AsString(pyptr).cast(); |
56 | debug_assert!(!buffer.is_null()); |
57 | // Zero-initialise the uninitialised bytearray |
58 | std::ptr::write_bytes(buffer, 0u8, len); |
59 | // (Further) Initialise the bytearray in init |
60 | // If init returns an Err, pypybytearray will automatically deallocate the buffer |
61 | init(std::slice::from_raw_parts_mut(buffer, len)).map(|_| pypybytearray.into_ref(py)) |
62 | } |
63 | } |
64 | |
65 | /// Creates a new Python `bytearray` object from another Python object that |
66 | /// implements the buffer protocol. |
67 | pub fn from(src: &PyAny) -> PyResult<&PyByteArray> { |
68 | unsafe { |
69 | src.py() |
70 | .from_owned_ptr_or_err(ffi::PyByteArray_FromObject(src.as_ptr())) |
71 | } |
72 | } |
73 | |
74 | /// Gets the length of the bytearray. |
75 | #[inline ] |
76 | pub fn len(&self) -> usize { |
77 | // non-negative Py_ssize_t should always fit into Rust usize |
78 | unsafe { ffi::PyByteArray_Size(self.as_ptr()) as usize } |
79 | } |
80 | |
81 | /// Checks if the bytearray is empty. |
82 | pub fn is_empty(&self) -> bool { |
83 | self.len() == 0 |
84 | } |
85 | |
86 | /// Gets the start of the buffer containing the contents of the bytearray. |
87 | /// |
88 | /// # Safety |
89 | /// |
90 | /// See the safety requirements of [`PyByteArray::as_bytes`] and [`PyByteArray::as_bytes_mut`]. |
91 | pub fn data(&self) -> *mut u8 { |
92 | unsafe { ffi::PyByteArray_AsString(self.as_ptr()).cast() } |
93 | } |
94 | |
95 | /// Extracts a slice of the `ByteArray`'s entire buffer. |
96 | /// |
97 | /// # Safety |
98 | /// |
99 | /// Mutation of the `bytearray` invalidates the slice. If it is used afterwards, the behavior is |
100 | /// undefined. |
101 | /// |
102 | /// These mutations may occur in Python code as well as from Rust: |
103 | /// - Calling methods like [`PyByteArray::as_bytes_mut`] and [`PyByteArray::resize`] will |
104 | /// invalidate the slice. |
105 | /// - Actions like dropping objects or raising exceptions can invoke `__del__`methods or signal |
106 | /// handlers, which may execute arbitrary Python code. This means that if Python code has a |
107 | /// reference to the `bytearray` you cannot safely use the vast majority of PyO3's API whilst |
108 | /// using the slice. |
109 | /// |
110 | /// As a result, this slice should only be used for short-lived operations without executing any |
111 | /// Python code, such as copying into a Vec. |
112 | /// |
113 | /// # Examples |
114 | /// |
115 | /// ```rust |
116 | /// use pyo3::prelude::*; |
117 | /// use pyo3::exceptions::PyRuntimeError; |
118 | /// use pyo3::types::PyByteArray; |
119 | /// |
120 | /// #[pyfunction] |
121 | /// fn a_valid_function(bytes: &PyByteArray) -> PyResult<()> { |
122 | /// let section = { |
123 | /// // SAFETY: We promise to not let the interpreter regain control |
124 | /// // or invoke any PyO3 APIs while using the slice. |
125 | /// let slice = unsafe { bytes.as_bytes() }; |
126 | /// |
127 | /// // Copy only a section of `bytes` while avoiding |
128 | /// // `to_vec` which copies the entire thing. |
129 | /// let section = slice |
130 | /// .get(6..11) |
131 | /// .ok_or_else(|| PyRuntimeError::new_err("input is not long enough" ))?; |
132 | /// Vec::from(section) |
133 | /// }; |
134 | /// |
135 | /// // Now we can do things with `section` and call PyO3 APIs again. |
136 | /// // ... |
137 | /// # assert_eq!(§ion, b"world" ); |
138 | /// |
139 | /// Ok(()) |
140 | /// } |
141 | /// # fn main() -> PyResult<()> { |
142 | /// # Python::with_gil(|py| -> PyResult<()> { |
143 | /// # let fun = wrap_pyfunction!(a_valid_function, py)?; |
144 | /// # let locals = pyo3::types::PyDict::new(py); |
145 | /// # locals.set_item("a_valid_function" , fun)?; |
