userver-framework / userver Goto Github PK

Production-ready C++ Asynchronous Framework with rich functionality

License: Apache License 2.0

CMake 1.59% Makefile 0.05% C++ 90.59% C 0.67% Python 5.23% Jinja 0.16% SWIG 0.07% Shell 0.19% CSS 0.96% HTML 0.12% JavaScript 0.29% Dockerfile 0.08%

async asynchronous coroutines database-access cpp17 cplusplus cplusplus-17 web server microservices

userver's Introduction

userver

Service Templates	Develop / Green Trunk	v1.0
Core:		[➚]
PostgreSQL:		[➚]
gRPC+PostgreSQL:		[➚]

userver is an open source asynchronous framework with a rich set of abstractions for fast and comfortable creation of C++ microservices, services and utilities.

The framework solves the problem of efficient I/O interactions transparently for the developers. Operations that would typically suspend the thread of execution do not suspend it. Instead of that, the thread processes other requests and tasks and returns to the handling of the operation only when it is guaranteed to execute immediately:

std::size_t Ins(storages::postgres::Transaction& tr, std::string_view key) {
  // Asynchronous execution of the SQL query in transaction. Current thread
  // handles other requests while the response from the DB is being received:
  auto res = tr.Execute("INSERT INTO keys VALUES ($1)", key);
  return res.RowsAffected();
}

As a result, with the framework you get straightforward source code, avoid CPU-consuming context switches from OS, efficiently utilize the CPU with a small amount of execution threads.

You can learn more about history and key features of userver from our articles on Medium (English) or Habr (Russian).

Other Features

Efficient asynchronous drivers for databases (MongoDB, PostgreSQL, Redis, ClickHouse, MySQL/MariaDB (experimental) ...) and data transfer protocols (HTTP, gRPC, AMQP 0-9-1 (experimental), TCP, TLS, WebSocket ...), tasks construction and cancellation.
Rich set of high-level components for caches, tasks, distributed locking, logging, tracing, statistics, metrics, JSON/YAML/BSON.
Functionality to change the service configuration on-the-fly.
On-the-fly configurable drivers, options of the deadline propagation, timeouts, congestion-control.
Comprehensive set of asynchronous low-level synchronization primitives and OS abstractions.

See the docs for more info.

userver's People

Contributors

Stargazers

Watchers

Forkers

vitek alexiprof apolukhin zzzabiyaka crackercat externalrepositories firsov179 dr-evilly ksanvat segoon leha-bot botogoske juggerd volerog lesf0 iamfork illidansr baajarmeh botanegg dengaleev evgeniibalabanov itrofimow kargatpwnz tuminsky anton3 nul1 korenandr izvolov svirex dimasidorenko jtstrs monad-one kenny-ngo k4sho profabvera strogo mrcodechef pavel-pimenov icodein zhengzhenxing wahello lyrl xiaohunqupo mu-l atifkarim codephilanthropist lovening cyberflamego yespon tonywork liduanwei swqhd jnhu76 rajabhinav1 ykyou load-net hundong2 ywjb duzhanyuan goga1992 sadjumb pvsur hei114 zzu-andrew hj3938 69567272 feiyunwill steparui ndevel yubobo linecode nybik nolinkspl szvyagin-gj sanyanya jorgenpo gerhobbelt syntezzz goyambokkaboy und3v3l0p3d korshunovdv sugarhl littleboyjohnny zmij blockspacer n3zhov sergej-samsonow flyarong fadeevsergey msavanan den818 x01af tregffa davebian ybogo druidfund maximsmolskiy aidankenney yoori jihadist

userver's Issues

Update moodycamel to control MAX_SEMA_SPINS

You see, moodycamel::BlockingConcurrentQueue spins here before going to kernel semaphore
and although it's preferable for maximum throughput in benchmarks,
it wastes CPU cycles in some real-life scenarios.

Since this commit there is a way to control this spinning behavior, which was very noticeable in some perfs, remember @apolukhin?

Lack of comparison benchmarks is concerning

First of all, thank you guys for this impressive piece of software.

As stated in the title, i'm a bit concerned about the lack of comparison benchmarks: guess its safe to assume that if someone wants a c++ web-framework they mostly strive for c++ performance, and only after that for simplicity of Python, coroutine model of Go or other things, however there's no evidence of "speed of c++" aside from using c++, and that kinda questions the point of the framework (imo)

Dynamic config referred as "Taxi config" in logs

Unhandled exception in components::Run: Cannot load fallback taxi config: Error opening '/etc/hello_service/dynamic_config_fallback.json'

should be "Cannot load fallback dynamic config"

mariadb support

Is there a plan to add mariadb support in future..

Compilation error when using clickhouse storage

In file included from /usr/include/c++/12.1.0/x86_64-pc-linux-gnu/bits/os_defines.h:39,
                 from /usr/include/c++/12.1.0/x86_64-pc-linux-gnu/bits/c++config.h:655,
                 from /usr/include/c++/12.1.0/type_traits:38,
                 from /usr/include/c++/12.1.0/optional:37,
                 from /root/code/userver/clickhouse/include/userver/storages/clickhouse/io/columns/nullable_column.hpp:7,
                 from /root/code/userver/clickhouse/src/storages/clickhouse/io/columns/nullable_column.cpp:1:
/usr/include/features.h:412:4: warning: #warning _FORTIFY_SOURCE requires compiling with optimization (-O) [-Wcpp]
  412 | #  warning _FORTIFY_SOURCE requires compiling with optimization (-O)
      |    ^~~~~~~
/root/code/userver/clickhouse/include/userver/storages/clickhouse/io/columns/nullable_column.hpp: In member function ‘userver::storages::clickhouse::io::columns::NullableColumn<T>::NullableDataHolder userver::storages::clickhouse::io::columns::NullableColumn<T>::NullableDataHolder::operator++(int)’:
/root/code/userver/clickhouse/include/userver/storages/clickhouse/io/columns/nullable_column.hpp:94:25: error: ‘exchange’ is not a member of ‘std’
   94 |   old.has_value_ = std::exchange(has_value_, false);
      |                         ^~~~~~~~
make[3]: *** [userver/clickhouse/CMakeFiles/userver-clickhouse.dir/build.make:328: userver/clickhouse/CMakeFiles/userver-clickhouse.dir/src/storages/clickhouse/io/columns/nullable_column.cpp.o] Error 1
make[2]: *** [CMakeFiles/Makefile2:1663: userver/clickhouse/CMakeFiles/userver-clickhouse.dir/all] Error 2
make[1]: *** [CMakeFiles/Makefile2:2325: samples/clickhouse_service/CMakeFiles/userver-samples-clickhouse_service.dir/rule] Error 2
make: *** [Makefile:868: userver-samples-clickhouse_service] Error 2

Seems like nullable_column.hpp does not include utility header

Using TaxiConfigClient?

