improve lookup time for matches_any_publishers().

[fujitatomoya]

Description

Fixes #3067

Is this user-facing behavior change?

No, just a minor performance improvement.

Did you use Generative AI?

Yes, Copilot Claude Sonnet 4.5

Additional Information

[fujitatomoya]

Pulls: #3068
Gist: https://gist.githubusercontent.com/fujitatomoya/c19e393b6ca155a3d6bf9012c60a8c2e/raw/76c8af93fb984113dd722b48638fdd6320a1b0e1/ros2.repos
BUILD args: --packages-above-and-dependencies rclcpp
TEST args: --packages-above rclcpp
ROS Distro: rolling
Job: ci_launcher
ci_launcher ran: https://ci.ros2.org/job/ci_launcher/18229

• Linux
• Linux-aarch64
• Linux-rhel
• Windows

[mini-1235]

@fujitatomoya I think this causes a regression in Nav2 (when turning intraprocess on)

backtrace:

Thread 14 "amcl" received signal SIGSEGV, Segmentation fault.                                                                                                       
[Switching to Thread 0x7fffcb7fe6c0 (LWP 983342)]
#0  0x00007ffff7dfd8b3 in std::_Hashtable<rmw_gid_s, std::pair<rmw_gid_s const, rclcpp::experimental::IntraProcessManager::PublisherInfo>, std::allocator<std::pair<rmw_gid_s const, rclcpp::experimental::IntraProcessManager::PublisherInfo> >, std::__detail::_Select1st, rclcpp::experimental::IntraProcessManager::rmw_gid_equal, rclcpp::experimental::IntraProcessManager::rmw_gid_hash, std::__detail::_Mod_range_hashing, std::__detail::_Default_ranged_hash, std::__detail::_Prime_rehash_policy, std::__detail::_Hashtable_traits<false, false, true> >::_M_find_before_node (this=0x55555565e448, __bkt=346849886, __k=..., 
    __code=<optimized out>) at /usr/include/c++/13/bits/hashtable.h:194
#1  0x00007ffff7e00b6c in std::_Hashtable<rmw_gid_s, std::pair<rmw_gid_s const, rclcpp::experimental::IntraProcessManager::PublisherInfo>, std::allocator<std::pair<rmw_gid_s const, rclcpp::experimental::IntraProcessManager::PublisherInfo> >, std::__detail::_Select1st, rclcpp::experimental::IntraProcessManager::rmw_gid_equal, rclcpp::experimental::IntraProcessManager::rmw_gid_hash, std::__detail::_Mod_range_hashing, std::__detail::_Default_ranged_hash, std::__detail::_Prime_rehash_policy, std::__detail::_Hashtable_traits<false, false, true> >::_M_find_node (__c=<optimized out>, __key=..., __bkt=<optimized out>, this=<optimized out>)
    at /usr/include/c++/13/bits/hashtable.h:815
#2  std::_Hashtable<rmw_gid_s, std::pair<rmw_gid_s const, rclcpp::experimental::IntraProcessManager::PublisherInfo>, std::allocator<std::pair<rmw_gid_s const, rclcpp::experimental::IntraProcessManager::PublisherInfo> >, std::__detail::_Select1st, rclcpp::experimental::IntraProcessManager::rmw_gid_equal, rclcpp::experimental::IntraProcessManager::rmw_gid_hash, std::__detail::_Mod_range_hashing, std--Type <RET> for more, q to quit, c to continue without paging--
::__detail::_Default_ranged_hash, std::__detail::_Prime_rehash_policy, std::__detail::_Hashtable_traits<false, false, true> >::find (__k=..., this=<optimized out>)
    at /usr/include/c++/13/bits/hashtable.h:1702
