beenull/ladybird
1
0
Fork 0
mirror of https://github.com/LadybirdBrowser/ladybird.git synced 2024-11-21 23:20:20 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 6
ladybird/AK/String.cpp

454 lines
13 KiB
C++
Raw Normal View History

AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
/*
* Copyright (c) 2018-2022, Andreas Kling <kling@serenityos.org>
*
* SPDX-License-Identifier: BSD-2-Clause
*/
#include <AK/Checked.h>
AK: Implement FlyString for the new String class This implements a FlyString that will de-duplicate String instances. The FlyString will store the raw encoded data of the String instance: If the String is a short string, FlyString holds the String::ShortString bytes; otherwise FlyString holds a pointer to the Detail::StringData. FlyString itself does not know about String's storage or how to refcount its Detail::StringData. It defers to String to implement these details.
2023-01-11 13:26:49 +00:00
#include <AK/FlyString.h>
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
#include <AK/Format.h>
#include <AK/String.h>
#include <AK/StringBuilder.h>
#include <AK/Utf8View.h>
AK: Add a somewhat naive implementation of String::reverse This will reverse the String's code points (i.e. not just its bytes), but is not aware of grapheme clusters.
2023-01-13 16:34:00 +00:00
#include <AK/Vector.h>
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
#include <stdlib.h>
namespace AK {
namespace Detail {
class StringData final : public RefCounted<StringData> {
public:
static ErrorOr<NonnullRefPtr<StringData>> create_uninitialized(size_t, u8*& buffer);
static ErrorOr<NonnullRefPtr<StringData>> create_substring(StringData const& superstring, size_t start, size_t byte_count);
static ErrorOr<NonnullRefPtr<StringData>> from_utf8(char const* utf8_bytes, size_t);
struct SubstringData {
StringData const* superstring { nullptr };
u32 start_offset { 0 };
};
void operator delete(void* ptr);
~StringData();
SubstringData const& substring_data() const
{
return *reinterpret_cast<SubstringData const*>(m_bytes_or_substring_data);
}
// NOTE: There is no guarantee about null-termination.
ReadonlyBytes bytes() const
{
if (m_substring) {
auto const& data = substring_data();
return data.superstring->bytes().slice(data.start_offset, m_byte_count);
}
return { &m_bytes_or_substring_data[0], m_byte_count };
}
StringView bytes_as_string_view() const { return { bytes() }; }
bool operator==(StringData const& other) const
{
return bytes_as_string_view() == other.bytes_as_string_view();
}
unsigned hash() const
{
if (!m_has_hash)
compute_hash();
return m_hash;
}
AK: Implement FlyString for the new String class This implements a FlyString that will de-duplicate String instances. The FlyString will store the raw encoded data of the String instance: If the String is a short string, FlyString holds the String::ShortString bytes; otherwise FlyString holds a pointer to the Detail::StringData. FlyString itself does not know about String's storage or how to refcount its Detail::StringData. It defers to String to implement these details.
2023-01-11 13:26:49 +00:00
bool is_fly_string() const { return m_is_fly_string; }
void set_fly_string(bool is_fly_string) { m_is_fly_string = is_fly_string; }
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
private:
explicit StringData(size_t byte_count);
StringData(StringData const& superstring, size_t start, size_t byte_count);
void compute_hash() const;
u32 m_byte_count { 0 };
mutable unsigned m_hash { 0 };
mutable bool m_has_hash { false };
bool m_substring { false };
AK: Implement FlyString for the new String class This implements a FlyString that will de-duplicate String instances. The FlyString will store the raw encoded data of the String instance: If the String is a short string, FlyString holds the String::ShortString bytes; otherwise FlyString holds a pointer to the Detail::StringData. FlyString itself does not know about String's storage or how to refcount its Detail::StringData. It defers to String to implement these details.
2023-01-11 13:26:49 +00:00
bool m_is_fly_string { false };
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
u8 m_bytes_or_substring_data[0];
};
void StringData::operator delete(void* ptr)
{
free(ptr);
}
StringData::StringData(size_t byte_count)
: m_byte_count(byte_count)
{
}
StringData::StringData(StringData const& superstring, size_t start, size_t byte_count)
: m_byte_count(byte_count)
, m_substring(true)
{
auto& data = const_cast<SubstringData&>(substring_data());
data.start_offset = start;
data.superstring = &superstring;
superstring.ref();
}
StringData::~StringData()
{
AK: Implement FlyString for the new String class This implements a FlyString that will de-duplicate String instances. The FlyString will store the raw encoded data of the String instance: If the String is a short string, FlyString holds the String::ShortString bytes; otherwise FlyString holds a pointer to the Detail::StringData. FlyString itself does not know about String's storage or how to refcount its Detail::StringData. It defers to String to implement these details.
