You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

605 lines
20 KiB
Rust

use core::{
borrow,
cmp::Ordering,
fmt,
hash::{Hash, Hasher},
mem::MaybeUninit,
ops, ptr, str,
};
#[cfg(not(feature = "std"))]
use core::convert::TryFrom;
#[cfg(all(feature = "alloc", not(feature = "std")))]
use alloc::{
borrow::{Cow, ToOwned},
boxed::Box,
string::String,
};
#[cfg(feature = "std")]
use std::borrow::Cow;
#[derive(Debug)]
pub struct CapacityError;
impl fmt::Display for CapacityError {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(f, "`CappedString` capacity exceeded")
}
}
#[cfg(feature = "std")]
impl std::error::Error for CapacityError {}
/// A string type which stores at most `N` bytes of string data. The string data is stored inline
/// rather than using a heap allocation.
///
/// ```
/// # use libshire::strings::CappedString;
/// # fn main() -> Result<(), libshire::strings::capped::Error> {
/// let s = CappedString::<16>::new("hello world")?;
/// assert_eq!(&*s, "hello world");
/// # Ok(())
/// # }
/// ```
#[derive(Clone)]
pub struct CappedString<const N: usize> {
buf: [MaybeUninit<u8>; N],
len: u8,
}
impl<const N: usize> CappedString<N> {
const MAX_LEN: u8 = {
#[allow(clippy::cast_possible_truncation, clippy::checked_conversions)]
if N <= u8::MAX as usize {
N as u8
} else {
panic!("`N` must be within the bounds of `u8`")
}
};
/// Creates a new `CappedString` from a given byte buffer and length.
///
/// # Safety
/// The first `len` bytes of `buf` (i.e. `buf[..len]`) must be initialised and valid UTF-8.
/// `len` must be less than or equal to `N`.
#[inline]
#[must_use]
pub const unsafe fn from_raw_parts(buf: [MaybeUninit<u8>; N], len: u8) -> Self {
Self { buf, len }
}
#[inline]
#[must_use]
pub const fn into_raw_parts(self) -> ([MaybeUninit<u8>; N], u8) {
(self.buf, self.len)
}
/// # Safety
/// `src` must point to `len` bytes of valid, UTF-8 string data. `len` must be less than or
/// equal to `N`.
#[inline]
unsafe fn from_raw_ptr(src: *const u8, len: u8) -> Self {
// `u8` has the same memory layout as `MaybeUninit<u8>`, so this cast is valid.
let src = src as *const MaybeUninit<u8>;
// SAFETY:
// `MaybeUninit::uninit()` is a valid value for `[MaybeUninit<u8>; N]`, since each element
// of the array is allowed to be uninitialised.
let mut buf = unsafe { MaybeUninit::<[MaybeUninit<u8>; N]>::uninit().assume_init() };
// SAFETY:
// The caller is responsible for ensuring that `src` points to a valid string, which means
// that it must not overlap with the new local variable `buf`. The caller is responsible
// for ensuring that `src` is valid for reads of `len` bytes. The caller is responsible for
// ensuring that `len <= N`, so `buf` is valid for writes of `len` bytes. `src` and `buf`
// are both trivially properly aligned, since they both have an alignment of 1.
unsafe { ptr::copy_nonoverlapping(src, buf.as_mut_ptr(), usize::from(len)); }
// SAFETY:
// The caller is responsible for ensuring that `src` points to `len` bytes of valid UTF-8
// data. `src` is copied into the start of `buf`, so the first `len` bytes of `buf` are
// valid UTF-8. The caller is responsible for ensuring that `len <= N`.
unsafe { Self::from_raw_parts(buf, len) }
}
/// # Safety
/// `self.len` must be less than `N`, so that there is space in the buffer to append the byte.
#[inline]
unsafe fn append_byte(&mut self, byte: u8) {
// SAFETY:
// The caller is responsible for ensuring that `self.len < N`.
let dst = unsafe { self.buf.get_unchecked_mut(usize::from(self.len)) };
*dst = MaybeUninit::new(byte);
self.len += 1;
}
/// # Safety
/// `src` must point to `len` bytes of valid UTF-8 string data. `len` must be less than or equal
/// to `N - self.len`.