#3  std::_Hashtable<rmw_gid_s, std::pair<rmw_gid_s const, rclcpp::experimental::IntraProcessManager::PublisherInfo>, std::allocator<std::pair<rmw_gid_s const, rclcpp::experimental::IntraProcessManager::PublisherInfo> >, std::__detail::_Select1st, rclcpp::experimental::IntraProcessManager::rmw_gid_equal, rclcpp::experimental::IntraProcessManager::rmw_gid_hash, std::__detail::_Mod_range_hashing, std::__detail::_Default_ranged_hash, std::__detail::_Prime_rehash_policy, std::__detail::_Hashtable_traits<false, false, true> >::find (this=<optimized out>, __k=...) at /usr/include/c++/13/bits/hashtable.h:1687
#4  0x00007ffff7dffb50 in std::unordered_map<rmw_gid_s, rclcpp::experimental::IntraProcessManager::PublisherInfo, rclcpp::experimental::IntraProcessManager::rmw_gid_hash, rclcpp::experimental::IntraProcessManager::rmw_gid_equal, std::allocator<std::pair<rmw_gid_s const, rclcpp::experimental::IntraProcessManager::PublisherInfo> > >::find (__x=..., this=0x55555565e448) at /usr/include/c++/13/bits/unordered_map.h:88
#5  rclcpp::experimental::IntraProcessManager::matches_any_publishers (this=0x55555565e360, id=<optimized out>)
    at /root/nav2_ws/src/rclcpp/rclcpp/src/rclcpp/intra_process_manager.cpp:132
#6  0x00007ffff7ec8511 in rclcpp::SubscriptionBase::matches_any_intra_process_publishers (this=this@entry=0x7fffbc1e12f0, sender_gid=<optimized out>)
    at /root/nav2_ws/src/rclcpp/rclcpp/src/rclcpp/subscription_base.cpp:403
#7  0x00007ffff7ec86a3 in rclcpp::SubscriptionBase::take_type_erased (this=0x7fffbc1e12f0, message_out=0x7fffbc204430, message_info_out=...)
    at /root/nav2_ws/src/rclcpp/rclcpp/src/rclcpp/subscription_base.cpp:256
#8  0x00007ffff7dba2dd in operator() (__closure=<synthetic pointer>) at /usr/include/c++/13/bits/shared_ptr_base.h:1665
#9  take_and_do_error_handling<rclcpp::Executor::execute_subscription(rclcpp::SubscriptionBase::SharedPtr)::<lambda()>, rclcpp::Executor::execute_subscription(rclcpp::SubscriptionBase::SharedPtr)::<lambda()> > (action_description=0x7ffff7f00226 "taking a message from topic", handle_action=..., take_action=..., 
    topic_or_service_name=0x7fffbc1e21d0 "/map") at /root/nav2_ws/src/rclcpp/rclcpp/src/rclcpp/executor.cpp:522
#10 rclcpp::Executor::execute_subscription (subscription=std::shared_ptr<rclcpp::SubscriptionBase> (use count 3, weak count 4) = {...})
    at /root/nav2_ws/src/rclcpp/rclcpp/src/rclcpp/executor.cpp:602
#11 0x00007ffff7dbb777 in rclcpp::Executor::execute_any_executable (this=this@entry=0x555555719fa0, any_exec=...)
    at /root/nav2_ws/src/rclcpp/rclcpp/src/rclcpp/executor.cpp:494
#12 0x00007ffff7de1802 in rclcpp::executors::SingleThreadedExecutor::spin (this=0x555555719fa0)
    at /root/nav2_ws/src/rclcpp/rclcpp/src/rclcpp/executors/single_threaded_executor.cpp:42
#13 0x00007ffff7a6fdb4 in ?? () from /lib/x86_64-linux-gnu/libstdc++.so.6
#14 0x00007ffff77dfaa4 in start_thread (arg=<optimized out>) at ./nptl/pthread_create.c:447
#15 0x00007ffff786cc6c in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:78

[fujitatomoya]

@mini-1235 can you give it a shot with #3073 to see if your problem can be fixed? all rclcpp CI and tests are passing, there could be other possibility...