Files
Cubed/third_party/asio/include/asio/basic_writable_pipe.hpp
zhenyan121 30f843ba6b feature: multiplayer (#25)
* build: add asio library

* feat(dev_panel): add Asio to about table

* refactor(gameplay): split Chunk into server/client variants and add networking

* feat(gameplay): add Session class and UUID generation utility

* feat(network): integrate protobuf for player sync and session management

Add Protobuf dependency, define proto messages for player requests, positions, and chunk data. Refactor Session to use strand and async write. Implement player join/exit and position sync in ServerWorld.

* refactor(proto): restructure protobuf definitions and update build system

* feat(protocol): add packet serialization and login handling

Introduce packet header and ID mapping for protobuf messages.
Refactor session and server_world to use new packet wrapper.
Fix missing semicolons in proto files.

* refactor(packet): serialize directly into packet buffer with ByteSizeLong

* feat(world): add player-based chunk loading and UUID support

Restructure world generation to trigger chunk loading based on player
movement. Replace player name with UUID for identification. Implement
chunk request/response protocol. Improve thread management for gen
thread.

* feat(server): add block change packet handling and increase reserved threads

* refactor(server): enhance thread safety and session management

* fix(session): catch std::exception in read_loop and log error

* fix(network-server): stop accept loop on shutdown flag

* fix(gameplay): correct unknown error log condition and missing chunk pos update

* feat(client): add client-side chunk, player, world and network classes

* feat(client-chunk): add greedy meshing, receive_chunk, and biome field

* feat(client_player): add name constructor param, remove chunk transition

* feat(gameplay): integrate network client and thread-safe player pos

* feat(gameplay): add block change sync and coordinate utils

* refactor(server): replace Chunk with ServerChunk

* refactor(gameplay): split world into client and server

Remove monolithic World, Player, Chunk classes. Introduce ClientWorld, ServerWorld, ClientPlayer, and related networked components. Add Abseil dependency for logging and checks. Rename old files to pre_remove_* and update all includes and references accordingly.

* feat(gameplay): add time sync and network improvements

* fix: correct neighbor block indexing and add time update support

* feat(gameplay): optimize chunk request with player chunk tracking

Add `update_player_chunk()` method to `ClientPlayer` that triggers a chunk request when the player moves more than 2 chunks away from the last recorded chunk. Introduce an atomic `m_requesting_chunk` flag in `ClientWorld` to prevent concurrent requests. Rename `m_player_chunk_pos` to `m_last_chunk_pos` for clarity.

* refactor(client_world): split pending queue into data and upload queues

* feat(client): add center-based chunk loading priority

* feat(app): add client/server CLI arguments and init guards

* feat(multiplayer): add rendering of other players as colored cubes

* feat(gameplay): add thread pool for asynchronous chunk data processing

* feat(gameplay): add player logout and disconnect handling

* fix(dev_panel): add protobuf to about table libraries

* build: add ZLIB dependency

* refactor(packet): replace zlib with zstd compression and restructure header

Increase header length to 12 bytes, add CompressType and PacketHeader struct, and implement decode_packet_header. Update CMake to find zstd and link against it, adding Findzstd module.

* refactor(network): extract packet header decoding into function

* feat(packet): implement zstd compression and refactor packet decoding

* perf(gameplay): optimize protobuf message allocation with Arena and limit thread pool size

Switch to Arena allocation for protobuf messages to reduce heap allocations. Pass ChunkDataRsp by value to enable move semantics. Cap thread pool size to hardware concurrency with max 4 threads.

* refactor(gameplay): use steady clock and sleep_until for tick loop timing

* refactor(gameplay): rename remote player types and add interpolation

* refactor(server_world): use thread pool for chunk request handling

* refactor(packet): replace if-constexpr chain with explicit template specializations

Refactor `get_packet_id()` by removing the if-constexpr chain and instead
providing explicit specializations for each packet type. This simplifies the
primary template (which now only contains a static assertion) and improves
compile-time dispatch clarity. The `always_false` helper is no longer needed
in the primary template.