Полагаю, такси все еще не отпускает)
https://github.com/userver-framework/userver/blob/develop/core/include/userver/dynamic_config/client/component.hpp 56

Userver

Support WebSocket

В ряде случаев WebSocket имеет преимущество перед HTTP. Было бы здорово увидеть его поддержку в userver.

Cannot find utest with conan

Some issues with conan and utest
Here is how to reproduce this issue

Build userver with conan, with option USERVER_FEATURE_UTEST=ON
Link executable with userver::userver or even CONAN_LIB::userver_userver-utest
File userver/utest/utest.hpp doesn't exist.
Apparently, there is only using_namespace_userver.hpp file in userver/utest

CMake Functions such as add_google_tests and userver_testsuite_add also cannot be seen.

Link to uservice-dynconf leads to GitHub 404 page

Hello and congratulations on open sourcing this amazing work!

There's a section on the page about dynamic configs that mentions userver-dynconf:

Note that there is a ready to use opensource uservice-dynconf dynamic configs service. Use it for your projects or just disable dynamic config updates and keep developing without a supplementary service.

Following the link to uservice-dynconf yields a GitHub 404 page, which is probably not expected.

Tasks do not cancel during shutdown

In this code coroutine won't be canceled if TCP connection is a live during shutdown process. Coroutine is waiting for wake-up, ev thread is waiting for socket events. The only way to close coroutine is to send something through connection or close it on remote site.

Expected behavior is to wake up all tasks and throw canceled exception if manager is shuting down.

static engine::Task task;

class StuckedTCPServer final : public userver::components::LoggableComponentBase {
 public:
  static constexpr std::string_view kName = "stucked-tcp-server";

  StuckedTCPServer(const userver::components::ComponentConfig& component_config,
                  const userver::components::ComponentContext& component_context)
      : userver::components::LoggableComponentBase(component_config, component_context) {
    userver::server::net::ListenerConfig listener;
    listener.port = 8080;
    task = utils::Async(
        "listener",
        [](auto socket) {
          while (!engine::current_task::ShouldCancel()) {
            auto s = socket.Accept({});
            std::array<char, 32> buf;
            s.RecvAll(buf.data(), buf.size(), {});
          }
        },
        userver::server::net::CreateSocket(listener));
  }

  void OnAllComponentsAreStopping() override {
    // close active connections
  }
};

int main(int argc, char* argv[]) {
  auto component_list = components::ComponentList()
                            .Append<os_signals::ProcessorComponent>()
                            .Append<components::Logging>()
                            .Append<components::Tracer>()
                            .Append<components::ManagerControllerComponent>()
                            .Append<components::StatisticsStorage>()
                            .Append<components::DynamicConfig>()
                            .Append<components::DynamicConfigFallbacks>()
                            .Append<StuckedTCPServer>();

  return userver::utils::DaemonMain(argc, argv, component_list);
}

`engine::Yield` doesn't actually yield

Due to moodycamel::ConcurrentQueue implementation details engine::Yield doesn't actually yield running thread back to the thread-pool:
since moodycamel consists of a set of thread-local FIFO queues and
every consumer checks for its paired producer queue first, when a coroutine-thread
calls engine::Yield task continuation is scheduled "for himself",
thus when it tries to dequeue a task it will get the exact same continuation it just yielded
(it's not always true and moodycamel does rotate consumer<->producer pairing from time to time).

Here is an example of what i'm talking about, and sadly this situation has some real-life analogues:
Consider PostgreCache: it uses engine::Yield when copying a container or iterating fetched values, and that could take plenty of iterations, basically "blocking" current thread, even though it yields.

Imagine you have 2 cores, 2 worker-threads and 150RPS, where each request takes 10ms of CPU times:
2 threads should deal with it no problem, leaving 250ms of CPU time to spare,
however as soon as one thread gets into "yield-loop" second thread gets overwhelmed instantly

What's even worse in aforementioned example,
PostgreCache does itself blocks a worker-thread when it deallocates cached data, but that's another problem i guess

Реализация конструктора запросов QueryBuilder

Часто приходится реализовывать запросы к БД, которые в зависимости от переданных параметров должны добавлять в запрос к БД дополнительные условия - фильтры.

Например, если передали строковый параметр, то нужно поискать в строковых полях, если передали коллекцию идентификаторов, то поискать с вхождением какого-то поля в эту коллекцию( массив).

Для этих целей удобно использовать QueryBuilder по типу такого:

https://docs.rs/sqlx/latest/sqlx/struct.QueryBuilder.html

Фактически он просто склеивает запрос, позволяя вставлять аргументы с эскейпингом для защиты от иньекций.

Добавляет escaped значение фильтра:
builder.bind(filter)
Вставляет строку в запрос как есть:
buidler.bind(" AND value IN ")

ну и другие удобные методы.

Есть ли что-то подобное в userver? Я не нашел. Если нет, было бы классно реализовать

Single-waiter mutex

There are some situations when a resource is accessed concurrently, but concurrency factor is limited by 2, - implications of socket r/w duality come to mind, - and engine::Mutex seems like a go to choice (and it is), however it operates over full-featured WaitList, when in this specific case WaitListLight would do, since there's at most 1 waiter at any given moment.

Improve ArchLinux support

Userver have a lot warnings with suggested disabling USERVER_CHECK_PACKAGE_VERSIONS
Also have a lot of problems on linkage time like #19 but different libs

I think its a bad way to mute this warnings.
We can improve situation by addition pacman version detector to DetectVersion.cmake

Reduce allocation count in PostgreSQL

userver/postgresql/src/storages/postgres/options.cpp

Line 23 in e37c78f

const std::unordered_map<TransactionOptions, std::string, HashOptions>

Why this global variables are not constexpr ? Seems like std::string_view may be used here without any problems.