2023-01-11 13:26:49 +00:00
if (m_is_fly_string)
FlyString::did_destroy_fly_string_data({}, bytes_as_string_view());
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
if (m_substring)
substring_data().superstring->unref();
}
constexpr size_t allocation_size_for_string_data(size_t length)
{
AK: Reduce String's allocated data by one byte This was copied from allocation_size_for_stringimpl, which had to ensure the string is null-terminated. String makes no such guarantee.
2023-01-22 20:04:35 +00:00
return sizeof(StringData) + (sizeof(char) * length);
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
}
ErrorOr<NonnullRefPtr<StringData>> StringData::create_uninitialized(size_t byte_count, u8*& buffer)
{
VERIFY(byte_count);
void* slot = malloc(allocation_size_for_string_data(byte_count));
if (!slot) {
return Error::from_errno(ENOMEM);
}
auto new_string_data = adopt_ref(*new (slot) StringData(byte_count));
buffer = const_cast<u8*>(new_string_data->bytes().data());
return new_string_data;
}
ErrorOr<NonnullRefPtr<StringData>> StringData::from_utf8(char const* utf8_data, size_t byte_count)
{
// Strings of MAX_SHORT_STRING_BYTE_COUNT bytes or less should be handled by the String short string optimization.
VERIFY(byte_count > String::MAX_SHORT_STRING_BYTE_COUNT);
Utf8View view(StringView(utf8_data, byte_count));
if (!view.validate())
return Error::from_string_literal("StringData::from_utf8: Input was not valid UTF-8");
VERIFY(utf8_data);
u8* buffer = nullptr;
auto new_string_data = TRY(create_uninitialized(byte_count, buffer));
memcpy(buffer, utf8_data, byte_count * sizeof(char));
return new_string_data;
}
ErrorOr<NonnullRefPtr<StringData>> StringData::create_substring(StringData const& superstring, size_t start, size_t byte_count)
{
// Strings of MAX_SHORT_STRING_BYTE_COUNT bytes or less should be handled by the String short string optimization.
VERIFY(byte_count > String::MAX_SHORT_STRING_BYTE_COUNT);
void* slot = malloc(sizeof(StringData) + sizeof(StringData::SubstringData));
if (!slot) {
return Error::from_errno(ENOMEM);
}
return adopt_ref(*new (slot) StringData(superstring, start, byte_count));
}
void StringData::compute_hash() const
{
auto bytes = this->bytes();
if (bytes.size() == 0)
m_hash = 0;
else
m_hash = string_hash(reinterpret_cast<char const*>(bytes.data()), bytes.size());
m_has_hash = true;
}
}
String::String(NonnullRefPtr<Detail::StringData> data)
: m_data(&data.leak_ref())
{
}
String::String(String const& other)
: m_data(other.m_data)
{
if (!is_short_string())
m_data->ref();
}
String::String(String&& other)
: m_data(exchange(other.m_data, nullptr))
{
AK: Change the moved-from String state to the empty short string The previous moved-from state was the null string. This violates both our invariant that String is never null, and also the C++ contract that the moved-from state must be valid but unspecified. The empty short string state is of course valid, so it satisfies both invariants. It also allows us to remove any extra checks for the null state. The reason this change is made is primarily because swap() requires moved-from objects to be reassignable (C++ allows this). Because the move assignment of String would not check the null state, it crashed trying to increment the data reference count (nullptr signals a non-short string). This meant that e.g. quick_sort'ing String would crash immediately.
2022-12-11 13:32:56 +00:00
other.m_short_string.byte_count_and_short_string_flag = SHORT_STRING_FLAG;
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
}
String& String::operator=(String&& other)
{
AK: Unref old m_data in String's move assignment We were overridding the data pointer without unreffing it, causing a memory leak when assigning a String.