* feat(gameplay): add packet to clear all chunks on server rebuild

Add new packet S2C_ClearAllChunks (ID 3005) to notify clients when the server rebuilds its world. Clients respond by clearing their chunk cache and requesting fresh chunks. Shift UPDATE_TIME packet ID to 3006 to accommodate.

* feat(dev_panel): split world tab; add chunk request logging and fix flag reset

* feat(tools): add RecentQueue to replace std::deque in server_world

* feat(gameplay): add task ID to chunk requests to discard stale responses

* refactor(block): move block source to gameplay directory

* feat(toml): add shared TOML utilities and handle ip.toml

* feat: add player name configuration and rename config file

- Add --player command-line argument and 'player' field in Arguments struct
- Rename ip.toml to server.toml and corresponding internal variable to 'server'
- Fix block.cpp source path in CMakeLists.txt

* build(cmake): modularize build configuration

* fix: build fail on windows

* refactor(renderer): rename Logger level enum value DEBUG to L_DEBUG

* refactor(world): rework chunk state machine and player chunk tracking

Introduce ChunkState enum and ChunkEntity struct to manage chunk lifecycle. Store chunks as shared_ptr to avoid move operations during generation. Add clear_unused_chunks to remove chunks not referenced by any player. Implement deferred chunk request queue for safe processing after generation completes. Update player chunk set during required chunk computation. Improve thread safety with mutexes on chunk and player maps. Fix m_gening flag not reset after generation and add assertions for correctness. Change need_gen to require a player UUID, removing std::optional. Add chunk_size query method for debugging.

* fix(server): ensure chunk is ready before sending or setting block

* refactor(server_world): separate gen and net thread pools

Introduce a second thread pool for network operations and a `ThreadPoolKind` enum to distinguish between gen and net pools. Rename `pool_threads()` to `gen_pool_threads()`, add `change_pool_threads()` overload that accepts the pool kind, and update the dev panel to use the gen pool. Adjust logging and initialization to handle both pools.

* feat(client_world): implement server exit acknowledgment with timeout

* feat(gameplay): implement server stop and client exit handling

Add server_stop flag to LogoutRsp protocol. Modify client_world to check for server stop or own logout to set exit flag. Add ServerWorld::stop() to broadcast stop and cleanly shut down. Refactor chunk ownership to unique_ptr. Remove name parameter from get_look_block_pos.

* fix(app): correct port validation and remove redundant TOML helper

* fix: address unused parameter warnings and missing port assignment

* refactor(gameplay): remove old pre_remove chunk, player, and world files

* refactor(server): replace chunk set with flat_hash_set and add ref count

- Replace std::unordered_set with absl::flat_hash_set for chunk position sets.
- Add ref_count field to ChunkEntity and implement update_ref_count().
- Remove clear_unused_chunks() and use ref counting for chunk lifetime.
- Add get_chunk_pos_set() accessors to ServerPlayer.

* fix(client-world): delay chunk request to wait for server central chunk generation

* refactor(gameplay): migrate chunk storage to tbb::concurrent_hash_map

* refactor(server_world): replace future-based chunk generation with queue

Remove `std::future` per-chunk tracking and poll_finished_chunks().
Instead, generation tasks push completed chunks to a concurrent queue
consumed during update. This eliminates wait_all_chunk_tasks() and
simplifies synchronization.

* refactor(client_world): use concurrent data structures for chunk management

Replace std::shared_mutex and std::unordered_set with tbb::concurrent_hash_map for chunk storage, tbb::concurrent_queue for pending uploads, and absl::flat_hash_set for player chunk positions. This improves thread safety and reduces locking overhead. Refactor receive_chunk and request_chunk to work with the new concurrent model.

* build(windows): add Nvidia and AMD GPU high performance exports

* fix(ci): update header file pattern in clang-format check

* style(clang-format): comment out AlignPPAndNotPP option

* chore: enable AlignPPAndNotPP in clang-format and pin clang version in CI

* ci(workflow): update format check to use LLVM 22

* ci(format-check): install clang-format-22

* perf(world): cull distant players from updates and rendering

Add distance2 utility function. On client, skip rendering other players if distance squared exceeds rendering distance. On server, only send player position and block change updates to players in relevant chunks.