146 | /// # |
147 | /// # py.run( |
148 | /// # r#"b = bytearray(b"hello world") |
149 | /// # a_valid_function(b) |
150 | /// # |
151 | /// # try: |
152 | /// # a_valid_function(bytearray()) |
153 | /// # except RuntimeError as e: |
154 | /// # assert str(e) == 'input is not long enough'"# , |
155 | /// # None, |
156 | /// # Some(locals), |
157 | /// # )?; |
158 | /// # |
159 | /// # Ok(()) |
160 | /// # }) |
161 | /// # } |
162 | /// ``` |
163 | /// |
164 | /// # Incorrect usage |
165 | /// |
166 | /// The following `bug` function is unsound ⚠️ |
167 | /// |
168 | /// ```rust,no_run |
169 | /// # use pyo3::prelude::*; |
170 | /// # use pyo3::types::PyByteArray; |
171 | /// |
172 | /// # #[allow (dead_code)] |
173 | /// #[pyfunction] |
174 | /// fn bug(py: Python<'_>, bytes: &PyByteArray) { |
175 | /// let slice = unsafe { bytes.as_bytes() }; |
176 | /// |
177 | /// // This explicitly yields control back to the Python interpreter... |
178 | /// // ...but it's not always this obvious. Many things do this implicitly. |
179 | /// py.allow_threads(|| { |
180 | /// // Python code could be mutating through its handle to `bytes`, |
181 | /// // which makes reading it a data race, which is undefined behavior. |
182 | /// println!("{:?}" , slice[0]); |
183 | /// }); |
184 | /// |
185 | /// // Python code might have mutated it, so we can not rely on the slice |
186 | /// // remaining valid. As such this is also undefined behavior. |
187 | /// println!("{:?}" , slice[0]); |
188 | /// } |
189 | /// ``` |
190 | pub unsafe fn as_bytes(&self) -> &[u8] { |
191 | slice::from_raw_parts(self.data(), self.len()) |
192 | } |
193 | |
194 | /// Extracts a mutable slice of the `ByteArray`'s entire buffer. |
195 | /// |
196 | /// # Safety |
197 | /// |
198 | /// Any other accesses of the `bytearray`'s buffer invalidate the slice. If it is used |
199 | /// afterwards, the behavior is undefined. The safety requirements of [`PyByteArray::as_bytes`] |
200 | /// apply to this function as well. |
201 | #[allow (clippy::mut_from_ref)] |
202 | pub unsafe fn as_bytes_mut(&self) -> &mut [u8] { |
203 | slice::from_raw_parts_mut(self.data(), self.len()) |
204 | } |
205 | |
206 | /// Copies the contents of the bytearray to a Rust vector. |
207 | /// |
208 | /// # Examples |
209 | /// |
210 | /// ``` |
211 | /// # use pyo3::prelude::*; |
212 | /// # use pyo3::types::PyByteArray; |
213 | /// # Python::with_gil(|py| { |
214 | /// let bytearray = PyByteArray::new(py, b"Hello World." ); |
215 | /// let mut copied_message = bytearray.to_vec(); |
216 | /// assert_eq!(b"Hello World." , copied_message.as_slice()); |
217 | /// |
218 | /// copied_message[11] = b'!' ; |
219 | /// assert_eq!(b"Hello World!" , copied_message.as_slice()); |
220 | /// |
221 | /// pyo3::py_run!(py, bytearray, "assert bytearray == b'Hello World.'" ); |
222 | /// # }); |
223 | /// ``` |
224 | pub fn to_vec(&self) -> Vec<u8> { |
225 | unsafe { self.as_bytes() }.to_vec() |
226 | } |
227 | |
228 | /// Resizes the bytearray object to the new length `len`. |
229 | /// |
230 | /// Note that this will invalidate any pointers obtained by [PyByteArray::data], as well as |
231 | /// any (unsafe) slices obtained from [PyByteArray::as_bytes] and [PyByteArray::as_bytes_mut]. |
232 | pub fn resize(&self, len: usize) -> PyResult<()> { |
233 | unsafe { |
234 | let result = ffi::PyByteArray_Resize(self.as_ptr(), len as ffi::Py_ssize_t); |
235 | if result == 0 { |
236 | Ok(()) |
237 | } else { |
238 | Err(PyErr::fetch(self.py())) |
239 | } |
240 | } |
241 | } |
242 | } |
243 | |
244 | #[cfg (test)] |
245 | mod tests { |
246 | use crate::exceptions; |
247 | use crate::types::PyByteArray; |
248 | use crate::{PyObject, Python}; |
