]> git.proxmox.com Git - rustc.git/blob - library/alloc/src/ffi/c_str.rs
New upstream version 1.62.1+dfsg1
[rustc.git] / library / alloc / src / ffi / c_str.rs
1 #[cfg(test)]
2 mod tests;
3
4 use crate::borrow::{Cow, ToOwned};
5 use crate::boxed::Box;
6 use crate::rc::Rc;
7 use crate::slice::hack::into_vec;
8 use crate::string::String;
9 use crate::vec::Vec;
10 use core::borrow::Borrow;
11 use core::ffi::{c_char, CStr};
12 use core::fmt;
13 use core::mem;
14 use core::num::NonZeroU8;
15 use core::ops;
16 use core::ptr;
17 use core::slice;
18 use core::slice::memchr;
19 use core::str::{self, Utf8Error};
20
21 #[cfg(target_has_atomic = "ptr")]
22 use crate::sync::Arc;
23
24 /// A type representing an owned, C-compatible, nul-terminated string with no nul bytes in the
25 /// middle.
26 ///
27 /// This type serves the purpose of being able to safely generate a
28 /// C-compatible string from a Rust byte slice or vector. An instance of this
29 /// type is a static guarantee that the underlying bytes contain no interior 0
30 /// bytes ("nul characters") and that the final byte is 0 ("nul terminator").
31 ///
32 /// `CString` is to <code>&[CStr]</code> as [`String`] is to <code>&[str]</code>: the former
33 /// in each pair are owned strings; the latter are borrowed
34 /// references.
35 ///
36 /// # Creating a `CString`
37 ///
38 /// A `CString` is created from either a byte slice or a byte vector,
39 /// or anything that implements <code>[Into]<[Vec]<[u8]>></code> (for
40 /// example, you can build a `CString` straight out of a [`String`] or
41 /// a <code>&[str]</code>, since both implement that trait).
42 ///
43 /// The [`CString::new`] method will actually check that the provided <code>&[[u8]]</code>
44 /// does not have 0 bytes in the middle, and return an error if it
45 /// finds one.
46 ///
47 /// # Extracting a raw pointer to the whole C string
48 ///
49 /// `CString` implements an [`as_ptr`][`CStr::as_ptr`] method through the [`Deref`]
50 /// trait. This method will give you a `*const c_char` which you can
51 /// feed directly to extern functions that expect a nul-terminated
52 /// string, like C's `strdup()`. Notice that [`as_ptr`][`CStr::as_ptr`] returns a
53 /// read-only pointer; if the C code writes to it, that causes
54 /// undefined behavior.
55 ///
56 /// # Extracting a slice of the whole C string
57 ///
58 /// Alternatively, you can obtain a <code>&[[u8]]</code> slice from a
59 /// `CString` with the [`CString::as_bytes`] method. Slices produced in this
60 /// way do *not* contain the trailing nul terminator. This is useful
61 /// when you will be calling an extern function that takes a `*const
62 /// u8` argument which is not necessarily nul-terminated, plus another
63 /// argument with the length of the string — like C's `strndup()`.
64 /// You can of course get the slice's length with its
65 /// [`len`][slice::len] method.
66 ///
67 /// If you need a <code>&[[u8]]</code> slice *with* the nul terminator, you
68 /// can use [`CString::as_bytes_with_nul`] instead.
69 ///
70 /// Once you have the kind of slice you need (with or without a nul
71 /// terminator), you can call the slice's own
72 /// [`as_ptr`][slice::as_ptr] method to get a read-only raw pointer to pass to
73 /// extern functions. See the documentation for that function for a
74 /// discussion on ensuring the lifetime of the raw pointer.
75 ///
76 /// [str]: prim@str "str"
77 /// [`Deref`]: ops::Deref
78 ///
79 /// # Examples
80 ///
81 /// ```ignore (extern-declaration)
82 /// # fn main() {
83 /// use std::ffi::CString;
84 /// use std::os::raw::c_char;
85 ///
86 /// extern "C" {
87 /// fn my_printer(s: *const c_char);
88 /// }
89 ///
90 /// // We are certain that our string doesn't have 0 bytes in the middle,
91 /// // so we can .expect()
92 /// let c_to_print = CString::new("Hello, world!").expect("CString::new failed");
93 /// unsafe {
94 /// my_printer(c_to_print.as_ptr());
95 /// }
96 /// # }
97 /// ```
98 ///
99 /// # Safety
100 ///
101 /// `CString` is intended for working with traditional C-style strings
102 /// (a sequence of non-nul bytes terminated by a single nul byte); the
103 /// primary use case for these kinds of strings is interoperating with C-like
104 /// code. Often you will need to transfer ownership to/from that external
105 /// code. It is strongly recommended that you thoroughly read through the
106 /// documentation of `CString` before use, as improper ownership management
107 /// of `CString` instances can lead to invalid memory accesses, memory leaks,
108 /// and other memory errors.
109 #[derive(PartialEq, PartialOrd, Eq, Ord, Hash, Clone)]
110 #[cfg_attr(not(test), rustc_diagnostic_item = "cstring_type")]
111 #[unstable(feature = "alloc_c_string", issue = "94079")]
112 pub struct CString {
113 // Invariant 1: the slice ends with a zero byte and has a length of at least one.
114 // Invariant 2: the slice contains only one zero byte.
115 // Improper usage of unsafe function can break Invariant 2, but not Invariant 1.
116 inner: Box<[u8]>,
117 }
118
119 /// An error indicating that an interior nul byte was found.
120 ///
121 /// While Rust strings may contain nul bytes in the middle, C strings
122 /// can't, as that byte would effectively truncate the string.
123 ///
124 /// This error is created by the [`new`][`CString::new`] method on
125 /// [`CString`]. See its documentation for more.
126 ///
127 /// # Examples
128 ///
129 /// ```
130 /// use std::ffi::{CString, NulError};
131 ///
132 /// let _: NulError = CString::new(b"f\0oo".to_vec()).unwrap_err();
133 /// ```
134 #[derive(Clone, PartialEq, Eq, Debug)]
135 #[unstable(feature = "alloc_c_string", issue = "94079")]
136 pub struct NulError(usize, Vec<u8>);
137
138 #[derive(Clone, PartialEq, Eq, Debug)]
139 enum FromBytesWithNulErrorKind {
140 InteriorNul(usize),
141 NotNulTerminated,
142 }
143
144 /// An error indicating that a nul byte was not in the expected position.