* refactor(math_tools): convert free functions to inline in header

* refactor(dev_panel): remove seed editing and server rebuild world, move client rebuild

- Remove text editing for perlin seed and filter function
- Remove ServerWorld::rebuild_world() method and its atomic flag
- Move 'Rebuild World' button to client world tab
- Remove unused perlin_noise_input_buffer

* fix(server_world): floor block position coordinates in handle_block_change

* refactor(gameplay): pass new_chunks vector as parameter instead of member variable

* perf(client): improve chunk rendering performance with snapshots and threaded generation

* feat(tools): add PriorityThreadPool with priority scheduling

Implement PriorityThreadPool supporting task priorities and FIFO ordering for same priority.
Update ClientWorld to use the new pool with explicit priority for chunk operations.
Fix ThreadPool stop logic with atomic exchange and remove unnecessary lambda capture.

* feat(networking): add priority and sequence ordering to packet send queues

* feat(server-world): use priority thread pool for chunk generation

Introduce PriorityThreadPool to process chunk generation tasks with priorities based on their distance from the player. Closer chunks receive higher priority, improving responsiveness.

* refactor(client): extract AABB creation and add collision check on block place

Make ClientPlayer::get_aabb and ClientWorld::get_block_aabb static.
Change player size to static constexpr M_SIZE.
Use shared_mutex for concurrent reads of other players.
Check for collision with other players before placing a block.