2022-12-08 17:30:04 +00:00
if (!is_short_string())
m_data->unref();
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
m_data = exchange(other.m_data, nullptr);
AK: Change the moved-from String state to the empty short string The previous moved-from state was the null string. This violates both our invariant that String is never null, and also the C++ contract that the moved-from state must be valid but unspecified. The empty short string state is of course valid, so it satisfies both invariants. It also allows us to remove any extra checks for the null state. The reason this change is made is primarily because swap() requires moved-from objects to be reassignable (C++ allows this). Because the move assignment of String would not check the null state, it crashed trying to increment the data reference count (nullptr signals a non-short string). This meant that e.g. quick_sort'ing String would crash immediately.
2022-12-11 13:32:56 +00:00
other.m_short_string.byte_count_and_short_string_flag = SHORT_STRING_FLAG;
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
return *this;
}
String& String::operator=(String const& other)
{
if (&other != this) {
m_data = other.m_data;
if (!is_short_string())
m_data->ref();
}
return *this;
}
AK: Support creating known short string literals at compile time In cases where we know a string literal will fit in the short string storage, we can do so at compile time without needing to handle error propagation. If the provided string literal is too long, a compilation error will be emitted due to the failed VERIFY statement being a non- constant expression.
2023-01-20 12:20:01 +00:00
void String::destroy_string()
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
{
AK: Change the moved-from String state to the empty short string The previous moved-from state was the null string. This violates both our invariant that String is never null, and also the C++ contract that the moved-from state must be valid but unspecified. The empty short string state is of course valid, so it satisfies both invariants. It also allows us to remove any extra checks for the null state. The reason this change is made is primarily because swap() requires moved-from objects to be reassignable (C++ allows this). Because the move assignment of String would not check the null state, it crashed trying to increment the data reference count (nullptr signals a non-short string). This meant that e.g. quick_sort'ing String would crash immediately.
2022-12-11 13:32:56 +00:00
if (!is_short_string())
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
m_data->unref();
}
ErrorOr<String> String::from_utf8(StringView view)
{
if (view.length() <= MAX_SHORT_STRING_BYTE_COUNT) {
ShortString short_string;
if (!view.is_empty())
memcpy(short_string.storage, view.characters_without_null_termination(), view.length());
short_string.byte_count_and_short_string_flag = (view.length() << 1) | SHORT_STRING_FLAG;
return String { short_string };
}
auto data = TRY(Detail::StringData::from_utf8(view.characters_without_null_termination(), view.length()));
return String { move(data) };
}
StringView String::bytes_as_string_view() const
{
return StringView(bytes());
}
ReadonlyBytes String::bytes() const
{
if (is_short_string())
return m_short_string.bytes();
return m_data->bytes();
}
bool String::is_empty() const
{
return bytes().size() == 0;
}
ErrorOr<String> String::vformatted(StringView fmtstr, TypeErasedFormatParams& params)
{
StringBuilder builder;
TRY(vformat(builder, fmtstr, params));
return builder.to_string();
}
AK: Add `split()` for `String`
2023-01-16 16:12:53 +00:00
ErrorOr<Vector<String>> String::split(u32 separator, SplitBehavior split_behavior) const
{
return split_limit(separator, 0, split_behavior);
}
ErrorOr<Vector<String>> String::split_limit(u32 separator, size_t limit, SplitBehavior split_behavior) const
{
Vector<String> result;
if (is_empty())
return result;
bool keep_empty = has_flag(split_behavior, SplitBehavior::KeepEmpty);
size_t substring_start = 0;
for (auto it = code_points().begin(); it != code_points().end() && (result.size() + 1) != limit; ++it) {
u32 code_point = *it;
if (code_point == separator) {
size_t substring_length = code_points().iterator_offset(it) - substring_start;
if (substring_length != 0 || keep_empty)
TRY(result.try_append(TRY(substring_from_byte_offset_with_shared_superstring(substring_start, substring_length))));
substring_start = code_points().iterator_offset(it) + it.underlying_code_point_length_in_bytes();
}
}
size_t tail_length = code_points().byte_length() - substring_start;
if (tail_length != 0 || keep_empty)
TRY(result.try_append(TRY(substring_from_byte_offset_with_shared_superstring(substring_start, tail_length))));
return result;
}
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
bool String::operator==(String const& other) const
{
if (is_short_string())
return m_data == other.m_data;
return bytes_as_string_view() == other.bytes_as_string_view();
}
AK: Implement FlyString for the new String class This implements a FlyString that will de-duplicate String instances. The FlyString will store the raw encoded data of the String instance: If the String is a short string, FlyString holds the String::ShortString bytes; otherwise FlyString holds a pointer to the Detail::StringData. FlyString itself does not know about String's storage or how to refcount its Detail::StringData. It defers to String to implement these details.