145 ///
146 /// The vector used to create a [`CString`] must have one and only one nul byte,
147 /// positioned at the end.
148 ///
149 /// This error is created by the [`CString::from_vec_with_nul`] method.
150 /// See its documentation for more.
151 ///
152 /// # Examples
153 ///
154 /// ```
155 /// use std::ffi::{CString, FromVecWithNulError};
156 ///
157 /// let _: FromVecWithNulError = CString::from_vec_with_nul(b"f\0oo".to_vec()).unwrap_err();
158 /// ```
159 #[derive(Clone, PartialEq, Eq, Debug)]
160 #[unstable(feature = "alloc_c_string", issue = "94079")]
161 pub struct FromVecWithNulError {
162 error_kind: FromBytesWithNulErrorKind,
163 bytes: Vec<u8>,
164 }
165
166 #[stable(feature = "cstring_from_vec_with_nul", since = "1.58.0")]
167 impl FromVecWithNulError {
168 /// Returns a slice of [`u8`]s bytes that were attempted to convert to a [`CString`].
169 ///
170 /// # Examples
171 ///
172 /// Basic usage:
173 ///
174 /// ```
175 /// use std::ffi::CString;
176 ///
177 /// // Some invalid bytes in a vector
178 /// let bytes = b"f\0oo".to_vec();
179 ///
180 /// let value = CString::from_vec_with_nul(bytes.clone());
181 ///
182 /// assert_eq!(&bytes[..], value.unwrap_err().as_bytes());
183 /// ```
184 #[must_use]
185 #[stable(feature = "cstring_from_vec_with_nul", since = "1.58.0")]
186 pub fn as_bytes(&self) -> &[u8] {
187 &self.bytes[..]
188 }
189
190 /// Returns the bytes that were attempted to convert to a [`CString`].
191 ///
192 /// This method is carefully constructed to avoid allocation. It will
193 /// consume the error, moving out the bytes, so that a copy of the bytes
194 /// does not need to be made.
195 ///
196 /// # Examples
197 ///
198 /// Basic usage:
199 ///
200 /// ```
201 /// use std::ffi::CString;
202 ///
203 /// // Some invalid bytes in a vector
204 /// let bytes = b"f\0oo".to_vec();
205 ///
206 /// let value = CString::from_vec_with_nul(bytes.clone());
207 ///
208 /// assert_eq!(bytes, value.unwrap_err().into_bytes());
209 /// ```
210 #[must_use = "`self` will be dropped if the result is not used"]
211 #[stable(feature = "cstring_from_vec_with_nul", since = "1.58.0")]
212 pub fn into_bytes(self) -> Vec<u8> {
213 self.bytes
214 }
215 }
216
217 /// An error indicating invalid UTF-8 when converting a [`CString`] into a [`String`].
218 ///
219 /// `CString` is just a wrapper over a buffer of bytes with a nul terminator;
220 /// [`CString::into_string`] performs UTF-8 validation on those bytes and may
221 /// return this error.
222 ///
223 /// This `struct` is created by [`CString::into_string()`]. See
224 /// its documentation for more.
225 #[derive(Clone, PartialEq, Eq, Debug)]
226 #[unstable(feature = "alloc_c_string", issue = "94079")]
227 pub struct IntoStringError {
228 inner: CString,
229 error: Utf8Error,
230 }
231
232 impl CString {
233 /// Creates a new C-compatible string from a container of bytes.
234 ///
235 /// This function will consume the provided data and use the
236 /// underlying bytes to construct a new string, ensuring that
237 /// there is a trailing 0 byte. This trailing 0 byte will be
238 /// appended by this function; the provided data should *not*
239 /// contain any 0 bytes in it.
240 ///
241 /// # Examples
242 ///
243 /// ```ignore (extern-declaration)
244 /// use std::ffi::CString;
245 /// use std::os::raw::c_char;
246 ///
247 /// extern "C" { fn puts(s: *const c_char); }
248 ///
249 /// let to_print = CString::new("Hello!").expect("CString::new failed");
250 /// unsafe {
251 /// puts(to_print.as_ptr());
252 /// }
253 /// ```
254 ///
255 /// # Errors
256 ///
257 /// This function will return an error if the supplied bytes contain an
258 /// internal 0 byte. The [`NulError`] returned will contain the bytes as well as
259 /// the position of the nul byte.
260 #[stable(feature = "rust1", since = "1.0.0")]
261 pub fn new<T: Into<Vec<u8>>>(t: T) -> Result<CString, NulError> {
262 trait SpecNewImpl {
263 fn spec_new_impl(self) -> Result<CString, NulError>;
264 }
265
266 impl<T: Into<Vec<u8>>> SpecNewImpl for T {
267 default fn spec_new_impl(self) -> Result<CString, NulError> {
268 let bytes: Vec<u8> = self.into();
269 match memchr::memchr(0, &bytes) {
270 Some(i) => Err(NulError(i, bytes)),
271 None => Ok(unsafe { CString::_from_vec_unchecked(bytes) }),
272 }
273 }
274 }
275
276 // Specialization for avoiding reallocation
277 #[inline(always)] // Without that it is not inlined into specializations
278 fn spec_new_impl_bytes(bytes: &[u8]) -> Result<CString, NulError> {
279 // We cannot have such large slice that we would overflow here
280 // but using `checked_add` allows LLVM to assume that capacity never overflows
281 // and generate twice shorter code.
282 // `saturating_add` doesn't help for some reason.
283 let capacity = bytes.len().checked_add(1).unwrap();
284
285 // Allocate before validation to avoid duplication of allocation code.
286 // We still need to allocate and copy memory even if we get an error.
287 let mut buffer = Vec::with_capacity(capacity);
288 buffer.extend(bytes);
289
290 // Check memory of self instead of new buffer.
291 // This allows better optimizations if lto enabled.