* feat(build): add build-time version system via CMake configure_file

* ci(release): add release build workflow for windows
2026-07-02 14:51:34 +08:00

625 lines
20 KiB
C++

//
// basic_writable_pipe.hpp
// ~~~~~~~~~~~~~~~~~~~~~~~
//
// Copyright (c) 2003-2026 Christopher M. Kohlhoff (chris at kohlhoff dot com)
//
// Distributed under the Boost Software License, Version 1.0. (See accompanying
// file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
//
#ifndef ASIO_BASIC_WRITABLE_PIPE_HPP
#define ASIO_BASIC_WRITABLE_PIPE_HPP
#if defined(_MSC_VER) && (_MSC_VER >= 1200)
# pragma once
#endif // defined(_MSC_VER) && (_MSC_VER >= 1200)
#include "asio/detail/config.hpp"
#if defined(ASIO_HAS_PIPE) \
|| defined(GENERATING_DOCUMENTATION)
#include <string>
#include <utility>
#include "asio/any_io_executor.hpp"
#include "asio/async_result.hpp"
#include "asio/detail/handler_type_requirements.hpp"
#include "asio/detail/io_object_impl.hpp"
#include "asio/detail/non_const_lvalue.hpp"
#include "asio/detail/throw_error.hpp"
#include "asio/detail/type_traits.hpp"
#include "asio/error.hpp"
#include "asio/execution_context.hpp"
#if defined(ASIO_HAS_IOCP)
# include "asio/detail/win_iocp_handle_service.hpp"
#elif defined(ASIO_HAS_IO_URING_AS_DEFAULT)
# include "asio/detail/io_uring_descriptor_service.hpp"
#else
# include "asio/detail/reactive_descriptor_service.hpp"
#endif
#include "asio/detail/push_options.hpp"
namespace asio {
ASIO_INLINE_NAMESPACE_BEGIN
/// Provides pipe functionality.
/**
* The basic_writable_pipe class provides a wrapper over pipe
* functionality.
*
* @par Thread Safety
* @e Distinct @e objects: Safe.@n
* @e Shared @e objects: Unsafe.
*/
template <typename Executor = any_io_executor>
class basic_writable_pipe
{
private:
class initiate_async_write_some;
public:
/// The type of the executor associated with the object.
typedef Executor executor_type;
/// Rebinds the pipe type to another executor.
template <typename Executor1>
struct rebind_executor
{
/// The pipe type when rebound to the specified executor.
typedef basic_writable_pipe<Executor1> other;
};
/// The native representation of a pipe.
#if defined(GENERATING_DOCUMENTATION)
typedef implementation_defined native_handle_type;
#elif defined(ASIO_HAS_IOCP)
typedef detail::win_iocp_handle_service::native_handle_type
native_handle_type;
#elif defined(ASIO_HAS_IO_URING_AS_DEFAULT)
typedef detail::io_uring_descriptor_service::native_handle_type
native_handle_type;
#else
typedef detail::reactive_descriptor_service::native_handle_type
native_handle_type;
#endif
/// A basic_writable_pipe is always the lowest layer.
typedef basic_writable_pipe lowest_layer_type;
/// Construct a basic_writable_pipe without opening it.
/**
* This constructor creates a pipe without opening it.
*
* @param ex The I/O executor that the pipe will use, by default, to dispatch
* handlers for any asynchronous operations performed on the pipe.
*/
explicit basic_writable_pipe(const executor_type& ex)
: impl_(0, ex)
{
}
/// Construct a basic_writable_pipe without opening it.
/**
* This constructor creates a pipe without opening it.
*
* @param context An execution context which provides the I/O executor that
* the pipe will use, by default, to dispatch handlers for any asynchronous
* operations performed on the pipe.
*/
template <typename ExecutionContext>
explicit basic_writable_pipe(ExecutionContext& context,
constraint_t<
is_convertible<ExecutionContext&, execution_context&>::value,
defaulted_constraint
> = defaulted_constraint())
: impl_(0, 0, context)
{
}
/// Construct a basic_writable_pipe on an existing native pipe.
/**
* This constructor creates a pipe object to hold an existing native
* pipe.
*
* @param ex The I/O executor that the pipe will use, by default, to
* dispatch handlers for any asynchronous operations performed on the
* pipe.
*
* @param native_pipe A native pipe.
*
* @throws asio::system_error Thrown on failure.
*/
basic_writable_pipe(const executor_type& ex,
const native_handle_type& native_pipe)
: impl_(0, ex)
{
asio::error_code ec;
impl_.get_service().assign(impl_.get_implementation(),
native_pipe, ec);
asio::detail::throw_error(ec, "assign");
}
/// Construct a basic_writable_pipe on an existing native pipe.
/**
* This constructor creates a pipe object to hold an existing native
* pipe.
*
* @param context An execution context which provides the I/O executor that
* the pipe will use, by default, to dispatch handlers for any
* asynchronous operations performed on the pipe.
*
* @param native_pipe A native pipe.
*
* @throws asio::system_error Thrown on failure.
*/
template <typename ExecutionContext>
basic_writable_pipe(ExecutionContext& context,
const native_handle_type& native_pipe,
constraint_t<
is_convertible<ExecutionContext&, execution_context&>::value
> = 0)
: impl_(0, 0, context)
{
asio::error_code ec;