What about unordred_map, do it constexpr may be tricky, but there are not many values and constexpr flat map can be easialy created with standard algorithms

Ошибка в определении оператора `!=`

Это ошибка вызвана, предположительно, копипастом. Определение оператора неравенства

userver/core/include/userver/server/auth/user_scopes.hpp

Lines 44 to 46 in cb8efc3

 inline bool operator!=(std::string_view lhs, const UserScope& rhs) { 

 return lhs == rhs.GetValue(); 

 }

Такое же, как и определение оператора равенства

userver/core/include/userver/server/auth/user_scopes.hpp

Lines 36 to 38 in cb8efc3

 inline bool operator==(std::string_view lhs, const UserScope& rhs) { 

 return lhs == rhs.GetValue(); 

 }

Очевидно, там подразумевалось != вместо ==

2G tmpfs isn't not enough for build whole project

Hello,

We have "build nodes" that are "docker nodes" with configurations tmpfs - 2G (for /tmp), ramfs - 8G (for build folder). For this project, our configuration is:

docker run -it \
--rm \
--name userver_build_release \
--mount type=tmpfs,destination=/tmp,tmpfs-size=2147483648 \
--volume /mnt/ramfs/userver:/userver/build_release \
ubuntu:22.04

And tried build:

apt update && apt install -y git

cd userver/
git init
git remote add origin https://github.com/userver-framework/userver.git
git pull
git checkout develop -f

# Environments for tzdata

export DEBIAN_FRONTEND=noninteractive 
export TZ=Etc/UTC

# https://userver.tech/d1/d03/md_en_userver_tutorial_build.html#installationinstructions

apt install -y $(cat scripts/docs/en/deps/ubuntu-22.04.md | tr '\n' ' ')
git config --global --add safe.directory $(pwd)/third_party/clickhouse-cpp
mkdir build_release
cd build_release
cmake -DCMAKE_BUILD_TYPE=Release ..

# Check available size in /tmp

df -h /tmp

# Build project

make -j12

We got errors :

lto-wrapper: fatal error: write: No space left on device
compilation terminated.
lto1: fatal error: error writing to /tmp/ccwzjvhU.s: No space left on device
compilation terminated.
/usr/bin/ld: error: lto-wrapper failed
lto-wrapper: fatal error: /usr/bin/c++ returned 1 exit status
compilation terminated.
collect2: error: ld returned 1 exit status
make[2]: *** [samples/clickhouse_service/CMakeFiles/userver-samples-clickhouse_service.dir/build.make:129: samples/clickhouse_service/userver-samples-clickhouse_service] Error 1
make[1]: *** [CMakeFiles/Makefile2:3842: samples/clickhouse_service/CMakeFiles/userver-samples-clickhouse_service.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....
[ 66%] Building CXX object userver/core/CMakeFiles/userver-utest.dir/testing/src/utest/dns_server_mock.cpp.o
/usr/bin/ld: error: lto-wrapper failed
collect2: error: ld returned 1 exit status
make[2]: *** [samples/http_caching/CMakeFiles/userver-samples-http_caching.dir/build.make:122: samples/http_caching/userver-samples-http_caching] Error 1
make[1]: *** [CMakeFiles/Makefile2:3397: samples/http_caching/CMakeFiles/userver-samples-http_caching.dir/all] Error 2

Of course we can disable tests, examples and USERVER_FEATURE_* but it looks strange

Getting following error

Ubuntu 20.04
while installing following error after given make -j$(nproc)