292 match memchr::memchr(0, bytes) {
293 Some(i) => Err(NulError(i, buffer)),
294 None => Ok(unsafe { CString::_from_vec_unchecked(buffer) }),
295 }
296 }
297
298 impl SpecNewImpl for &'_ [u8] {
299 fn spec_new_impl(self) -> Result<CString, NulError> {
300 spec_new_impl_bytes(self)
301 }
302 }
303
304 impl SpecNewImpl for &'_ str {
305 fn spec_new_impl(self) -> Result<CString, NulError> {
306 spec_new_impl_bytes(self.as_bytes())
307 }
308 }
309
310 impl SpecNewImpl for &'_ mut [u8] {
311 fn spec_new_impl(self) -> Result<CString, NulError> {
312 spec_new_impl_bytes(self)
313 }
314 }
315
316 t.spec_new_impl()
317 }
318
319 /// Creates a C-compatible string by consuming a byte vector,
320 /// without checking for interior 0 bytes.
321 ///
322 /// Trailing 0 byte will be appended by this function.
323 ///
324 /// This method is equivalent to [`CString::new`] except that no runtime
325 /// assertion is made that `v` contains no 0 bytes, and it requires an
326 /// actual byte vector, not anything that can be converted to one with Into.
327 ///
328 /// # Examples
329 ///
330 /// ```
331 /// use std::ffi::CString;
332 ///
333 /// let raw = b"foo".to_vec();
334 /// unsafe {
335 /// let c_string = CString::from_vec_unchecked(raw);
336 /// }
337 /// ```
338 #[must_use]
339 #[stable(feature = "rust1", since = "1.0.0")]
340 pub unsafe fn from_vec_unchecked(v: Vec<u8>) -> Self {
341 debug_assert!(memchr::memchr(0, &v).is_none());
342 unsafe { Self::_from_vec_unchecked(v) }
343 }
344
345 unsafe fn _from_vec_unchecked(mut v: Vec<u8>) -> Self {
346 v.reserve_exact(1);
347 v.push(0);
348 Self { inner: v.into_boxed_slice() }
349 }
350
351 /// Retakes ownership of a `CString` that was transferred to C via
352 /// [`CString::into_raw`].
353 ///
354 /// Additionally, the length of the string will be recalculated from the pointer.
355 ///
356 /// # Safety
357 ///
358 /// This should only ever be called with a pointer that was earlier
359 /// obtained by calling [`CString::into_raw`]. Other usage (e.g., trying to take
360 /// ownership of a string that was allocated by foreign code) is likely to lead
361 /// to undefined behavior or allocator corruption.
362 ///
363 /// It should be noted that the length isn't just "recomputed," but that
364 /// the recomputed length must match the original length from the
365 /// [`CString::into_raw`] call. This means the [`CString::into_raw`]/`from_raw`
366 /// methods should not be used when passing the string to C functions that can
367 /// modify the string's length.
368 ///
369 /// > **Note:** If you need to borrow a string that was allocated by
370 /// > foreign code, use [`CStr`]. If you need to take ownership of
371 /// > a string that was allocated by foreign code, you will need to
372 /// > make your own provisions for freeing it appropriately, likely
373 /// > with the foreign code's API to do that.
374 ///
375 /// # Examples
376 ///
377 /// Creates a `CString`, pass ownership to an `extern` function (via raw pointer), then retake
378 /// ownership with `from_raw`:
379 ///
380 /// ```ignore (extern-declaration)
381 /// use std::ffi::CString;
382 /// use std::os::raw::c_char;
383 ///
384 /// extern "C" {
385 /// fn some_extern_function(s: *mut c_char);
386 /// }
387 ///
388 /// let c_string = CString::new("Hello!").expect("CString::new failed");
389 /// let raw = c_string.into_raw();
390 /// unsafe {
391 /// some_extern_function(raw);
392 /// let c_string = CString::from_raw(raw);
393 /// }
394 /// ```
395 #[must_use = "call `drop(from_raw(ptr))` if you intend to drop the `CString`"]
396 #[stable(feature = "cstr_memory", since = "1.4.0")]
397 pub unsafe fn from_raw(ptr: *mut c_char) -> CString {
398 // SAFETY: This is called with a pointer that was obtained from a call
399 // to `CString::into_raw` and the length has not been modified. As such,
400 // we know there is a NUL byte (and only one) at the end and that the
401 // information about the size of the allocation is correct on Rust's
402 // side.
403 unsafe {
404 extern "C" {
405 /// Provided by libc or compiler_builtins.
406 fn strlen(s: *const c_char) -> usize;
407 }
408 let len = strlen(ptr) + 1; // Including the NUL byte
409 let slice = slice::from_raw_parts_mut(ptr, len as usize);
410 CString { inner: Box::from_raw(slice as *mut [c_char] as *mut [u8]) }
411 }
412 }
413
414 /// Consumes the `CString` and transfers ownership of the string to a C caller.
415 ///
416 /// The pointer which this function returns must be returned to Rust and reconstituted using
417 /// [`CString::from_raw`] to be properly deallocated. Specifically, one
418 /// should *not* use the standard C `free()` function to deallocate
419 /// this string.
420 ///
421 /// Failure to call [`CString::from_raw`] will lead to a memory leak.
422 ///
423 /// The C side must **not** modify the length of the string (by writing a
424 /// `null` somewhere inside the string or removing the final one) before
425 /// it makes it back into Rust using [`CString::from_raw`]. See the safety section
426 /// in [`CString::from_raw`].
427 ///
428 /// # Examples
429 ///
430 /// ```
431 /// use std::ffi::CString;
432 ///
433 /// let c_string = CString::new("foo").expect("CString::new failed");
434 ///
435 /// let ptr = c_string.into_raw();
436 ///
437 /// unsafe {
438 /// assert_eq!(b'f', *ptr as u8);
439 /// assert_eq!(b'o', *ptr.offset(1) as u8);
440 /// assert_eq!(b'o', *ptr.offset(2) as u8);
441 /// assert_eq!(b'\0', *ptr.offset(3) as u8);
442 ///
443 /// // retake pointer to free memory
444 /// let _ = CString::from_raw(ptr);
445 /// }
446 /// ```
447 #[inline]
448 #[must_use = "`self` will be dropped if the result is not used"]
449 #[stable(feature = "cstr_memory", since = "1.4.0")]
450 pub fn into_raw(self) -> *mut c_char {
451 Box::into_raw(self.into_inner()) as *mut c_char
452 }
453
454 /// Converts the `CString` into a [`String`] if it contains valid UTF-8 data.
