Various performance improvements #224

hikinggrass · 2024-12-21T20:49:19Z

contains performance enhancements in:
- schema loading, config parsing and validation
- async MQTT config distribution
- publishing of manifests and module names
- handling of types of config items where until now the whole schema was published but only the type was needed
Additionally addresses some clang-tidy issues
Adds a new ctor for everestpy to allow MQTT based config distribution to be used in everest-testing more easily
Remove retained topics after startup and add flag that prevents this from happening for introspection

Some quick-and-dirty performance comparisons starting up the config-sil with a recent everest-core main. This mostly shows that config loading got faster (by cleaning up the usage of schema validators) and a decrease in individual module startup time. Interestingly the total startup time is pretty similar which warrants further investigation:

(v0.19.2 vs v0.19.1)

manager          :: everest-framework 0.19.2 feature/performance-improvements@v0.19.1-35-g64e9ae2
manager          :: Config loading completed in 188ms
manager          :: Starting 16 modules
api:API          :: Module api initialized [53ms]
auth:Auth        :: Module auth initialized [55ms]
connector_1:Evs  :: Module connector_1 initialized [81ms]
connector_1_pow  :: Module connector_1_powerpath initialized [57ms]
energy_manager:  :: Module energy_manager initialized [44ms]
error_history:E  :: Module error_history initialized [61ms]
ev_manager:EvMa  :: Module ev_manager initialized [47ms]
evse_security:E  :: Module evse_security initialized [64ms]
grid_connection  :: Module grid_connection_point initialized [68ms]
iso15118_car     :: Module iso15118_car initialized [45ms]
iso15118_charge  :: Module iso15118_charger initialized [61ms]
persistent_stor  :: Module persistent_store initialized [45ms]
setup:Setup      :: Module setup initialized [38ms]
slac:SlacSimula  :: Module slac initialized [42ms]
token_provider:  :: Module token_provider initialized [38ms]
token_validator  :: Module token_validator initialized [35ms]
manager          :: Clearing retained topics published by manager during startup
manager          :: 🚙🚙🚙 All modules are initialized. EVerest up and running [1700ms] 🚙🚙🚙

vs.

manager          :: everest-framework 0.19.1 @v0.19.1
manager          :: Config loading completed in 336ms
api:API          :: Module api initialized [84ms]
auth:Auth        :: Module auth initialized [65ms]
connector_1:Evs  :: Module connector_1 initialized [86ms]
energy_manager:  :: Module energy_manager initialized [76ms]
error_history:E  :: Module error_history initialized [94ms]
ev_manager:EvMa  :: Module ev_manager initialized [89ms]
evse_security:E  :: Module evse_security initialized [96ms]
grid_connection  :: Module grid_connection_point initialized [79ms]
iso15118_charge  :: Module iso15118_charger initialized [87ms]
persistent_stor  :: Module persistent_store initialized [63ms]
setup:Setup      :: Module setup initialized [60ms]
slac:SlacSimula  :: Module slac initialized [70ms]
token_provider:  :: Module token_provider initialized [67ms]
token_validator  :: Module token_validator initialized [65ms]
manager          :: 🚙🚙🚙 All modules are initialized. EVerest up and running [1721ms] 🚙🚙🚙

Comparing the amount of threads a module uses shows a measurable decrease (refactoring the error handling to use less topics/handlers will have an even big impact on eg. ErrorHistory #225 )

(v0.19.2 vs v0.19.1)

manager            35
controller          2
api:API            22
auth:Auth          22
connector_1:Evs   121
connector_1_pow    15
energy_manager:     6
error_history:E   119
ev_manager:EvMa    16
evse_security:E     6
grid_connection     7
python3             5
iso15118_charge     8
persistent_stor     5
setup:Setup        12
slac:SlacSimula     7
token_provider:    13
token_validator     5

vs

manager            35
controller          2
api:API            55
auth:Auth          55
connector_1:Evs   156
connector_1_pow    47
energy_manager:    39
error_history:E   151
ev_manager:EvMa    50
evse_security:E    38
grid_connection    40
python3            37
iso15118_charge    41
persistent_stor    37
setup:Setup        45
slac:SlacSimula    39
token_provider:    45
token_validator    37