2023-01-11 13:26:49 +00:00
bool String::operator==(FlyString const& other) const
{
if (reinterpret_cast<uintptr_t>(m_data) == other.data({}))
return true;
return bytes_as_string_view() == other.bytes_as_string_view();
}
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
bool String::operator==(StringView other) const
{
return bytes_as_string_view() == other;
}
ErrorOr<String> String::substring_from_byte_offset(size_t start, size_t byte_count) const
{
if (!byte_count)
return String {};
return String::from_utf8(bytes_as_string_view().substring_view(start, byte_count));
}
ErrorOr<String> String::substring_from_byte_offset_with_shared_superstring(size_t start, size_t byte_count) const
{
if (!byte_count)
return String {};
if (byte_count <= MAX_SHORT_STRING_BYTE_COUNT)
return String::from_utf8(bytes_as_string_view().substring_view(start, byte_count));
return String { TRY(Detail::StringData::create_substring(*m_data, start, byte_count)) };
}
bool String::operator==(char const* c_string) const
{
return bytes_as_string_view() == c_string;
}
u32 String::hash() const
{
if (is_short_string()) {
auto bytes = this->bytes();
return string_hash(reinterpret_cast<char const*>(bytes.data()), bytes.size());
}
return m_data->hash();
}
Utf8View String::code_points() const
{
return Utf8View(bytes_as_string_view());
}
ErrorOr<void> Formatter<String>::format(FormatBuilder& builder, String const& utf8_string)
{
return Formatter<StringView>::format(builder, utf8_string.bytes_as_string_view());
}
ErrorOr<String> String::replace(StringView needle, StringView replacement, ReplaceMode replace_mode) const
{
return StringUtils::replace(*this, needle, replacement, replace_mode);
}
AK: Add a somewhat naive implementation of String::reverse This will reverse the String's code points (i.e. not just its bytes), but is not aware of grapheme clusters.
2023-01-13 16:34:00 +00:00
ErrorOr<String> String::reverse() const
{
// FIXME: This handles multi-byte code points, but not e.g. grapheme clusters.
// FIXME: We could avoid allocating a temporary vector if Utf8View supports reverse iteration.
auto code_point_length = code_points().length();
Vector<u32> code_points;
TRY(code_points.try_ensure_capacity(code_point_length));
for (auto code_point : this->code_points())
code_points.unchecked_append(code_point);
auto builder = TRY(StringBuilder::create(code_point_length * sizeof(u32)));
while (!code_points.is_empty())
TRY(builder.try_append_code_point(code_points.take_last()));
return builder.to_string();
}
AK: Add String::contains
2023-01-14 15:17:32 +00:00
bool String::contains(StringView needle, CaseSensitivity case_sensitivity) const
{
return StringUtils::contains(bytes_as_string_view(), needle, case_sensitivity);
}
bool String::contains(char needle, CaseSensitivity case_sensitivity) const
{
return contains(StringView { &needle, 1 }, case_sensitivity);
}
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
bool String::is_short_string() const
{
AK: Implement FlyString for the new String class This implements a FlyString that will de-duplicate String instances. The FlyString will store the raw encoded data of the String instance: If the String is a short string, FlyString holds the String::ShortString bytes; otherwise FlyString holds a pointer to the Detail::StringData. FlyString itself does not know about String's storage or how to refcount its Detail::StringData. It defers to String to implement these details.
2023-01-11 13:26:49 +00:00
return has_short_string_bit(reinterpret_cast<uintptr_t>(m_data));
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
}
ReadonlyBytes String::ShortString::bytes() const
{
return { storage, byte_count() };
}
size_t String::ShortString::byte_count() const
{
return byte_count_and_short_string_flag >> 1;
}
unsigned Traits<String>::hash(String const& string)
{
return string.hash();
}
AK: Implement FlyString for the new String class This implements a FlyString that will de-duplicate String instances. The FlyString will store the raw encoded data of the String instance: If the String is a short string, FlyString holds the String::ShortString bytes; otherwise FlyString holds a pointer to the Detail::StringData. FlyString itself does not know about String's storage or how to refcount its Detail::StringData. It defers to String to implement these details.