455 ///
456 /// On failure, ownership of the original `CString` is returned.
457 ///
458 /// # Examples
459 ///
460 /// ```
461 /// use std::ffi::CString;
462 ///
463 /// let valid_utf8 = vec![b'f', b'o', b'o'];
464 /// let cstring = CString::new(valid_utf8).expect("CString::new failed");
465 /// assert_eq!(cstring.into_string().expect("into_string() call failed"), "foo");
466 ///
467 /// let invalid_utf8 = vec![b'f', 0xff, b'o', b'o'];
468 /// let cstring = CString::new(invalid_utf8).expect("CString::new failed");
469 /// let err = cstring.into_string().err().expect("into_string().err() failed");
470 /// assert_eq!(err.utf8_error().valid_up_to(), 1);
471 /// ```
472 #[stable(feature = "cstring_into", since = "1.7.0")]
473 pub fn into_string(self) -> Result<String, IntoStringError> {
474 String::from_utf8(self.into_bytes()).map_err(|e| IntoStringError {
475 error: e.utf8_error(),
476 inner: unsafe { Self::_from_vec_unchecked(e.into_bytes()) },
477 })
478 }
479
480 /// Consumes the `CString` and returns the underlying byte buffer.
481 ///
482 /// The returned buffer does **not** contain the trailing nul
483 /// terminator, and it is guaranteed to not have any interior nul
484 /// bytes.
485 ///
486 /// # Examples
487 ///
488 /// ```
489 /// use std::ffi::CString;
490 ///
491 /// let c_string = CString::new("foo").expect("CString::new failed");
492 /// let bytes = c_string.into_bytes();
493 /// assert_eq!(bytes, vec![b'f', b'o', b'o']);
494 /// ```
495 #[must_use = "`self` will be dropped if the result is not used"]
496 #[stable(feature = "cstring_into", since = "1.7.0")]
497 pub fn into_bytes(self) -> Vec<u8> {
498 let mut vec = into_vec(self.into_inner());
499 let _nul = vec.pop();
500 debug_assert_eq!(_nul, Some(0u8));
501 vec
502 }
503
504 /// Equivalent to [`CString::into_bytes()`] except that the
505 /// returned vector includes the trailing nul terminator.
506 ///
507 /// # Examples
508 ///
509 /// ```
510 /// use std::ffi::CString;
511 ///
512 /// let c_string = CString::new("foo").expect("CString::new failed");
513 /// let bytes = c_string.into_bytes_with_nul();
514 /// assert_eq!(bytes, vec![b'f', b'o', b'o', b'\0']);
515 /// ```
516 #[must_use = "`self` will be dropped if the result is not used"]
517 #[stable(feature = "cstring_into", since = "1.7.0")]
518 pub fn into_bytes_with_nul(self) -> Vec<u8> {
519 into_vec(self.into_inner())
520 }
521
522 /// Returns the contents of this `CString` as a slice of bytes.
523 ///
524 /// The returned slice does **not** contain the trailing nul
525 /// terminator, and it is guaranteed to not have any interior nul
526 /// bytes. If you need the nul terminator, use
527 /// [`CString::as_bytes_with_nul`] instead.
528 ///
529 /// # Examples
530 ///
531 /// ```
532 /// use std::ffi::CString;
533 ///
534 /// let c_string = CString::new("foo").expect("CString::new failed");
535 /// let bytes = c_string.as_bytes();
536 /// assert_eq!(bytes, &[b'f', b'o', b'o']);
537 /// ```
538 #[inline]
539 #[must_use]
540 #[stable(feature = "rust1", since = "1.0.0")]
541 pub fn as_bytes(&self) -> &[u8] {
542 // SAFETY: CString has a length at least 1
543 unsafe { self.inner.get_unchecked(..self.inner.len() - 1) }
544 }
545
546 /// Equivalent to [`CString::as_bytes()`] except that the
547 /// returned slice includes the trailing nul terminator.
548 ///
549 /// # Examples
550 ///
551 /// ```
552 /// use std::ffi::CString;
553 ///
554 /// let c_string = CString::new("foo").expect("CString::new failed");
555 /// let bytes = c_string.as_bytes_with_nul();
556 /// assert_eq!(bytes, &[b'f', b'o', b'o', b'\0']);
557 /// ```
558 #[inline]
559 #[must_use]
560 #[stable(feature = "rust1", since = "1.0.0")]
561 pub fn as_bytes_with_nul(&self) -> &[u8] {
562 &self.inner
563 }
564
565 /// Extracts a [`CStr`] slice containing the entire string.
566 ///
567 /// # Examples
568 ///
569 /// ```
570 /// use std::ffi::{CString, CStr};
571 ///
572 /// let c_string = CString::new(b"foo".to_vec()).expect("CString::new failed");
573 /// let cstr = c_string.as_c_str();
574 /// assert_eq!(cstr,
575 /// CStr::from_bytes_with_nul(b"foo\0").expect("CStr::from_bytes_with_nul failed"));
576 /// ```
577 #[inline]
578 #[must_use]
579 #[stable(feature = "as_c_str", since = "1.20.0")]
580 pub fn as_c_str(&self) -> &CStr {
581 &*self
582 }
583
584 /// Converts this `CString` into a boxed [`CStr`].
585 ///
586 /// # Examples
587 ///
588 /// ```
589 /// use std::ffi::{CString, CStr};
590 ///
591 /// let c_string = CString::new(b"foo".to_vec()).expect("CString::new failed");
592 /// let boxed = c_string.into_boxed_c_str();
593 /// assert_eq!(&*boxed,
594 /// CStr::from_bytes_with_nul(b"foo\0").expect("CStr::from_bytes_with_nul failed"));
595 /// ```
596 #[must_use = "`self` will be dropped if the result is not used"]
597 #[stable(feature = "into_boxed_c_str", since = "1.20.0")]
598 pub fn into_boxed_c_str(self) -> Box<CStr> {
599 unsafe { Box::from_raw(Box::into_raw(self.into_inner()) as *mut CStr) }
600 }
601
602 /// Bypass "move out of struct which implements [`Drop`] trait" restriction.