#[inline]
unsafe fn append_bytes(&mut self, src: *const u8, len: u8) {
// `u8` has the same memory layout as `MaybeUninit<u8>`, so this cast is valid.
let src = src as *const MaybeUninit<u8>;
// SAFETY:
// `self.len <= N` is an invariant of `CappedString`, so `self.len..` is a valid range over
// `self.buf`.
let dst = unsafe { self.buf.get_unchecked_mut(usize::from(self.len)..) };
// SAFETY:
// The caller is responsible for ensuring that `src` points to a valid string, which means
// that it cannot overlap with the new local variable `buf`. The caller is responsible for
// ensuring that `src` is valid for reads of `len` bytes. The caller is responsible for
// ensuring that `len <= N - self.len`, so the destination `dst = self.buf[self.len..]` is
// valid for writes of `len` bytes. `src` and `dst` are both trivially properly aligned,
// since they both have an alignment of 1.
unsafe { ptr::copy_nonoverlapping(src, dst.as_mut_ptr(), usize::from(len)); }
self.len += len;
}
/// Returns a new empty `CappedString`.
///
/// ```
/// # use libshire::strings::CappedString;
/// let s = CappedString::<8>::empty();
/// assert!(s.is_empty());
/// assert_eq!(s.len(), 0);
/// assert_eq!(&*s, "");
/// ```
#[inline]
#[must_use]
pub const fn empty() -> Self {
// SAFETY:
// `MaybeUninit::uninit()` is a valid value for `[MaybeUninit<u8>; N]`, since each element
// of the array is allowed to be uninitialised.
let buf = unsafe { MaybeUninit::<[MaybeUninit<u8>; N]>::uninit().assume_init() };
// SAFETY:
// It is vacuously true that the first 0 bytes of the buffer are initialised and valid
// UTF-8.
unsafe { Self::from_raw_parts(buf, 0) }
}
/// Returns a new `CappedString` containing the given string data. The string data will be
/// stored inline; no heap allocation is used. An error will be returned if the length of the
/// provided string exceeds the `CappedString`'s maximum length, `N`.
///
/// ```
/// # use libshire::strings::CappedString;
/// # fn main() -> Result<(), libshire::strings::capped::Error> {
/// let s = CappedString::<16>::new("hello world")?;
/// assert_eq!(&*s, "hello world");
/// # Ok(())
/// # }
/// ```
#[inline]
pub fn new<S>(src: &S) -> Result<Self, CapacityError>
where
S: AsRef<str> + ?Sized,
{
// Convert the string to a byte slice, which is guaranteed to be valid UTF-8 since this is
// an invariant of `str`.
let src = <S as AsRef<str>>::as_ref(src);
// If the length of the `src` string does not fit into a `u8` or is greater than
// `Self::MAX_LEN`, we can't fit it into the new `CappedString` so return an error.
let len = match u8::try_from(src.len()) {
Ok(len) if len <= Self::MAX_LEN => len,
_ => return Err(CapacityError),
};
// SAFETY:
// `src.as_ptr()` points to `len` bytes of valid UTF-8 string data since `src` is a `&str`
// and `len` is its length. `len` is less than or equal to `Self::MAX_LEN`, which is equal
// to `N`.
unsafe { Ok(Self::from_raw_ptr(src.as_ptr(), len)) }
}
#[inline]
#[must_use]
pub fn new_truncating<S>(src: &S) -> Self
where
S: AsRef<str> + ?Sized,
{
let src = <S as AsRef<str>>::as_ref(src);
let (src, len) = truncate_str(src, Self::MAX_LEN);
// SAFETY:
// It is part of the contract of `truncate_str` that it returns a pointer to a valid UTF-8
// string of length `len`, and that `len` is less than or equal to the provided maximum
// length, which is `Self::MAX_LEN` (which is equal to `N`) in this case.
unsafe { Self::from_raw_ptr(src, len) }
}
/// Appends the given character to the end of this `CappedString`, returning an error if there
/// is insufficient capacity remaining to do so.