impl_.get_service().assign(impl_.get_implementation(),
native_pipe, ec);
asio::detail::throw_error(ec, "assign");
}
/// Move-construct a basic_writable_pipe from another.
/**
* This constructor moves a pipe from one object to another.
*
* @param other The other basic_writable_pipe object from which the move will
* occur.
*
* @note Following the move, the moved-from object is in the same state as if
* constructed using the @c basic_writable_pipe(const executor_type&)
* constructor.
*/
basic_writable_pipe(basic_writable_pipe&& other)
: impl_(std::move(other.impl_))
{
}
/// Move-assign a basic_writable_pipe from another.
/**
* This assignment operator moves a pipe from one object to another.
*
* @param other The other basic_writable_pipe object from which the move will
* occur.
*
* @note Following the move, the moved-from object is in the same state as if
* constructed using the @c basic_writable_pipe(const executor_type&)
* constructor.
*/
basic_writable_pipe& operator=(basic_writable_pipe&& other)
{
impl_ = std::move(other.impl_);
return *this;
}
// All pipes have access to each other's implementations.
template <typename Executor1>
friend class basic_writable_pipe;
/// Move-construct a basic_writable_pipe from a pipe of another executor type.
/**
* This constructor moves a pipe from one object to another.
*
* @param other The other basic_writable_pipe object from which the move will
* occur.
*
* @note Following the move, the moved-from object is in the same state as if
* constructed using the @c basic_writable_pipe(const executor_type&)
* constructor.
*/
template <typename Executor1>
basic_writable_pipe(basic_writable_pipe<Executor1>&& other,
constraint_t<
is_convertible<Executor1, Executor>::value,
defaulted_constraint
> = defaulted_constraint())
: impl_(std::move(other.impl_))
{
}
/// Move-assign a basic_writable_pipe from a pipe of another executor type.
/**
* This assignment operator moves a pipe from one object to another.
*
* @param other The other basic_writable_pipe object from which the move will
* occur.
*
* @note Following the move, the moved-from object is in the same state as if
* constructed using the @c basic_writable_pipe(const executor_type&)
* constructor.
*/
template <typename Executor1>
constraint_t<
is_convertible<Executor1, Executor>::value,
basic_writable_pipe&
> operator=(basic_writable_pipe<Executor1>&& other)
{
basic_writable_pipe tmp(std::move(other));
impl_ = std::move(tmp.impl_);
return *this;
}
/// Destroys the pipe.
/**
* This function destroys the pipe, cancelling any outstanding
* asynchronous wait operations associated with the pipe as if by
* calling @c cancel.
*/
~basic_writable_pipe()
{
}
/// Get the executor associated with the object.
const executor_type& get_executor() noexcept
{
return impl_.get_executor();
}
/// Get a reference to the lowest layer.
/**
* This function returns a reference to the lowest layer in a stack of
* layers. Since a basic_writable_pipe cannot contain any further layers, it
* simply returns a reference to itself.
*
* @return A reference to the lowest layer in the stack of layers. Ownership
* is not transferred to the caller.
*/
lowest_layer_type& lowest_layer()
{
return *this;
}
/// Get a const reference to the lowest layer.
/**
* This function returns a const reference to the lowest layer in a stack of
* layers. Since a basic_writable_pipe cannot contain any further layers, it
* simply returns a reference to itself.
*
* @return A const reference to the lowest layer in the stack of layers.
* Ownership is not transferred to the caller.
*/
const lowest_layer_type& lowest_layer() const
{
return *this;
}
/// Assign an existing native pipe to the pipe.
/*
* This function opens the pipe to hold an existing native pipe.
*
* @param native_pipe A native pipe.
*
* @throws asio::system_error Thrown on failure.
*/
void assign(const native_handle_type& native_pipe)
{
asio::error_code ec;
impl_.get_service().assign(impl_.get_implementation(), native_pipe, ec);
asio::detail::throw_error(ec, "assign");
}
/// Assign an existing native pipe to the pipe.
/*
* This function opens the pipe to hold an existing native pipe.
*
* @param native_pipe A native pipe.
*
* @param ec Set to indicate what error occurred, if any.
*/
ASIO_SYNC_OP_VOID assign(const native_handle_type& native_pipe,
asio::error_code& ec)
{
impl_.get_service().assign(impl_.get_implementation(), native_pipe, ec);
ASIO_SYNC_OP_VOID_RETURN(ec);
}
/// Determine whether the pipe is open.
bool is_open() const
{
return impl_.get_service().is_open(impl_.get_implementation());
}
/// Close the pipe.
/**
* This function is used to close the pipe. Any asynchronous write operations
* will be cancelled immediately, and will complete with the
* asio::error::operation_aborted error.
*
* @throws asio::system_error Thrown on failure.
*/
void close()
{
asio::error_code ec;