603 #[inline]
604 fn into_inner(self) -> Box<[u8]> {
605 // Rationale: `mem::forget(self)` invalidates the previous call to `ptr::read(&self.inner)`
606 // so we use `ManuallyDrop` to ensure `self` is not dropped.
607 // Then we can return the box directly without invalidating it.
608 // See https://github.com/rust-lang/rust/issues/62553.
609 let this = mem::ManuallyDrop::new(self);
610 unsafe { ptr::read(&this.inner) }
611 }
612
613 /// Converts a <code>[Vec]<[u8]></code> to a [`CString`] without checking the
614 /// invariants on the given [`Vec`].
615 ///
616 /// # Safety
617 ///
618 /// The given [`Vec`] **must** have one nul byte as its last element.
619 /// This means it cannot be empty nor have any other nul byte anywhere else.
620 ///
621 /// # Example
622 ///
623 /// ```
624 /// use std::ffi::CString;
625 /// assert_eq!(
626 /// unsafe { CString::from_vec_with_nul_unchecked(b"abc\0".to_vec()) },
627 /// unsafe { CString::from_vec_unchecked(b"abc".to_vec()) }
628 /// );
629 /// ```
630 #[must_use]
631 #[stable(feature = "cstring_from_vec_with_nul", since = "1.58.0")]
632 pub unsafe fn from_vec_with_nul_unchecked(v: Vec<u8>) -> Self {
633 debug_assert!(memchr::memchr(0, &v).unwrap() + 1 == v.len());
634 unsafe { Self::_from_vec_with_nul_unchecked(v) }
635 }
636
637 unsafe fn _from_vec_with_nul_unchecked(v: Vec<u8>) -> Self {
638 Self { inner: v.into_boxed_slice() }
639 }
640
641 /// Attempts to converts a <code>[Vec]<[u8]></code> to a [`CString`].
642 ///
643 /// Runtime checks are present to ensure there is only one nul byte in the
644 /// [`Vec`], its last element.
645 ///
646 /// # Errors
647 ///
648 /// If a nul byte is present and not the last element or no nul bytes
649 /// is present, an error will be returned.
650 ///
651 /// # Examples
652 ///
653 /// A successful conversion will produce the same result as [`CString::new`]
654 /// when called without the ending nul byte.
655 ///
656 /// ```
657 /// use std::ffi::CString;
658 /// assert_eq!(
659 /// CString::from_vec_with_nul(b"abc\0".to_vec())
660 /// .expect("CString::from_vec_with_nul failed"),
661 /// CString::new(b"abc".to_vec()).expect("CString::new failed")
662 /// );
663 /// ```
664 ///
665 /// An incorrectly formatted [`Vec`] will produce an error.
666 ///
667 /// ```
668 /// use std::ffi::{CString, FromVecWithNulError};
669 /// // Interior nul byte
670 /// let _: FromVecWithNulError = CString::from_vec_with_nul(b"a\0bc".to_vec()).unwrap_err();
671 /// // No nul byte
672 /// let _: FromVecWithNulError = CString::from_vec_with_nul(b"abc".to_vec()).unwrap_err();
673 /// ```
674 #[stable(feature = "cstring_from_vec_with_nul", since = "1.58.0")]
675 pub fn from_vec_with_nul(v: Vec<u8>) -> Result<Self, FromVecWithNulError> {
676 let nul_pos = memchr::memchr(0, &v);
677 match nul_pos {
678 Some(nul_pos) if nul_pos + 1 == v.len() => {
679 // SAFETY: We know there is only one nul byte, at the end
680 // of the vec.
681 Ok(unsafe { Self::_from_vec_with_nul_unchecked(v) })
682 }
683 Some(nul_pos) => Err(FromVecWithNulError {
684 error_kind: FromBytesWithNulErrorKind::InteriorNul(nul_pos),
685 bytes: v,
686 }),
687 None => Err(FromVecWithNulError {
688 error_kind: FromBytesWithNulErrorKind::NotNulTerminated,
689 bytes: v,
690 }),
691 }
692 }
693 }
694
695 // Turns this `CString` into an empty string to prevent
696 // memory-unsafe code from working by accident. Inline
697 // to prevent LLVM from optimizing it away in debug builds.
698 #[stable(feature = "cstring_drop", since = "1.13.0")]
699 impl Drop for CString {
700 #[inline]
701 fn drop(&mut self) {
702 unsafe {
703 *self.inner.get_unchecked_mut(0) = 0;
704 }
705 }
706 }
707
708 #[stable(feature = "rust1", since = "1.0.0")]
709 impl ops::Deref for CString {
710 type Target = CStr;
711
712 #[inline]
713 fn deref(&self) -> &CStr {
714 unsafe { CStr::from_bytes_with_nul_unchecked(self.as_bytes_with_nul()) }
715 }
716 }
717
718 #[stable(feature = "rust1", since = "1.0.0")]
719 impl fmt::Debug for CString {
720 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
721 fmt::Debug::fmt(&**self, f)
722 }
723 }
724
725 #[stable(feature = "cstring_into", since = "1.7.0")]
726 impl From<CString> for Vec<u8> {
727 /// Converts a [`CString`] into a <code>[Vec]<[u8]></code>.
728 ///
729 /// The conversion consumes the [`CString`], and removes the terminating NUL byte.
730 #[inline]
731 fn from(s: CString) -> Vec<u8> {
732 s.into_bytes()
733 }
734 }
735
736 #[stable(feature = "cstr_default", since = "1.10.0")]
737 impl Default for CString {
738 /// Creates an empty `CString`.
739 fn default() -> CString {
740 let a: &CStr = Default::default();
741 a.to_owned()
742 }
743 }
744
745 #[stable(feature = "cstr_borrow", since = "1.3.0")]
746 impl Borrow<CStr> for CString {
747 #[inline]
748 fn borrow(&self) -> &CStr {
749 self
750 }
751 }
752
753 #[stable(feature = "cstring_from_cow_cstr", since = "1.28.0")]
754 impl<'a> From<Cow<'a, CStr>> for CString {
755 /// Converts a `Cow<'a, CStr>` into a `CString`, by copying the contents if they are
756 /// borrowed.