[ 52%] Built target netcat
[ 52%] Building CXX object tools/congestion_control_emulator/CMakeFiles/congestion_control_emulator.dir/congestion_control_emulator.cpp.o
[ 52%] Linking CXX executable congestion_control_emulator
[ 52%] Linking CXX executable dns-resolver
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans22.ltrans.o: in function `bool boost::regex_search<char const*, char, boost::regex_traits<char, boost::cpp_regex_traits<char> > >(char const*, char const*, boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, boost::regex_constants::_match_flags) [clone .constprop.0]':
/usr/include/boost/regex/v4/perl_matcher.hpp:386: undefined reference to `boost::re_detail_106700::perl_matcher<char const*, std::allocator<boost::sub_match<char const*> >, boost::regex_traits<char, boost::cpp_regex_traits<char> > >::construct_init(boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, boost::regex_constants::_match_flags)'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans22.ltrans.o: in function `bool boost::regex_search<char const*, char, boost::regex_traits<char, boost::cpp_regex_traits<char> > >(char const*, char const*, boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, boost::regex_constants::_match_flags) [clone .constprop.0]':
/usr/include/boost/regex/v4/perl_matcher_non_recursive.hpp:107: undefined reference to `boost::re_detail_106700::get_mem_block()'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans22.ltrans.o: in function `bool boost::regex_search<char const*, char, boost::regex_traits<char, boost::cpp_regex_traits<char> > >(char const*, char const*, boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, boost::regex_constants::_match_flags) [clone .constprop.0]':
/usr/include/boost/regex/v4/perl_matcher_common.hpp:305: undefined reference to `boost::re_detail_106700::verify_options(unsigned int, boost::regex_constants::_match_flags)'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans22.ltrans.o: in function `bool boost::regex_search<char const*, char, boost::regex_traits<char, boost::cpp_regex_traits<char> > >(char const*, char const*, boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, boost::regex_constants::_match_flags) [clone .constprop.0]':
/usr/include/boost/regex/v4/perl_matcher_non_recursive.hpp:115: undefined reference to `boost::re_detail_106700::put_mem_block(void*)'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans22.ltrans.o: in function `bool boost::regex_search<char const*, char, boost::regex_traits<char, boost::cpp_regex_traits<char> > >(char const*, char const*, boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, boost::regex_constants::_match_flags) [clone .constprop.0] [clone .cold]':
/usr/include/boost/regex/v4/perl_matcher_non_recursive.hpp:115: undefined reference to `boost::re_detail_106700::put_mem_block(void*)'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans24.ltrans.o: in function `char const* boost::re_detail_106700::re_is_set_member<char const*, char, boost::regex_traits<char, boost::cpp_regex_traits<char> >, unsigned int>(char const*, char const*, boost::re_detail_106700::re_set_long<unsigned int> const*, boost::re_detail_106700::regex_data<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, bool)':
/usr/include/boost/regex/v4/cpp_regex_traits.hpp:966: undefined reference to `boost::re_detail_106700::cpp_regex_traits_implementation<char>::transform_primary[abi:cxx11](char const*, char const*) const'
/usr/bin/ld: /usr/include/boost/regex/v4/cpp_regex_traits.hpp:962: undefined reference to `boost::re_detail_106700::cpp_regex_traits_implementation<char>::transform[abi:cxx11](char const*, char const*) const'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans24.ltrans.o: in function `void boost::re_detail_106700::raise_error<boost::regex_traits_wrapper<boost::regex_traits<char, boost::cpp_regex_traits<char> > > >(boost::regex_traits_wrapper<boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, boost::regex_constants::error_type)':
/usr/include/boost/regex/v4/cpp_regex_traits.hpp:447: undefined reference to `boost::re_detail_106700::get_default_error_string(boost::regex_constants::error_type)'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans24.ltrans.o: in function `void boost::re_detail_106700::raise_error<boost::regex_traits_wrapper<boost::regex_traits<char, boost::cpp_regex_traits<char> > > >(boost::regex_traits_wrapper<boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, boost::regex_constants::error_type)':
/usr/include/boost/regex/pattern_except.hpp:75: undefined reference to `boost::re_detail_106700::raise_runtime_error(std::runtime_error const&)'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans24.ltrans.o: in function `boost::re_detail_106700::perl_matcher<char const*, std::allocator<boost::sub_match<char const*> >, boost::regex_traits<char, boost::cpp_regex_traits<char> > >::extend_stack()':
/usr/include/boost/regex/v4/perl_matcher_non_recursive.hpp:233: undefined reference to `boost::re_detail_106700::get_mem_block()'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans24.ltrans.o: in function `boost::re_detail_106700::perl_matcher<char const*, std::allocator<boost::sub_match<char const*> >, boost::regex_traits<char, boost::cpp_regex_traits<char> > >::match_then()':
/usr/include/boost/regex/v4/perl_matcher_non_recursive.hpp:233: undefined reference to `boost::re_detail_106700::get_mem_block()'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans24.ltrans.o: in function `boost::re_detail_106700::perl_matcher<char const*, std::allocator<boost::sub_match<char const*> >, boost::regex_traits<char, boost::cpp_regex_traits<char> > >::match_commit()':
/usr/include/boost/regex/v4/perl_matcher_non_recursive.hpp:233: undefined reference to `boost::re_detail_106700::get_mem_block()'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans24.ltrans.o: in function `boost::re_detail_106700::perl_matcher<char const*, std::allocator<boost::sub_match<char const*> >, boost::regex_traits<char, boost::cpp_regex_traits<char> > >::match_toggle_case()':
/usr/include/boost/regex/v4/perl_matcher_non_recursive.hpp:233: undefined reference to `boost::re_detail_106700::get_mem_block()'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans25.ltrans.o: in function `boost::re_detail_106700::perl_matcher<char const*, std::allocator<boost::sub_match<char const*> >, boost::regex_traits<char, boost::cpp_regex_traits<char> > >::unwind_extra_block(bool)':
/usr/include/boost/regex/v4/perl_matcher_non_recursive.hpp:1368: undefined reference to `boost::re_detail_106700::put_mem_block(void*)'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans25.ltrans.o: in function `userver::utils::regex_match(std::basic_string_view<char, std::char_traits<char> >, userver::utils::regex const&)':
/usr/include/boost/regex/v4/perl_matcher.hpp:386: undefined reference to `boost::re_detail_106700::perl_matcher<char const*, std::allocator<boost::sub_match<char const*> >, boost::regex_traits<char, boost::cpp_regex_traits<char> > >::construct_init(boost::basic_regex<char, boost::regex_traits<char, boost::cpp_regex_traits<char> > > const&, boost::regex_constants::_match_flags)'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans25.ltrans.o: in function `userver::utils::regex_match(std::basic_string_view<char, std::char_traits<char> >, userver::utils::regex const&)':
/usr/include/boost/regex/v4/perl_matcher_non_recursive.hpp:107: undefined reference to `boost::re_detail_106700::get_mem_block()'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans25.ltrans.o: in function `userver::utils::regex_match(std::basic_string_view<char, std::char_traits<char> >, userver::utils::regex const&)':
/usr/include/boost/regex/v4/perl_matcher_common.hpp:219: undefined reference to `boost::re_detail_106700::verify_options(unsigned int, boost::regex_constants::_match_flags)'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans25.ltrans.o: in function `userver::utils::regex_match(std::basic_string_view<char, std::char_traits<char> >, userver::utils::regex const&)':
/usr/include/boost/regex/v4/perl_matcher_non_recursive.hpp:115: undefined reference to `boost::re_detail_106700::put_mem_block(void*)'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans25.ltrans.o: in function `userver::utils::regex_match(std::basic_string_view<char, std::char_traits<char> >, userver::utils::regex const&) [clone .cold]':
/usr/include/boost/regex/v4/perl_matcher_non_recursive.hpp:115: undefined reference to `boost::re_detail_106700::put_mem_block(void*)'
/usr/bin/ld: /tmp/userver-universal_unittest.fzUonP.ltrans26.ltrans.o: in function `userver::fs::blocking::(anonymous namespace)::RemoveDirectory(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)':
/usr/include/boost/system/error_code.hpp:461: undefined reference to `boost::system::system_category()'
collect2: error: ld returned 1 exit status
make[2]: *** [userver/universal/CMakeFiles/userver-universal_unittest.dir/build.make:1441: userver/universal/userver-universal_unittest] Error 1
make[1]: *** [CMakeFiles/Makefile2:3212: userver/universal/CMakeFiles/userver-universal_unittest.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....
[ 52%] Built target httpclient_perf
[ 52%] Built target congestion_control_emulator
[ 52%] Built target dns-resolver
make: *** [Makefile:146: all] Error 2

Requiring Jinja2 for building

Does it really require jinja2 at the CMake configuration step? I'd like to try building this without having to install any Python dependencies if possible.

Improve the documentation for `macOS`

Please improve the documentation about the pre-requirements packages for macOS here

Now, the configure process is not easy for macOS. For instance, I faced some problems when realpath wasn't installed on my PC. So, I had to install coreutils.

Поддержка коллекций в запросах к БД

На данный момент есть возможность передавать параметр типа std::vector в Execute и он эскейпится в значение массива.

Хорошо бы добавить возможность такого преобразования для google::protobuf::RepeatedField и для других коллекций.
В идеале было бы очень удобно, если бы была реализована функция Escape для любых типов у которых есть begin и end, которая работала бы как для std::vector.

Jemalloc Installation Hint

Hi!
I found an error in the hint text, when building the project in debug on debian, the hint suggested installing the corporate version of the Jemalloc library.

And finally, thank you very much for putting your framework in Open Source.

Странное определение `operator bool()`

userver/core/include/userver/server/http/form_data_arg.hpp

Lines 15 to 22 in cb8efc3

 struct FormDataArg { 

 std::string_view value; 

 std::string_view content_disposition; 

 std::optional<std::string> filename; 

 std::optional<std::string> default_charset; 

 std::optional<std::string_view> content_type; 

 explicit operator bool() const { return !!value.data(); }

Непонятное назначение двух инверсий. Ведь value.data(), указатель и сам скастуется в bool при необходимости. Это преобразование не считается чем-либо постыдным и не вызывает предупреждений ни у каких компиляторов.