impl_.get_service().close(impl_.get_implementation(), ec);
asio::detail::throw_error(ec, "close");
}
/// Close the pipe.
/**
* This function is used to close the pipe. Any asynchronous write operations
* will be cancelled immediately, and will complete with the
* asio::error::operation_aborted error.
*
* @param ec Set to indicate what error occurred, if any.
*/
ASIO_SYNC_OP_VOID close(asio::error_code& ec)
{
impl_.get_service().close(impl_.get_implementation(), ec);
ASIO_SYNC_OP_VOID_RETURN(ec);
}
/// Release ownership of the underlying native pipe.
/**
* This function causes all outstanding asynchronous write operations to
* finish immediately, and the handlers for cancelled operations will be
* passed the asio::error::operation_aborted error. Ownership of the
* native pipe is then transferred to the caller.
*
* @throws asio::system_error Thrown on failure.
*
* @note This function is unsupported on Windows versions prior to Windows
* 8.1, and will fail with asio::error::operation_not_supported on
* these platforms.
*/
#if defined(ASIO_MSVC) && (ASIO_MSVC >= 1400) \
&& (!defined(_WIN32_WINNT) || _WIN32_WINNT < 0x0603)
__declspec(deprecated("This function always fails with "
"operation_not_supported when used on Windows versions "
"prior to Windows 8.1."))
#endif
native_handle_type release()
{
asio::error_code ec;
native_handle_type s = impl_.get_service().release(
impl_.get_implementation(), ec);
asio::detail::throw_error(ec, "release");
return s;
}
/// Release ownership of the underlying native pipe.
/**
* This function causes all outstanding asynchronous write operations to
* finish immediately, and the handlers for cancelled operations will be
* passed the asio::error::operation_aborted error. Ownership of the
* native pipe is then transferred to the caller.
*
* @param ec Set to indicate what error occurred, if any.
*
* @note This function is unsupported on Windows versions prior to Windows
* 8.1, and will fail with asio::error::operation_not_supported on
* these platforms.
*/
#if defined(ASIO_MSVC) && (ASIO_MSVC >= 1400) \
&& (!defined(_WIN32_WINNT) || _WIN32_WINNT < 0x0603)
__declspec(deprecated("This function always fails with "
"operation_not_supported when used on Windows versions "
"prior to Windows 8.1."))
#endif
native_handle_type release(asio::error_code& ec)
{
return impl_.get_service().release(impl_.get_implementation(), ec);
}
/// Get the native pipe representation.
/**
* This function may be used to obtain the underlying representation of the
* pipe. This is intended to allow access to native pipe
* functionality that is not otherwise provided.
*/
native_handle_type native_handle()
{
return impl_.get_service().native_handle(impl_.get_implementation());
}
/// Cancel all asynchronous operations associated with the pipe.
/**
* This function causes all outstanding asynchronous write operations to
* finish immediately, and the handlers for cancelled operations will be
* passed the asio::error::operation_aborted error.
*
* @throws asio::system_error Thrown on failure.
*/
void cancel()
{
asio::error_code ec;
impl_.get_service().cancel(impl_.get_implementation(), ec);
asio::detail::throw_error(ec, "cancel");
}
/// Cancel all asynchronous operations associated with the pipe.
/**
* This function causes all outstanding asynchronous write operations to
* finish immediately, and the handlers for cancelled operations will be
* passed the asio::error::operation_aborted error.
*
* @param ec Set to indicate what error occurred, if any.
*/
ASIO_SYNC_OP_VOID cancel(asio::error_code& ec)
{
impl_.get_service().cancel(impl_.get_implementation(), ec);
ASIO_SYNC_OP_VOID_RETURN(ec);
}
/// Write some data to the pipe.
/**
* This function is used to write data to the pipe. The function call will
* block until one or more bytes of the data has been written successfully,
* or until an error occurs.
*
* @param buffers One or more data buffers to be written to the pipe.
*
* @returns The number of bytes written.
*
* @throws asio::system_error Thrown on failure. An error code of
* asio::error::eof indicates that the connection was closed by the
* peer.
*
* @note The write_some operation may not transmit all of the data to the
* peer. Consider using the @ref write function if you need to ensure that
* all data is written before the blocking operation completes.
*
* @par Example
* To write a single data buffer use the @ref buffer function as follows:
* @code
* pipe.write_some(asio::buffer(data, size));
* @endcode
* See the @ref buffer documentation for information on writing multiple
* buffers in one go, and how to use it with arrays, boost::array or
* std::vector.
*/
template <typename ConstBufferSequence>
std::size_t write_some(const ConstBufferSequence& buffers)
{
asio::error_code ec;
std::size_t s = impl_.get_service().write_some(
impl_.get_implementation(), buffers, ec);