757 #[inline]
758 fn from(s: Cow<'a, CStr>) -> Self {
759 s.into_owned()
760 }
761 }
762
763 #[cfg(not(test))]
764 #[stable(feature = "box_from_c_str", since = "1.17.0")]
765 impl From<&CStr> for Box<CStr> {
766 /// Converts a `&CStr` into a `Box<CStr>`,
767 /// by copying the contents into a newly allocated [`Box`].
768 fn from(s: &CStr) -> Box<CStr> {
769 let boxed: Box<[u8]> = Box::from(s.to_bytes_with_nul());
770 unsafe { Box::from_raw(Box::into_raw(boxed) as *mut CStr) }
771 }
772 }
773
774 #[stable(feature = "box_from_cow", since = "1.45.0")]
775 impl From<Cow<'_, CStr>> for Box<CStr> {
776 /// Converts a `Cow<'a, CStr>` into a `Box<CStr>`,
777 /// by copying the contents if they are borrowed.
778 #[inline]
779 fn from(cow: Cow<'_, CStr>) -> Box<CStr> {
780 match cow {
781 Cow::Borrowed(s) => Box::from(s),
782 Cow::Owned(s) => Box::from(s),
783 }
784 }
785 }
786
787 #[stable(feature = "c_string_from_box", since = "1.18.0")]
788 impl From<Box<CStr>> for CString {
789 /// Converts a <code>[Box]<[CStr]></code> into a [`CString`] without copying or allocating.
790 #[inline]
791 fn from(s: Box<CStr>) -> CString {
792 let raw = Box::into_raw(s) as *mut [u8];
793 CString { inner: unsafe { Box::from_raw(raw) } }
794 }
795 }
796
797 #[stable(feature = "cstring_from_vec_of_nonzerou8", since = "1.43.0")]
798 impl From<Vec<NonZeroU8>> for CString {
799 /// Converts a <code>[Vec]<[NonZeroU8]></code> into a [`CString`] without
800 /// copying nor checking for inner null bytes.
801 #[inline]
802 fn from(v: Vec<NonZeroU8>) -> CString {
803 unsafe {
804 // Transmute `Vec<NonZeroU8>` to `Vec<u8>`.
805 let v: Vec<u8> = {
806 // SAFETY:
807 // - transmuting between `NonZeroU8` and `u8` is sound;
808 // - `alloc::Layout<NonZeroU8> == alloc::Layout<u8>`.
809 let (ptr, len, cap): (*mut NonZeroU8, _, _) = Vec::into_raw_parts(v);
810 Vec::from_raw_parts(ptr.cast::<u8>(), len, cap)
811 };
812 // SAFETY: `v` cannot contain null bytes, given the type-level
813 // invariant of `NonZeroU8`.
814 Self::_from_vec_unchecked(v)
815 }
816 }
817 }
818
819 #[cfg(not(test))]
820 #[stable(feature = "more_box_slice_clone", since = "1.29.0")]
821 impl Clone for Box<CStr> {
822 #[inline]
823 fn clone(&self) -> Self {
824 (**self).into()
825 }
826 }
827
828 #[stable(feature = "box_from_c_string", since = "1.20.0")]
829 impl From<CString> for Box<CStr> {
830 /// Converts a [`CString`] into a <code>[Box]<[CStr]></code> without copying or allocating.
831 #[inline]
832 fn from(s: CString) -> Box<CStr> {
833 s.into_boxed_c_str()
834 }
835 }
836
837 #[stable(feature = "cow_from_cstr", since = "1.28.0")]
838 impl<'a> From<CString> for Cow<'a, CStr> {
839 /// Converts a [`CString`] into an owned [`Cow`] without copying or allocating.
840 #[inline]
841 fn from(s: CString) -> Cow<'a, CStr> {
842 Cow::Owned(s)
843 }
844 }
845
846 #[stable(feature = "cow_from_cstr", since = "1.28.0")]
847 impl<'a> From<&'a CStr> for Cow<'a, CStr> {
848 /// Converts a [`CStr`] into a borrowed [`Cow`] without copying or allocating.
849 #[inline]
850 fn from(s: &'a CStr) -> Cow<'a, CStr> {
851 Cow::Borrowed(s)
852 }
853 }
854
855 #[stable(feature = "cow_from_cstr", since = "1.28.0")]
856 impl<'a> From<&'a CString> for Cow<'a, CStr> {
857 /// Converts a `&`[`CString`] into a borrowed [`Cow`] without copying or allocating.
858 #[inline]
859 fn from(s: &'a CString) -> Cow<'a, CStr> {
860 Cow::Borrowed(s.as_c_str())
861 }
862 }
863
864 #[cfg(target_has_atomic = "ptr")]
865 #[stable(feature = "shared_from_slice2", since = "1.24.0")]
866 impl From<CString> for Arc<CStr> {
867 /// Converts a [`CString`] into an <code>[Arc]<[CStr]></code> by moving the [`CString`]
868 /// data into a new [`Arc`] buffer.
869 #[inline]
870 fn from(s: CString) -> Arc<CStr> {
871 let arc: Arc<[u8]> = Arc::from(s.into_inner());
872 unsafe { Arc::from_raw(Arc::into_raw(arc) as *const CStr) }
873 }
874 }
875
876 #[cfg(target_has_atomic = "ptr")]
877 #[stable(feature = "shared_from_slice2", since = "1.24.0")]
878 impl From<&CStr> for Arc<CStr> {
879 /// Converts a `&CStr` into a `Arc<CStr>`,
880 /// by copying the contents into a newly allocated [`Arc`].
881 #[inline]
882 fn from(s: &CStr) -> Arc<CStr> {
883 let arc: Arc<[u8]> = Arc::from(s.to_bytes_with_nul());
884 unsafe { Arc::from_raw(Arc::into_raw(arc) as *const CStr) }
885 }
886 }
887
888 #[stable(feature = "shared_from_slice2", since = "1.24.0")]
889 impl From<CString> for Rc<CStr> {
890 /// Converts a [`CString`] into an <code>[Rc]<[CStr]></code> by moving the [`CString`]
891 /// data into a new [`Arc`] buffer.
892 #[inline]
893 fn from(s: CString) -> Rc<CStr> {
894 let rc: Rc<[u8]> = Rc::from(s.into_inner());
895 unsafe { Rc::from_raw(Rc::into_raw(rc) as *const CStr) }
896 }
897 }
898
899 #[stable(feature = "shared_from_slice2", since = "1.24.0")]
900 impl From<&CStr> for Rc<CStr> {
901 /// Converts a `&CStr` into a `Rc<CStr>`,
902 /// by copying the contents into a newly allocated [`Rc`].