///
/// If you do not care whether or not the append succeeds, see [`Self::push_truncating`].
#[inline]
pub fn push(&mut self, c: char) -> Result<(), CapacityError> {
let mut char_buf = [0u8; 4];
let encoded = c.encode_utf8(&mut char_buf);
match encoded.len() {
1 => {
if self.len == Self::MAX_LEN {
return Err(CapacityError);
}
// SAFETY:
//
unsafe { self.append_byte(encoded.as_bytes()[0]) }
Ok(())
},
_ => self.push_str(encoded),
}
}
/// Appends the given character to the end of this `CappedString`, failing silently if there is
/// insufficient capacity remaining to do so.
///
/// If you would like to know whether or not the append succeeds, see [`Self::push`].
#[inline]
pub fn push_truncating(&mut self, c: char) {
// Unlike `Self::push_str_truncating`, we can just use `Self::push` and swallow the error
// because a single character will never be partially pushed; it is either pushed or it
// isn't.
self.push(c).ok();
}
/// Appends the given string slice to the end of this `CappedString`, returning an error if
/// there is insufficient capacity remaining to do so.
///
/// If you would like a version which cannot fail, see [`Self::push_str_truncating`].
#[inline]
pub fn push_str<S>(&mut self, src: &S) -> Result<(), CapacityError>
where
S: AsRef<str> + ?Sized,
{
let src = <S as AsRef<str>>::as_ref(src);
let len = match u8::try_from(src.len()) {
Ok(len) if len <= Self::MAX_LEN - self.len => len,
_ => return Err(CapacityError),
};
// SAFETY:
// `src` is a valid string slice with length `len`. We have checked that
// `len <= N - self.len` holds above (note that `Self::MAX_LEN == N`).
unsafe { self.append_bytes(src.as_ptr(), len); }
Ok(())
}
/// Appends as many of the characters of the given string slice to the end of this
/// `CappedString` as can fit. Any remaining characters will not be added.
///
/// If you would like a version which returns an error if there is not enough capacity remaining
/// to append the entire string slice, see [`Self::push_str`].
#[inline]
pub fn push_str_truncating<S>(&mut self, src: &S)
where
S: AsRef<str> + ?Sized,
{
let remaining_cap = Self::MAX_LEN - self.len;
// Short-circuit if we have no space left to copy into.
if remaining_cap == 0 {
return;
}
let src = <S as AsRef<str>>::as_ref(src);
// Find the longest valid UTF-8 prefix which fits into the remaining space.
let (src, len) = truncate_str(src, remaining_cap);
// SAFETY:
// `truncate_str` returns a pointer to `len` bytes of valid UTF-8 string data. The returned
// `len` will always be less than or equal to `remaining_cap`, which is equal to
// `N - self.len` (note that `Self::MAX_LEN == N`).
unsafe { self.append_bytes(src, len); }
}
/// Returns a string slice pointing to the underlying string data.
#[inline]
#[must_use]
pub fn as_str(&self) -> &str {
// SAFETY:
// The first `self.len` bytes of `self.buf` (which is returned by `Self::as_bytes`) being
// valid UTF-8 is an invariant of `CappedString`.
unsafe { str::from_utf8_unchecked(self.as_bytes()) }
}
/// Returns a mutable string slice pointing to the underlying string data.
#[inline]
#[must_use]
pub fn as_str_mut(&mut self) -> &mut str {
// SAFETY:
// The first `self.len` bytes of `self.buf` (which is returned by `Self::as_bytes_mut`)
// being valid UTF-8 is an invariant of `CappedString`.
unsafe { str::from_utf8_unchecked_mut(self.as_bytes_mut()) }
}
#[inline]
#[must_use]
pub fn as_bytes(&self) -> &[u8] {
// Get the slice of the buffer containing initialised string data.