Не уверен, что проброс value.data() это лучший способ реализовать operator bool() для данного типа, ведь поле value теоретически может быть пустой строкой:

form_data_arg.value = ""

И тогда operator bool() вернет true, хотя, возможно, должен вернуть false. Я недостаточно хорошо знаю код, чтобы понять, какая семантика требуется.

Ошибка при билде, Ubuntu 20.08,

При билде тестового сервиса созданного из шаблона получаю такую ошибку:
File "/media/sf_userver/test_service/third_party/userver/plugins/external_deps/impl/cmake_generator.py", line 211, in
main()
File "/media/sf_userver/test_service/third_party/userver/plugins/external_deps/impl/cmake_generator.py", line 206, in main
with open(file_path, 'w') as output_file:
FileNotFoundError: [Errno 2] No such file or directory: '/media/sf_userver/test_service/build_debug/cmake_generated/FindIconv::Iconv.cmake'
CMake Error at third_party/userver/CMakeLists.txt:70 (message):
Generating cmake files failed with exit code: 1
Я вижу, что не генерится какой-то файл с путями к какой-то либе, но её название абсолютное невзрачное. Можете подсказать, что я делаю не так?

Readme documentation

Hi,

On Readme documentation i see a block of code about postgres, but i don't see nothing related to the server.

Can you add to readme the source code to start the server or something like this?

Thanks.

utils::Async from non-worker threads

Is utils::Async(...) supposed to be callable from non-worker threads (say, ev)?

I got the following code

UTEST(UtilsAsync, FromNonWorkerThread) {
  auto& ev_thread = engine::current_task::GetEventThread();
  auto& task_processor = engine::current_task::GetTaskProcessor();
  engine::TaskWithResult<void> task;

  ev_thread.RunInEvLoopSync([&task_processor, &task] {
    task = utils::Async(task_processor, "just_a_task", []{});
  });

  task.Wait();
}

which asserts somewhere in Span internals https://pastebin.com/raw/JdLrwQ5x,
and i feel like this behavior should either be documented or changed in a way that Span gets instantiated in a thread that actually runs a task, not a thread that created it.
Or am i missing something?

P.S. engine::AsyncNoSpan works as expected

Compile error when CentOS 7.8 based on 4.19 kernel

I tried to compile userver on CentOS 7.8 based on 4.19 kernel and got the following error, on the same machine using https://github.com/userver-framework/docker-userver-build-base/blob/develop/ The compilation in the dockerfile generated by Dockerfile does not have this problem.

$ make
[  1%] Built target userver-uboost-coro
[  1%] Built target fmt
[  6%] Built target c-ares
[ 18%] Built target libcurl
[ 32%] Built target cryptopp-static
[ 33%] Built target cctz
[ 33%] Built target userver-stacktrace
Scanning dependencies of target userver-core
[ 33%] Building CXX object userver/core/CMakeFiles/userver-core.dir/__/shared/src/crypto/base.cpp.o
[ 33%] Building CXX object userver/core/CMakeFiles/userver-core.dir/__/shared/src/crypto/base64.cpp.o
[ 33%] Building CXX object userver/core/CMakeFiles/userver-core.dir/__/shared/src/crypto/certificate.cpp.o
[ 33%] Building CXX object userver/core/CMakeFiles/userver-core.dir/__/shared/src/crypto/hash.cpp.o
[ 33%] Building CXX object userver/core/CMakeFiles/userver-core.dir/__/shared/src/crypto/helpers.cpp.o
[ 34%] Building CXX object userver/core/CMakeFiles/userver-core.dir/__/shared/src/crypto/private_key.cpp.o
[ 34%] Building CXX object userver/core/CMakeFiles/userver-core.dir/__/shared/src/crypto/public_key.cpp.o
[ 34%] Building CXX object userver/core/CMakeFiles/userver-core.dir/__/shared/src/crypto/signers.cpp.o
[ 34%] Building CXX object userver/core/CMakeFiles/userver-core.dir/__/shared/src/crypto/verifiers.cpp.o
[ 34%] Building CXX object userver/core/CMakeFiles/userver-core.dir/__/shared/src/formats/json/impl/mutable_value_wrapper.cpp.o
In file included from userver/core/../shared/include/userver/formats/json/impl/mutable_value_wrapper.hpp:9,
                 from userver/shared/src/formats/json/impl/mutable_value_wrapper.cpp:1:
userver/core/../shared/include/userver/utils/fast_pimpl.hpp: In instantiation of 'static void userver::utils::FastPimpl<T, Size, Alignment, Strict>::Validate() [with long unsigned int ActualSize = 64; long unsigned int ActualAlignment = 8; T = userver::formats::json::impl::MutableValueWrapper::Impl; long unsigned int Size = 88; long unsigned int Alignment = 8; bool Strict = true]':
userver/core/../shared/include/userver/utils/fast_pimpl.hpp:86:36:   required from 'userver::utils::FastPimpl<T, Size, Alignment, Strict>::~FastPimpl() [with T = userver::formats::json::impl::MutableValueWrapper::Impl; long unsigned int Size = 88; long unsigned int Alignment = 8; bool Strict = true]'
userver/shared/src/formats/json/impl/mutable_value_wrapper.cpp:48:1:   required from here
userver/core/../shared/include/userver/utils/fast_pimpl.hpp:95:27: error: static assertion failed: invalid Size: Size == sizeof(T) failed
   95 |     static_assert(!Strict || Size == ActualSize,
      |                   ~~~~~~~~^~~~~~~~~~~~~~~~~~~~~
userver/core/../shared/include/userver/utils/fast_pimpl.hpp:95:27: note: '(false || (88 == 64))' evaluates to false
make[2]: *** [userver/core/CMakeFiles/userver-core.dir/build.make:317: userver/core/CMakeFiles/userver-core.dir/__/shared/src/formats/json/impl/mutable_value_wrapper.cpp.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:858: userver/core/CMakeFiles/userver-core.dir/all] Error 2
make: *** [Makefile:161: all] Error 2

more info

$ lscpu
Architecture:          x86_64
CPU op-mode(s):        32-bit, 64-bit
Byte Order:            Little Endian
CPU(s):                80
On-line CPU(s) list:   0-79
Thread(s) per core:    2
Core(s) per socket:    20
Socket(s):             2
NUMA node(s):          2
Vendor ID:             GenuineIntel
CPU family:            6
Model:                 85
Model name:            Intel(R) Xeon(R) Gold 5218R CPU @ 2.10GHz
Stepping:              7
CPU MHz:               2099.996
CPU max MHz:           2100.0000
CPU min MHz:           800.0000
BogoMIPS:              4200.00
Virtualization:        VT-x
L1d cache:             32K
L1i cache:             32K
L2 cache:              1024K
L3 cache:              28160K
NUMA node0 CPU(s):     0-19,40-59
NUMA node1 CPU(s):     20-39,60-79
Flags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cdp_l3 invpcid_single intel_ppin ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb intel_pt avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req pku ospke avx512_vnni md_clear flush_l1d arch_capabilities