903 #[inline]
904 fn from(s: &CStr) -> Rc<CStr> {
905 let rc: Rc<[u8]> = Rc::from(s.to_bytes_with_nul());
906 unsafe { Rc::from_raw(Rc::into_raw(rc) as *const CStr) }
907 }
908 }
909
910 #[cfg(not(test))]
911 #[stable(feature = "default_box_extra", since = "1.17.0")]
912 impl Default for Box<CStr> {
913 fn default() -> Box<CStr> {
914 let boxed: Box<[u8]> = Box::from([0]);
915 unsafe { Box::from_raw(Box::into_raw(boxed) as *mut CStr) }
916 }
917 }
918
919 impl NulError {
920 /// Returns the position of the nul byte in the slice that caused
921 /// [`CString::new`] to fail.
922 ///
923 /// # Examples
924 ///
925 /// ```
926 /// use std::ffi::CString;
927 ///
928 /// let nul_error = CString::new("foo\0bar").unwrap_err();
929 /// assert_eq!(nul_error.nul_position(), 3);
930 ///
931 /// let nul_error = CString::new("foo bar\0").unwrap_err();
932 /// assert_eq!(nul_error.nul_position(), 7);
933 /// ```
934 #[must_use]
935 #[stable(feature = "rust1", since = "1.0.0")]
936 pub fn nul_position(&self) -> usize {
937 self.0
938 }
939
940 /// Consumes this error, returning the underlying vector of bytes which
941 /// generated the error in the first place.
942 ///
943 /// # Examples
944 ///
945 /// ```
946 /// use std::ffi::CString;
947 ///
948 /// let nul_error = CString::new("foo\0bar").unwrap_err();
949 /// assert_eq!(nul_error.into_vec(), b"foo\0bar");
950 /// ```
951 #[must_use = "`self` will be dropped if the result is not used"]
952 #[stable(feature = "rust1", since = "1.0.0")]
953 pub fn into_vec(self) -> Vec<u8> {
954 self.1
955 }
956 }
957
958 #[stable(feature = "rust1", since = "1.0.0")]
959 impl fmt::Display for NulError {
960 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
961 write!(f, "nul byte found in provided data at position: {}", self.0)
962 }
963 }
964
965 #[stable(feature = "cstring_from_vec_with_nul", since = "1.58.0")]
966 impl fmt::Display for FromVecWithNulError {
967 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
968 match self.error_kind {
969 FromBytesWithNulErrorKind::InteriorNul(pos) => {
970 write!(f, "data provided contains an interior nul byte at pos {pos}")
971 }
972 FromBytesWithNulErrorKind::NotNulTerminated => {
973 write!(f, "data provided is not nul terminated")
974 }
975 }
976 }
977 }
978
979 impl IntoStringError {
980 /// Consumes this error, returning original [`CString`] which generated the
981 /// error.
982 #[must_use = "`self` will be dropped if the result is not used"]
983 #[stable(feature = "cstring_into", since = "1.7.0")]
984 pub fn into_cstring(self) -> CString {
985 self.inner
986 }
987
988 /// Access the underlying UTF-8 error that was the cause of this error.
989 #[must_use]
990 #[stable(feature = "cstring_into", since = "1.7.0")]
991 pub fn utf8_error(&self) -> Utf8Error {
992 self.error
993 }
994
995 #[doc(hidden)]
996 #[unstable(feature = "cstr_internals", issue = "none")]
997 pub fn __source(&self) -> &Utf8Error {
998 &self.error
999 }
1000 }
1001
1002 impl IntoStringError {
1003 fn description(&self) -> &str {
1004 "C string contained non-utf8 bytes"
1005 }
1006 }
1007
1008 #[stable(feature = "cstring_into", since = "1.7.0")]
1009 impl fmt::Display for IntoStringError {
1010 #[allow(deprecated, deprecated_in_future)]
1011 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1012 self.description().fmt(f)
1013 }
1014 }
1015
1016 #[stable(feature = "cstr_borrow", since = "1.3.0")]
1017 impl ToOwned for CStr {
1018 type Owned = CString;
1019
1020 fn to_owned(&self) -> CString {
1021 CString { inner: self.to_bytes_with_nul().into() }
1022 }
1023
1024 fn clone_into(&self, target: &mut CString) {
1025 let mut b = into_vec(mem::take(&mut target.inner));
1026 self.to_bytes_with_nul().clone_into(&mut b);
1027 target.inner = b.into_boxed_slice();
1028 }
1029 }
1030
1031 #[stable(feature = "cstring_asref", since = "1.7.0")]
1032 impl From<&CStr> for CString {
1033 fn from(s: &CStr) -> CString {
1034 s.to_owned()
1035 }
1036 }
1037
1038 #[stable(feature = "cstring_asref", since = "1.7.0")]
1039 impl ops::Index<ops::RangeFull> for CString {
1040 type Output = CStr;
1041
1042 #[inline]
1043 fn index(&self, _index: ops::RangeFull) -> &CStr {
1044 self
1045 }
1046 }
1047
1048 #[stable(feature = "cstring_asref", since = "1.7.0")]
1049 impl AsRef<CStr> for CString {
1050 #[inline]
1051 fn as_ref(&self) -> &CStr {
1052 self
1053 }
1054 }
1055
1056 #[cfg(bootstrap)]
1057 #[doc(hidden)]
1058 #[unstable(feature = "cstr_internals", issue = "none")]
1059 pub trait CStrExt {
1060 /// Converts a `CStr` into a <code>[Cow]<[str]></code>.
1061 ///
1062 /// If the contents of the `CStr` are valid UTF-8 data, this
1063 /// function will return a <code>[Cow]::[Borrowed]\(&[str])</code>
1064 /// with the corresponding <code>&[str]</code> slice. Otherwise, it will
1065 /// replace any invalid UTF-8 sequences with
1066 /// [`U+FFFD REPLACEMENT CHARACTER`][U+FFFD] and return a
1067 /// <code>[Cow]::[Owned]\(&[str])</code> with the result.