// SAFETY:
// It is an invariant of `CappedString` that `self.len <= N`, so `..self.len` is a valid
// range over `self.buf`.
let data_slice = unsafe { self.buf.get_unchecked(..usize::from(self.len)) };
// Convert the `&[MaybeUninit<u8>]` to a `&[u8]`.
// SAFETY:
// `MaybeUninit<u8>` has the same memory layout as `u8`, and the first `self.len` bytes of
// the buffer are initialised, so this conversion is valid.
unsafe { &*(data_slice as *const [MaybeUninit<u8>] as *const [u8]) }
}
/// # Safety
/// The caller is responsible for ensuring that the slice is valid UTF-8 when the mutable
/// borrow ends.
#[inline]
#[must_use]
pub unsafe fn as_bytes_mut(&mut self) -> &mut [u8] {
// Get the slice of the buffer containing initialised string data.
// SAFETY:
// It is an invariant of `CappedString` that `self.len <= N`, so `..self.len` is a valid
// range over `self.buf`.
let data_slice = unsafe { self.buf.get_unchecked_mut(..usize::from(self.len)) };
// Convert the `&[MaybeUninit<u8>]` to a `&[u8]`.
// SAFETY:
// `MaybeUninit<u8>` has the same memory layout as `u8`, and the first `self.len` bytes of
// the buffer are initialised, so this conversion is valid.
unsafe { &mut *(data_slice as *mut [MaybeUninit<u8>] as *mut [u8]) }
}
#[inline]
#[must_use]
pub fn len(&self) -> usize {
usize::from(self.len)
}
#[inline]
#[must_use]
pub fn is_empty(&self) -> bool {
self.len == 0
}
}
#[cfg(feature = "alloc")]
impl<const N: usize> CappedString<N> {
#[inline]
#[must_use]
pub fn into_boxed_str(self) -> Box<str> {
self.as_str().into()
}
#[inline]
#[must_use]
pub fn into_string(self) -> String {
self.as_str().to_owned()
}
}
impl<const N: usize> Default for CappedString<N> {
#[inline]
fn default() -> Self {
Self::empty()
}
}
impl<const N: usize> ops::Deref for CappedString<N> {
type Target = str;
#[inline]
fn deref(&self) -> &Self::Target {
self.as_str()
}
}
impl<const N: usize> ops::DerefMut for CappedString<N> {
#[inline]
fn deref_mut(&mut self) -> &mut Self::Target {
self.as_str_mut()
}
}
impl<const N: usize> AsRef<str> for CappedString<N> {
#[inline]
fn as_ref(&self) -> &str {
self
}
}
impl<const N: usize> AsMut<str> for CappedString<N> {
#[inline]
fn as_mut(&mut self) -> &mut str {
self
}
}
impl<const N: usize> borrow::Borrow<str> for CappedString<N> {
#[inline]
fn borrow(&self) -> &str {
self
}
}
impl<const N: usize> borrow::BorrowMut<str> for CappedString<N> {
#[inline]
fn borrow_mut(&mut self) -> &mut str {
self
}
}
impl<'a, const N: usize> TryFrom<&'a str> for CappedString<N> {
type Error = CapacityError;
#[inline]
fn try_from(s: &'a str) -> Result<Self, Self::Error> {
Self::new(s)
}
}
#[cfg(feature = "alloc")]
impl<const N: usize> TryFrom<String> for CappedString<N> {
type Error = CapacityError;
#[inline]
fn try_from(s: String) -> Result<Self, Self::Error> {
Self::new(&s)
}
}
#[cfg(feature = "alloc")]
impl<const N: usize> TryFrom<Box<str>> for CappedString<N> {
type Error = CapacityError;
#[inline]
fn try_from(s: Box<str>) -> Result<Self, Self::Error> {
Self::new(&s)
}
}
#[cfg(feature = "alloc")]
impl<'a, const N: usize> TryFrom<Cow<'a, str>> for CappedString<N> {
type Error = CapacityError;
#[inline]
fn try_from(s: Cow<'a, str>) -> Result<Self, Self::Error> {
Self::new(&s)
}
}
#[cfg(feature = "alloc")]
impl<const N: usize> From<CappedString<N>> for String {
#[inline]