$ cat /etc/centos-release
CentOS Linux release 7.8.2003 (Core)

$ uname -a
Linux no.docker 4.19.136 #1 SMP Fri Nov 6 17:54:05 CST 2020 x86_64 x86_64 x86_64 GNU/Linux

$ c++ -v
Using built-in specs.
COLLECT_GCC=c++
COLLECT_LTO_WRAPPER=/opt/rh/devtoolset-11/root/usr/libexec/gcc/x86_64-redhat-linux/11/lto-wrapper
Target: x86_64-redhat-linux
Configured with: ../configure --enable-bootstrap --enable-languages=c,c++,fortran,lto --prefix=/opt/rh/devtoolset-11/root/usr --mandir=/opt/rh/devtoolset-11/root/usr/share/man --infodir=/opt/rh/devtoolset-11/root/usr/share/info --with-bugurl=http://bugzilla.redhat.com/bugzilla --enable-shared --enable-threads=posix --enable-checking=release --enable-multilib --with-system-zlib --enable-__cxa_atexit --disable-libunwind-exceptions --enable-gnu-unique-object --enable-linker-build-id --with-gcc-major-version-only --with-linker-hash-style=gnu --with-default-libstdcxx-abi=gcc4-compatible --enable-plugin --enable-initfini-array --with-isl=/builddir/build/BUILD/gcc-11.2.1-20220127/obj-x86_64-redhat-linux/isl-install --enable-gnu-indirect-function --with-tune=generic --with-arch_32=x86-64 --build=x86_64-redhat-linux
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 11.2.1 20220127 (Red Hat 11.2.1-9) (GCC)

$ cmake --version
cmake3 version 3.17.5

CMake suite maintained and supported by Kitware (kitware.com/cmake).
$ cmake                                          
-DCMAKE_INSTALL_PREFIX=$HOME/local           \
-DBUILD_EXAMPLES=0                           \
-DOPENSSL_ROOT_DIR=/usr/include/openssl11/   \
-DBoost_INCLUDE_DIR=/usr/include/boost169/   \
-DBoost_LIBRARY_DIR=/usr/lib64/boost169/     \
                                             \
-DUSERVER_DOWNLOAD_PACKAGES=1                \
-DUSERVER_FEATURE_GRPC=0                     \
-DUSERVER_FEATURE_REDIS=0                    \
-DUSERVER_FEATURE_MONGODB=0                  \
-DUSERVER_FEATURE_POSTGRESQL=0               \
-DUSERVER_FEATURE_CLICKHOUSE=0               \
-DUSERVER_FEATURE_STACKTRACE=0               \
-DUSERVER_FEATURE_SPDLOG_TCP_SINK=0          \
..

Can't build clickhouse service sample with test suite disabled

I'm trying to build userver without test suite enabled (DUSERVER_FEATURE_TESTSUITE=OFF), but clickhouse_service sample fails to build because of non-existent target testsuite-userver-samples-clickhouse_service.

Related Cmake log:

-- Testsuite target testsuite-userver-samples-clickhouse_service is disabled
CMake Error at samples/clickhouse_service/CMakeLists.txt:7 (set_tests_properties):
  set_tests_properties Can not find test to add properties to:
  testsuite-userver-samples-clickhouse_service


-- Configuring incomplete, errors occurred!
See also "/home/popov/projects/github/userver/cmake-build-debug/CMakeFiles/CMakeOutput.log".
See also "/home/popov/projects/github/userver/cmake-build-debug/CMakeFiles/CMakeError.log".

[Failed to reload]

It would be nice if the sample would work without test suite enabled. Maybe some check can be added before using set_tests_properties in clickhouse_service

Performance comparison with our project

Do you have a tcp echo server that I can run with? So I can add the figures into this list.

Consider renaming NonFifoQueues

I feel like NonFifo is very unclear in terms of what a queue actually represents and leaves too wide of a room for an interpretation: stack is NonFifo, dequeing random element is NonFifo etc.
Maybe something in the lines of CasualOrder/NonLinearizable would fit better?

To add to that, NonFifoSp[s|m]cQueue is actually fifo, because it serializes writes through a single producer token, as suggested in moodycamel documentation:

However, if for some reason you do extra explicit synchronization between the two producer threads yourself, thus defining a total order between enqueue operations, you might expect that the elements would come out in the same total order, which is a guarantee my queue does not offer. At that point, though, there semantically aren't really two separate producers, but rather one that happens to be spread across multiple threads. In this case, you can still establish a total ordering with my queue by creating a single producer token, and using that from both threads to enqueue (taking care to synchronize access to the token, of course, but there was already extra synchronization involved anyway).

There also exists https://github.com/cameron314/readerwriterqueue, which might be a better fit for spsc queue version

[feature-request] Add option to disable TCP_NODELAY for connection socket

Doing -3 lines of code makes wonders when running TechEmpower pipelined plaintext benchmarks on 24-cores VM (16 worker threads, 4 ev threads, 6 threads for wrk):

with TCP_NODELAY:

Concurrency: 1024 for plaintext
 wrk -H 'Host: tfb-server' -H 'Accept: text/plain,text/html;q=0.9,application/xhtml+xml;q=0.9,application/xml;q=0.8,*/*;q=0.7' -H 'Connection: keep-alive' --latency -d 15 -c 1024 --timeout 8 -t 24 http://tfb-server:8090/plaintext -s pipeline.lua -- 16
---------------------------------------------------------
Running 15s test @ http://tfb-server:8090/plaintext
  24 threads and 1024 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    23.35ms   18.20ms 164.49ms   77.32%
    Req/Sec    20.61k     3.52k   78.08k    81.24%
  Latency Distribution
     50%   19.16ms
     75%   31.32ms
     90%   45.71ms
     99%   90.12ms
  7417080 requests in 15.09s, 2.19GB read
Requests/sec: 491395.99
Transfer/sec:    148.56MB

without TCP_NODELAY:

Concurrency: 1024 for plaintext
 wrk -H 'Host: tfb-server' -H 'Accept: text/plain,text/html;q=0.9,application/xhtml+xml;q=0.9,application/xml;q=0.8,*/*;q=0.7' -H 'Connection: keep-alive' --latency -d 15 -c 1024 --timeout 8 -t 24 http://tfb-server:8090/plaintext -s pipeline.lua -- 16
---------------------------------------------------------
Running 15s test @ http://tfb-server:8090/plaintext
  24 threads and 1024 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency    20.91ms   16.76ms 192.81ms   74.69%
    Req/Sec    24.92k     3.27k   93.68k    88.27%
  Latency Distribution
     50%   15.95ms
     75%   28.38ms
     90%   46.78ms
     99%   71.08ms
  8957662 requests in 15.09s, 2.64GB read
Requests/sec: 593597.99
Transfer/sec:    179.45MB

I do understand that the case is somewhat specific, but hey, it's 20% speedup!