asio::detail::throw_error(ec, "write_some");
return s;
}
/// Write some data to the pipe.
/**
* This function is used to write data to the pipe. The function call will
* block until one or more bytes of the data has been written successfully,
* or until an error occurs.
*
* @param buffers One or more data buffers to be written to the pipe.
*
* @param ec Set to indicate what error occurred, if any.
*
* @returns The number of bytes written. Returns 0 if an error occurred.
*
* @note The write_some operation may not transmit all of the data to the
* peer. Consider using the @ref write function if you need to ensure that
* all data is written before the blocking operation completes.
*/
template <typename ConstBufferSequence>
std::size_t write_some(const ConstBufferSequence& buffers,
asio::error_code& ec)
{
return impl_.get_service().write_some(
impl_.get_implementation(), buffers, ec);
}
/// Start an asynchronous write.
/**
* This function is used to asynchronously write data to the pipe. It is an
* initiating function for an @ref asynchronous_operation, and always returns
* immediately.
*
* @param buffers One or more data buffers to be written to the pipe.
* Although the buffers object may be copied as necessary, ownership of the
* underlying memory blocks is retained by the caller, which must guarantee
* that they remain valid until the completion handler is called.
*
* @param token The @ref completion_token that will be used to produce a
* completion handler, which will be called when the write completes.
* Potential completion tokens include @ref use_future, @ref use_awaitable,
* @ref yield_context, or a function object with the correct completion
* signature. The function signature of the completion handler must be:
* @code void handler(
* const asio::error_code& error, // Result of operation.
* std::size_t bytes_transferred // Number of bytes written.
* ); @endcode
* Regardless of whether the asynchronous operation completes immediately or
* not, the completion handler will not be invoked from within this function.
* On immediate completion, invocation of the handler will be performed in a
* manner equivalent to using asio::async_immediate().
*
* @par Completion Signature
* @code void(asio::error_code, std::size_t) @endcode
*
* @note The write operation may not transmit all of the data to the peer.
* Consider using the @ref async_write function if you need to ensure that all
* data is written before the asynchronous operation completes.
*
* @par Example
* To write a single data buffer use the @ref buffer function as follows:
* @code
* pipe.async_write_some(asio::buffer(data, size), handler);
* @endcode
* See the @ref buffer documentation for information on writing multiple
* buffers in one go, and how to use it with arrays, boost::array or
* std::vector.
*/
template <typename ConstBufferSequence,
ASIO_COMPLETION_TOKEN_FOR(void (asio::error_code,
std::size_t)) WriteToken = default_completion_token_t<executor_type>>
auto async_write_some(const ConstBufferSequence& buffers,
WriteToken&& token = default_completion_token_t<executor_type>())
-> decltype(
async_initiate<WriteToken,
void (asio::error_code, std::size_t)>(
declval<initiate_async_write_some>(), token, buffers))
{
return async_initiate<WriteToken,
void (asio::error_code, std::size_t)>(
initiate_async_write_some(this), token, buffers);
}
private:
// Disallow copying and assignment.
basic_writable_pipe(const basic_writable_pipe&) = delete;
basic_writable_pipe& operator=(const basic_writable_pipe&) = delete;
class initiate_async_write_some
{
public:
typedef Executor executor_type;
explicit initiate_async_write_some(basic_writable_pipe* self)
: self_(self)
{
}
const executor_type& get_executor() const noexcept
{
return self_->get_executor();
}
template <typename WriteHandler, typename ConstBufferSequence>
void operator()(WriteHandler&& handler,
const ConstBufferSequence& buffers) const
{
// If you get an error on the following line it means that your handler
// does not meet the documented type requirements for a WriteHandler.
ASIO_WRITE_HANDLER_CHECK(WriteHandler, handler) type_check;
detail::non_const_lvalue<WriteHandler> handler2(handler);
self_->impl_.get_service().async_write_some(
self_->impl_.get_implementation(), buffers,
handler2.value, self_->impl_.get_executor());
}
private:
basic_writable_pipe* self_;
};
#if defined(ASIO_HAS_IOCP)
detail::io_object_impl<detail::win_iocp_handle_service, Executor> impl_;
#elif defined(ASIO_HAS_IO_URING_AS_DEFAULT)
detail::io_object_impl<detail::io_uring_descriptor_service, Executor> impl_;
#else
detail::io_object_impl<detail::reactive_descriptor_service, Executor> impl_;
#endif
};
ASIO_INLINE_NAMESPACE_END
} // namespace asio
#include "asio/detail/pop_options.hpp"
#endif // defined(ASIO_HAS_PIPE)
// || defined(GENERATING_DOCUMENTATION)
#endif // ASIO_BASIC_WRITABLE_PIPE_HPP