fn from(s: CappedString<N>) -> Self {
s.into_string()
}
}
#[cfg(feature = "alloc")]
impl<const N: usize> From<CappedString<N>> for Box<str> {
#[inline]
fn from(s: CappedString<N>) -> Self {
s.into_boxed_str()
}
}
impl<const N: usize, const M: usize> PartialEq<CappedString<M>> for CappedString<N> {
#[inline]
fn eq(&self, other: &CappedString<M>) -> bool {
**self == **other
}
}
impl<const N: usize> Eq for CappedString<N> {}
impl<const N: usize, const M: usize> PartialOrd<CappedString<M>> for CappedString<N> {
#[inline]
fn partial_cmp(&self, other: &CappedString<M>) -> Option<Ordering> {
(**self).partial_cmp(&**other)
}
}
impl<const N: usize> Ord for CappedString<N> {
#[inline]
fn cmp(&self, other: &Self) -> Ordering {
(**self).cmp(&**other)
}
}
impl<const N: usize> Hash for CappedString<N> {
#[inline]
fn hash<H: Hasher>(&self, state: &mut H) {
(**self).hash(state);
}
}
impl<const N: usize> str::FromStr for CappedString<N> {
type Err = CapacityError;
#[inline]
fn from_str(s: &str) -> Result<Self, Self::Err> {
Self::new(s)
}
}
impl<const N: usize> fmt::Debug for CappedString<N> {
#[inline]
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
fmt::Debug::fmt(&**self, f)
}
}
impl<const N: usize> fmt::Display for CappedString<N> {
#[inline]
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
fmt::Display::fmt(&**self, f)
}
}
/// Returns a pointer to the longest prefix of `src` which is valid UTF-8 and whose length is
/// less than or equal to `max_len`, and returns the length of this prefix.
#[inline]
fn truncate_str(src: &str, max_len: u8) -> (*const u8, u8) {
match u8::try_from(src.len()) {
// If the length of the `src` string is less than or equal to `max_len`, there is no need to
// truncate it.
Ok(src_len) if src_len <= max_len => (src.as_ptr(), src_len),
// If the length of `src` is greater than `max_len`, we need to truncate it. Note that
// `u8::try_from` returning an error means that `src.len() > max_len`, since `max_len` is a
// `u8` and `src.len()` is a `usize`.
_ => {
let src = src.as_bytes();
let mut i = max_len;
// Find the rightmost codepoint which starts at an index less than or equal to
// `max_len`. Everything to the left of this will be valid UTF-8 with a length less
// than or equal to `max_len`. We only need to do 3 iterations because codepoints have
// a maximum length of 4 bytes.
for _ in 0..3 {
// The first byte in the string must always be the start of a codepoint.
if i == 0 {
break;
}
// SAFETY:
// `i <= max_len`, since it is never incremented. If this branch is run, then either
// `src.len(): usize` does not fit into a `u8`, in which case it must be greater
// than `max_len: u8`, or it does fit into a `u8` but it is greater than `max_len`.
// Therefore, `src.len() > max_len` must hold. Substitution gives `i < src.len()`,
// so `i` is a valid index into `src`.
let byte = unsafe { *src.get_unchecked(usize::from(i)) };
// If the byte is not of the form 0b10xxxxxx, then it is the start of a codepoint.
if byte & 0xc0 != 0x80 {
break;
}
i -= 1;
}
// `i < src.len()` always holds as discussed above, so the pointer `src.as_ptr()` is
// valid for reads of `i` bytes. `i` is the index of the start of a codepoint, and
// codepoints are contiguous, so the `i` bytes being pointed to must be valid UTF-8.
(src.as_ptr(), i)
},
}
}
#[cfg(test)]
mod tests {}