P.S. would be great if someone could recheck me here

Compile error when Python < 3.6

Please, pre-check Python version to prevent this error. Thank you.

I have compiled using Python 3.4:

File "userver/plugins/external_deps/impl/cmake_generator.py", line 81

result: dict = {}
SyntaxError: invalid syntax
CMake Error at CMakeLists.txt:70 (message):
Generating cmake files failed with exit code: 1

Tsan is unhappy with `engine::io::Socket`

Running socket benchmarks (--benchmark_filter='socket*') produces this tsan output
socket_tsan.txt

Is it FP or is some bug actually lurking?

TlsWrapper busy-loops on some SSL errors

Setup to reproduce:

take this haproxy config (it's just a default one with some backends+frontends added)
start haproxy with it (disable haproxy.service beforehand) sudo /usr/sbin/haproxy -f /etc/haproxy/haproxy.cfg
run nc -l 5672
take this test and run it
wait a couple of seconds and kill -9 your haproxy process from step 2
code loops here with read_ret = 0 and ssl_error = 2 taking ~100% of CPU

itrofimow@itrofimow-RedmiBook-16:~$ openssl version
OpenSSL 1.1.1  11 Sep 2018
itrofimow@itrofimow-RedmiBook-16:~$ cat /etc/lsb-release 
DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=18.04
DISTRIB_CODENAME=bionic
DISTRIB_DESCRIPTION="Ubuntu 18.04.6 LTS"

Consider removing `Task::Detach` from public API

I'm yet too see a correct usage of Detach in user code, and there were (and could still be) some misuses even in the framework itself

People coming from .Net might think 'hey, it's Task.Run()!', people coming from Rust might find it similar to tokio::spawn or the likes of, however there is one very important part missing: lifetimes.
.Net takes care of this with GC, rust does it's best with borrow checker, but neither is present in userver

Detaching a task basically gives it 'static lifetime, thus capturing anything by reference without proper synchronization is straight UB, and even capturing by value is very dangerous - is there something holding a reference/pointer down the ownership chain? - You never know.
Talking about proper synchronization - there is BackgroundTaskStorage, which got quite some love in performance regards recently, and it solves all the problems. Honestly i can't come up with a case that can not be solved with BTS but could be solved by Detach, in user code especially.

So my proposal is to either remove .Detach from public API completely, or warn a user about all the problematic implications of detaching a task in documentation

RabbitMQ driver enhancements

Would be great to see the following features implemented in RabbitMQ driver:

Dynamic config with some sane timeout defaults
Headers in both publish and admin methods (headers for publish methods might be convenient for additional metadata, headers for admin methods allow one to do this , for example)
Metrics - as of now they are mostly just a bunch of placeholders, and they could be way more useful

Continuation of #44, basically

Build error if .proto-file has no rpc declarations

make[3]: *** No rule to make target 'samples/grpc_service/proto/contracts/common_client.usrv.pb.cpp', needed by 'samples/grpc_service/CMakeFiles/userver-samples-grpc_service_proto.dir/proto/contracts/common_client.usrv.pb.cpp.o'. Stop.

example

If you add any rpc declaration to the .proto file, it will be built

Arch Build Fail - `cannot find -lPostgreSQLInternal_LIBRARIES_libpq_a-NOTFOUND`

Hello. I would like to report an issue on Arch Linux.

Build system was generated with this command, as suggested in docs:
cmake -DUSERVER_FEATURE_PATCH_LIBPQ=0 -DCMAKE_BUILD_TYPE=Release ..
Build error:

[ 65%] Linking CXX executable userver-samples-postgres_service
/usr/sbin/ld: cannot find -lPostgreSQLInternal_LIBRARIES_libpq_a-NOTFOUND: No such file or directory
clang-14: error: linker command failed with exit code 1 (use -v to see invocation)
make[2]: *** [samples/postgres_service/CMakeFiles/userver-samples-postgres_service.dir/build.make:128: samples/postgres_service/userver-samples-postgres_service] Error 1
make[1]: *** [CMakeFiles/Makefile2:3265: samples/postgres_service/CMakeFiles/userver-samples-postgres_service.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....

As I can see, PG's internal lib libpq was not found at configure process, but build system generation step was not failed and the issue was not reported. Tested on two different machines. I tries to install / reinstall packages from docs both with pacman and aur tool.

My system info:

clang version 14.0.6
NAME="Arch Linux"
PRETTY_NAME="Arch Linux"
ID=arch
BUILD_ID=rolling
IMAGE_VERSION=2022.02.01

I also tried to generate build system with this command cmake -DCMAKE_BUILD_TYPE=Release .. (without turning off patch for libpq).

This time generating build system fails with the following output:

-- Installed version postgresql-libs: 14.5
CMake Error at /usr/share/cmake/Modules/FindPackageHandleStandardArgs.cmake:230 (message):
  Could not find `PostgreSQL` package.

        Debian: sudo apt update && sudo apt install libpq-dev

        MacOS: brew install postgresql@14

        Fedora: sudo dnf install postgresql-static

        ArchLinux: sudo pacman -S postgresql-libs

   (missing: PostgreSQL_LIBRARIES) (Required is at least version "12")
Call Stack (most recent call first):
  /usr/share/cmake/Modules/FindPackageHandleStandardArgs.cmake:594 (_FPHSA_FAILURE_MESSAGE)
  build_release/cmake_generated/FindPostgreSQL.cmake:69 (find_package_handle_standard_args)
  postgresql/pq-extra/CMakeLists.txt:14 (find_package)

The required package is installed.

❯ sudo pacman -S postgresql-libs
warning: postgresql-libs-14.5-1 is up to date -- reinstalling
resolving dependencies...
looking for conflicting packages...