249 | |
250 | #[test ] |
251 | fn test_len() { |
252 | Python::with_gil(|py| { |
253 | let src = b"Hello Python" ; |
254 | let bytearray = PyByteArray::new(py, src); |
255 | assert_eq!(src.len(), bytearray.len()); |
256 | }); |
257 | } |
258 | |
259 | #[test ] |
260 | fn test_as_bytes() { |
261 | Python::with_gil(|py| { |
262 | let src = b"Hello Python" ; |
263 | let bytearray = PyByteArray::new(py, src); |
264 | |
265 | let slice = unsafe { bytearray.as_bytes() }; |
266 | assert_eq!(src, slice); |
267 | assert_eq!(bytearray.data() as *const _, slice.as_ptr()); |
268 | }); |
269 | } |
270 | |
271 | #[test ] |
272 | fn test_as_bytes_mut() { |
273 | Python::with_gil(|py| { |
274 | let src = b"Hello Python" ; |
275 | let bytearray = PyByteArray::new(py, src); |
276 | |
277 | let slice = unsafe { bytearray.as_bytes_mut() }; |
278 | assert_eq!(src, slice); |
279 | assert_eq!(bytearray.data(), slice.as_mut_ptr()); |
280 | |
281 | slice[0..5].copy_from_slice(b"Hi..." ); |
282 | |
283 | assert_eq!( |
284 | bytearray.str().unwrap().to_str().unwrap(), |
285 | "bytearray(b'Hi... Python')" |
286 | ); |
287 | }); |
288 | } |
289 | |
290 | #[test ] |
291 | fn test_to_vec() { |
292 | Python::with_gil(|py| { |
293 | let src = b"Hello Python" ; |
294 | let bytearray = PyByteArray::new(py, src); |
295 | |
296 | let vec = bytearray.to_vec(); |
297 | assert_eq!(src, vec.as_slice()); |
298 | }); |
299 | } |
300 | |
301 | #[test ] |
302 | fn test_from() { |
303 | Python::with_gil(|py| { |
304 | let src = b"Hello Python" ; |
305 | let bytearray = PyByteArray::new(py, src); |
306 | |
307 | let ba: PyObject = bytearray.into(); |
308 | let bytearray = PyByteArray::from(ba.as_ref(py)).unwrap(); |
309 | |
310 | assert_eq!(src, unsafe { bytearray.as_bytes() }); |
311 | }); |
312 | } |
313 | |
314 | #[test ] |
315 | fn test_from_err() { |
316 | Python::with_gil(|py| { |
317 | if let Err(err) = PyByteArray::from(py.None().as_ref(py)) { |
318 | assert!(err.is_instance_of::<exceptions::PyTypeError>(py)); |
319 | } else { |
320 | panic!("error" ); |
321 | } |
322 | }); |
323 | } |
324 | |
325 | #[test ] |
326 | fn test_resize() { |
327 | Python::with_gil(|py| { |
328 | let src = b"Hello Python" ; |
329 | let bytearray = PyByteArray::new(py, src); |
330 | |
331 | bytearray.resize(20).unwrap(); |
332 | assert_eq!(20, bytearray.len()); |
333 | }); |
334 | } |
335 | |
336 | #[test ] |
337 | fn test_byte_array_new_with() -> super::PyResult<()> { |
338 | Python::with_gil(|py| -> super::PyResult<()> { |
339 | let py_bytearray = PyByteArray::new_with(py, 10, |b: &mut [u8]| { |
340 | b.copy_from_slice(b"Hello Rust" ); |
341 | Ok(()) |
342 | })?; |
343 | let bytearray: &[u8] = unsafe { py_bytearray.as_bytes() }; |
344 | assert_eq!(bytearray, b"Hello Rust" ); |
345 | Ok(()) |
346 | }) |
347 | } |
348 | |
349 | #[test ] |
350 | fn test_byte_array_new_with_zero_initialised() -> super::PyResult<()> { |
351 | Python::with_gil(|py| -> super::PyResult<()> { |
352 | let py_bytearray = PyByteArray::new_with(py, 10, |_b: &mut [u8]| Ok(()))?; |
353 | let bytearray: &[u8] = unsafe { py_bytearray.as_bytes() }; |
354 | assert_eq!(bytearray, &[0; 10]); |
355 | Ok(()) |
356 | }) |
357 | } |
358 | |
359 | #[test ] |
360 | fn test_byte_array_new_with_error() { |
361 | use crate::exceptions::PyValueError; |
362 | Python::with_gil(|py| { |
363 | let py_bytearray_result = PyByteArray::new_with(py, 10, |_b: &mut [u8]| { |
364 | Err(PyValueError::new_err("Hello Crustaceans!" )) |
365 | }); |
366 | assert!(py_bytearray_result.is_err()); |
367 | assert!(py_bytearray_result |
368 | .err() |
369 | .unwrap() |
370 | .is_instance_of::<PyValueError>(py)); |
371 | }) |
372 | } |
373 | } |
374 | |