2023-01-11 13:26:49 +00:00
String String::fly_string_data_to_string(Badge<FlyString>, uintptr_t const& data)
{
if (has_short_string_bit(data))
return String { *reinterpret_cast<ShortString const*>(&data) };
auto const* string_data = reinterpret_cast<Detail::StringData const*>(data);
return String { NonnullRefPtr<Detail::StringData>(*string_data) };
}
StringView String::fly_string_data_to_string_view(Badge<FlyString>, uintptr_t const& data)
{
if (has_short_string_bit(data)) {
auto const* short_string = reinterpret_cast<ShortString const*>(&data);
return short_string->bytes();
}
auto const* string_data = reinterpret_cast<Detail::StringData const*>(data);
return string_data->bytes_as_string_view();
}
uintptr_t String::to_fly_string_data(Badge<FlyString>) const
{
return reinterpret_cast<uintptr_t>(m_data);
}
void String::ref_fly_string_data(Badge<FlyString>, uintptr_t data)
{
if (has_short_string_bit(data))
return;
auto const* string_data = reinterpret_cast<Detail::StringData const*>(data);
string_data->ref();
}
void String::unref_fly_string_data(Badge<FlyString>, uintptr_t data)
{
if (has_short_string_bit(data))
return;
auto const* string_data = reinterpret_cast<Detail::StringData const*>(data);
string_data->unref();
}
void String::did_create_fly_string(Badge<FlyString>) const
{
VERIFY(!is_short_string());
m_data->set_fly_string(true);
}
AK: Introduce the new String, replacement for DeprecatedString DeprecatedString (formerly String) has been with us since the start, and it has served us well. However, it has a number of shortcomings that I'd like to address. Some of these issues are hard if not impossible to solve incrementally inside of DeprecatedString, so instead of doing that, let's build a new String class and then incrementally move over to it instead. Problems in DeprecatedString: - It assumes string allocation never fails. This makes it impossible to use in allocation-sensitive contexts, and is the reason we had to ban DeprecatedString from the kernel entirely. - The awkward null state. DeprecatedString can be null. It's different from the empty state, although null strings are considered empty. All code is immediately nicer when using Optional<DeprecatedString> but DeprecatedString came before Optional, which is how we ended up like this. - The encoding of the underlying data is ambiguous. For the most part, we use it as if it's always UTF-8, but there have been cases where we pass around strings in other encodings (e.g ISO8859-1) - operator[] and length() are used to iterate over DeprecatedString one byte at a time. This is done all over the codebase, and will *not* give the right results unless the string is all ASCII. How we solve these issues in the new String: - Functions that may allocate now return ErrorOr<String> so that ENOMEM errors can be passed to the caller. - String has no null state. Use Optional<String> when needed. - String is always UTF-8. This is validated when constructing a String. We may need to add a bypass for this in the future, for cases where you have a known-good string, but for now: validate all the things! - There is no operator[] or length(). You can get the underlying data with bytes(), but for iterating over code points, you should be using an UTF-8 iterator. Furthermore, it has two nifty new features: - String implements a small string optimization (SSO) for strings that can fit entirely within a pointer. This means up to 3 bytes on 32-bit platforms, and 7 bytes on 64-bit platforms. Such small strings will not be heap-allocated. - String can create substrings without making a deep copy of the substring. Instead, the superstring gets +1 refcount from the substring, and it acts like a view into the superstring. To make substrings like this, use the substring_with_shared_superstring() API. One caveat: - String does not guarantee that the underlying data is null-terminated like DeprecatedString does today. While this was nifty in a handful of places where we were calling C functions, it did stand in the way of shared-superstring substrings.
2022-12-01 12:27:43 +00:00
DeprecatedString String::to_deprecated_string() const
{
return DeprecatedString(bytes_as_string_view());
}
ErrorOr<String> String::from_deprecated_string(DeprecatedString const& deprecated_string)
{
Utf8View view(deprecated_string);
if (!view.validate())
return Error::from_string_literal("String::from_deprecated_string: Input was not valid UTF-8");
return String::from_utf8(deprecated_string.view());
}
}
Reference in a new issue Copy permalink