1068 ///
1069 /// [str]: prim@str "str"
1070 /// [Borrowed]: Cow::Borrowed
1071 /// [Owned]: Cow::Owned
1072 /// [U+FFFD]: crate::char::REPLACEMENT_CHARACTER "std::char::REPLACEMENT_CHARACTER"
1073 ///
1074 /// # Examples
1075 ///
1076 /// Calling `to_string_lossy` on a `CStr` containing valid UTF-8:
1077 ///
1078 /// ```
1079 /// use std::borrow::Cow;
1080 /// use std::ffi::CStr;
1081 ///
1082 /// let cstr = CStr::from_bytes_with_nul(b"Hello World\0")
1083 /// .expect("CStr::from_bytes_with_nul failed");
1084 /// assert_eq!(cstr.to_string_lossy(), Cow::Borrowed("Hello World"));
1085 /// ```
1086 ///
1087 /// Calling `to_string_lossy` on a `CStr` containing invalid UTF-8:
1088 ///
1089 /// ```
1090 /// use std::borrow::Cow;
1091 /// use std::ffi::CStr;
1092 ///
1093 /// let cstr = CStr::from_bytes_with_nul(b"Hello \xF0\x90\x80World\0")
1094 /// .expect("CStr::from_bytes_with_nul failed");
1095 /// assert_eq!(
1096 /// cstr.to_string_lossy(),
1097 /// Cow::Owned(String::from("Hello �World")) as Cow<'_, str>
1098 /// );
1099 /// ```
1100 #[must_use = "this returns the result of the operation, \
1101 without modifying the original"]
1102 #[stable(feature = "cstr_to_str", since = "1.4.0")]
1103 fn to_string_lossy(&self) -> Cow<'_, str>;
1104
1105 /// Converts a <code>[Box]<[CStr]></code> into a [`CString`] without copying or allocating.
1106 ///
1107 /// # Examples
1108 ///
1109 /// ```
1110 /// use std::ffi::CString;
1111 ///
1112 /// let c_string = CString::new(b"foo".to_vec()).expect("CString::new failed");
1113 /// let boxed = c_string.into_boxed_c_str();
1114 /// assert_eq!(boxed.into_c_string(), CString::new("foo").expect("CString::new failed"));
1115 /// ```
1116 #[must_use = "`self` will be dropped if the result is not used"]
1117 #[stable(feature = "into_boxed_c_str", since = "1.20.0")]
1118 fn into_c_string(self: Box<Self>) -> CString;
1119 }
1120
1121 #[cfg(bootstrap)]
1122 #[unstable(feature = "cstr_internals", issue = "none")]
1123 impl CStrExt for CStr {
1124 fn to_string_lossy(&self) -> Cow<'_, str> {
1125 String::from_utf8_lossy(self.to_bytes())
1126 }
1127
1128 fn into_c_string(self: Box<Self>) -> CString {
1129 CString::from(self)
1130 }
1131 }
1132
1133 #[cfg(not(bootstrap))]
1134 #[cfg(not(test))]
1135 impl CStr {
1136 /// Converts a `CStr` into a <code>[Cow]<[str]></code>.
1137 ///
1138 /// If the contents of the `CStr` are valid UTF-8 data, this
1139 /// function will return a <code>[Cow]::[Borrowed]\(&[str])</code>
1140 /// with the corresponding <code>&[str]</code> slice. Otherwise, it will
1141 /// replace any invalid UTF-8 sequences with
1142 /// [`U+FFFD REPLACEMENT CHARACTER`][U+FFFD] and return a
1143 /// <code>[Cow]::[Owned]\(&[str])</code> with the result.
1144 ///
1145 /// [str]: prim@str "str"
1146 /// [Borrowed]: Cow::Borrowed
1147 /// [Owned]: Cow::Owned
1148 /// [U+FFFD]: core::char::REPLACEMENT_CHARACTER "std::char::REPLACEMENT_CHARACTER"
1149 ///
1150 /// # Examples
1151 ///
1152 /// Calling `to_string_lossy` on a `CStr` containing valid UTF-8:
1153 ///
1154 /// ```
1155 /// use std::borrow::Cow;
1156 /// use std::ffi::CStr;
1157 ///
1158 /// let cstr = CStr::from_bytes_with_nul(b"Hello World\0")
1159 /// .expect("CStr::from_bytes_with_nul failed");
1160 /// assert_eq!(cstr.to_string_lossy(), Cow::Borrowed("Hello World"));
1161 /// ```
1162 ///
1163 /// Calling `to_string_lossy` on a `CStr` containing invalid UTF-8:
1164 ///
1165 /// ```
1166 /// use std::borrow::Cow;
1167 /// use std::ffi::CStr;
1168 ///
1169 /// let cstr = CStr::from_bytes_with_nul(b"Hello \xF0\x90\x80World\0")
1170 /// .expect("CStr::from_bytes_with_nul failed");
1171 /// assert_eq!(
1172 /// cstr.to_string_lossy(),
1173 /// Cow::Owned(String::from("Hello �World")) as Cow<'_, str>
1174 /// );
1175 /// ```
1176 #[rustc_allow_incoherent_impl]
1177 #[must_use = "this returns the result of the operation, \
1178 without modifying the original"]
1179 #[stable(feature = "cstr_to_str", since = "1.4.0")]
1180 pub fn to_string_lossy(&self) -> Cow<'_, str> {
1181 String::from_utf8_lossy(self.to_bytes())
1182 }
1183
1184 /// Converts a <code>[Box]<[CStr]></code> into a [`CString`] without copying or allocating.
1185 ///
1186 /// # Examples
1187 ///
1188 /// ```
1189 /// use std::ffi::CString;
1190 ///
1191 /// let c_string = CString::new(b"foo".to_vec()).expect("CString::new failed");
1192 /// let boxed = c_string.into_boxed_c_str();
1193 /// assert_eq!(boxed.into_c_string(), CString::new("foo").expect("CString::new failed"));
1194 /// ```
1195 #[rustc_allow_incoherent_impl]
1196 #[must_use = "`self` will be dropped if the result is not used"]
1197 #[stable(feature = "into_boxed_c_str", since = "1.20.0")]
1198 pub fn into_c_string(self: Box<Self>) -> CString {
1199 CString::from(self)
1200 }
1201 }