Packages (1) postgresql-libs-14.5-1

Total Installed Size:  6.90 MiB
Net Upgrade Size:      0.00 MiB

:: Proceed with installation? [Y/n] ^

This may relate to #88

Socket functions do not always honor deadline

Deadline is only checked if a socket would block and some waiting should be performed, however it's completely possible for read/write to go through without blocking.

One could expect that any attempt to do some work after deadline is reached would fail, but as of now it's not the case with Socket

Wrong documentation link in README

Hi!

Thanks for the library! I guess now it is kinda proven that userver really exists :)

Regarding the problem. In README you have a link to the documentation: See the docs for more info. Please change the link to the actual documentation which is public accessible.

Thanks in advance!

[arch linux] ld can't find -lHttp_Parser

[100%] Linking CXX executable userver-samples-hello_service
/usr/bin/ld: невозможно найти -lHttp_Parser: Нет такого файла или каталога
collect2: ошибка: выполнение ld завершилось с кодом возврата 1

 ❯ pacman -Ql http-parser
http-parser /usr/
http-parser /usr/include/
http-parser /usr/include/http_parser.h
http-parser /usr/lib/
http-parser /usr/lib/libhttp_parser.so
http-parser /usr/lib/libhttp_parser.so.2.9
http-parser /usr/lib/libhttp_parser.so.2.9.4
http-parser /usr/share/
http-parser /usr/share/licenses/
http-parser /usr/share/licenses/http-parser/
http-parser /usr/share/licenses/http-parser/LICENSE-MIT

Apparently it's case-sensitive and package for arch linux comes with different case.
Manual replacement of -lHttp_Parser with -lhttp_parser in build files helped.

SQL dbms support

First of all, thank you for userver. That's one small step for man, one giant leap for mankind.
I see you use postgres and mongo. Is there any chances for other dbms such as mysql, mssql or oracle?

Submodules URLs

In .gitmodules there are links to the internal Yandex repository inaccessible from the outer web

Consider using `moodycamel::readerwriterqueue` for SpscQueue

There exists https://github.com/cameron314/readerwriterqueue, which might be a better fit for concurrent::SpscQueue – it has to be faster than full-featured moodycamel::ConcurrentQueue and it should also require less memory.

There is some confusion between readme and docs:
In readme it's stated that

It only supports a two-thread use case (one consuming, and one producing). The threads can't switch roles, ...

However inline docs elaborate that, leaving a room for experimentation.

Error from 'dynamic-config-client-updater' updater: Field 'USERVER_HTTP_PROXY' is of a wrong type. Expected: stringValue, actual: objectValue

Hey guys,
when connecting a dynamic-config-client-updater to a config server that's configured to use a postgresql db this error appears:

tskv timestamp=2022-08-26T17:33:41.127367 level=ERROR module=DoRun ( third_party/userver/core/src/components/run.cpp:164 ) task_id=0 thread_id=0x00007F1E59D55AC0 text=Loading failed: Error from 'dynamic-config-client-updater' updater: Field 'USERVER_HTTP_PROXY' is of a wrong type. Expected: stringValue, actual: objectValue (std::runtime_error) tskv timestamp=2022-08-26T17:33:41.129100 level=ERROR module=DaemonMain ( third_party/userver/core/src/utils/daemon_run.cpp:72 ) task_id=0 thread_id=0x00007F1E59D55AC0 text=Unhandled exception in components::Run: Error from 'dynamic-config-client-updater' updater: Field 'USERVER_HTTP_PROXY' is of a wrong type. Expected: stringValue, actual: objectValue Unhandled exception in components::Run: Error from 'dynamic-config-client-updater' updater: Field 'USERVER_HTTP_PROXY' is of a wrong type. Expected: stringValue, actual: objectValue

It's expecting a string but config_value is a json column.
The error doesn't happen if switching to the dynamic_config_fallback.json file there the value of USERVER_HTTP_PROXY is correctly a string.
Thanks

[feature-request] Replace `std::unordered_map` with flat map for http headers

Looking at flamegraphs from #141 (comment) one can see that unordered_map takes ~10-11% off total time, with most of it coming from http internals (and some from TimeStorage).

There is an idea that it could very well be replaced with flat map, that would fall back to unordered_map after some threshold, and chances are it would perform better - not only std::unordered_map is pretty slow and linear search might perform better for small number of keys, but there also is that HashDOS-preventing hash, which ain't the fastest. Also my experience tells me that it's not that common to have, say, 50+ headers in http.

What is also promising is that this abstract HttpHeadersMap could be implemented via on stack buffer containing data in http-format right away! - just add some array of string_views which indexes into that buffer. Not sure whether that would work for requests, but should be possible for responses, and for responses we could just pass it's content in Socket::SendAll as is

Maybe make it available in Conan and Vcpkg?

Even though it's pretty convenient to build because of the great documentation you guys provide. But it's still helpful if it's available in the famous package management. I do think Conan and Vcpkg should be considered.
And by the way, thank you and it's a brilliant framework.

userver build failed

hi:
When I try to build userver, got errors below, It seems something error with libabsl?

operating system: Arch linux
cc: g++ (GCC) 12.1.1 20220730

cmake command: cmake -DUSERVER_FEATURE_PATCH_LIBPQ=0 -DCMAKE_BUILD_TYPE=Release ..
make output(errors):

/usr/bin/ld: libuserver-samples-grpc_service_proto.a(greeter.grpc.pb.cc.o): undefined reference to symbol '_ZN4absl12lts_202206235MutexD1Ev'
/usr/bin/ld: /usr/lib/libabsl_synchronization.so.2206.0.0: error adding symbols: DSO missing from command line
collect2: error: ld returned 1 exit status
make[2]: *** [samples/grpc_service/CMakeFiles/userver-samples-grpc_service.dir/build.make:129: samples/grpc_service/userver-samples-grpc_service] Error 1

RabbitMQ (amqp-protocol) support

Thanks for the long awaited framework. You have done a great job!

I would really like to get support for the async amqp protocol,
are there any plans to add it in the near future? Thank you!

	inline bool operator!=(std::string_view lhs, const UserScope& rhs) {
	return lhs == rhs.GetValue();
	}

	inline bool operator==(std::string_view lhs, const UserScope& rhs) {
	return lhs == rhs.GetValue();
	}

	struct FormDataArg {
	std::string_view value;
	std::string_view content_disposition;
	std::optional<std::string> filename;
	std::optional<std::string> default_charset;
	std::optional<std::string_view> content_type;

	explicit operator bool() const { return !!value.data(); }