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

The same functionality can be achieved without first including the main config in the serialized config and then deleting it... Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Now when a schema is loaded a validator for it is instantiated as well Then validators can then be re-used instead of instantiating new ones for the same schema multiple times Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Skip storing it in a temporary variable Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

…ount of space in the string the file content gets read into Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

…imum length Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

…etain Previously this was sent to every module individually which isn't necessary Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Add type to parsed_config_map and remove config entry from manifest sent via MQTT Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

refactor interfaces, manifests and types get_async Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

…e is required Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

…logging configuration directly Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

…sion ctor that accepts config file Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

…rtup Remove them by default now but allow them to be explicitly kept for inspection Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

…ed by the MQTT abstraction Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

This drastically reduces the amount of lingering threads just hanging around and doing absolutely nothing Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

…mprovements Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

a-w50

These are quite a lot of commits. I would appreciate, if these could be split into several PRs, focused on their real use case like refactor, bugfix, features etc...

a-w50 · 2025-01-07T07:23:54Z

everestpy/src/everest/everestpy.cpp

@@ -130,7 +130,7 @@ PYBIND11_MODULE(everestpy, m) {
        .def(py::init<const std::string&, const RuntimeSession&>())
        .def("say_hello", &Module::say_hello)
        .def("init_done", py::overload_cast<>(&Module::init_done))
-        .def("init_done", py::overload_cast<std::function<void()>>(&Module::init_done))
+        .def("init_done", py::overload_cast<const std::function<void()>&>(&Module::init_done))


This might be a matter of taste. I'm just wondering if now any appearance of any std::function will be passed by reference instead of value?

a-w50 · 2025-01-07T07:25:54Z

everestpy/src/everest/module.cpp

@@ -12,7 +12,7 @@
 std::unique_ptr<Everest::Everest>
 Module::create_everest_instance(const std::string& module_id, const Everest::Config& config,
                                const Everest::RuntimeSettings& rs,
-                                std::shared_ptr<Everest::MQTTAbstraction> mqtt_abstraction) {
+                                const std::shared_ptr<Everest::MQTTAbstraction>& mqtt_abstraction) {


A constant reference to a shared pointer is almost the same as the pointer itself. Everest::MQTTAbstraction& would be the better choice here.

Here it would have probably been fine because Module will outlive Everest, but I've reverted the change in #227

a-w50 · 2025-01-07T07:26:37Z

include/framework/everest.hpp

@@ -66,7 +66,7 @@ class Everest {
    ///
    /// \brief Allows a module to indicate that it provides the given command \p cmd
    ///
-    void provide_cmd(const std::string impl_id, const std::string cmd_name, const JsonCommand handler);
+    void provide_cmd(const std::string& impl_id, const std::string cmd_name, const JsonCommand& handler);


Why const ref'ing only impl_id and not cmd_name?

Addressed in #227

a-w50 · 2025-01-07T07:27:34Z

include/framework/runtime.hpp

@@ -194,7 +194,7 @@ class ModuleLoader {
    explicit ModuleLoader(int argc, char* argv[], ModuleCallbacks callbacks) :
        ModuleLoader(argc, argv, callbacks, {"undefined project", "undefined version", "undefined git version"}){};
    explicit ModuleLoader(int argc, char* argv[], ModuleCallbacks callbacks,
-                          const VersionInformation version_information);
+                          const VersionInformation& version_information);


This should potentially passed by value, as this information is probably only moved to the ModuleLoader.

Addressed in #227

a-w50 · 2025-01-07T07:28:49Z

include/utils/message_queue.hpp

@@ -79,10 +79,10 @@ class MessageHandler {
    void stop();

    /// \brief Adds a \p handler that will receive messages from the queue.
-    void add_handler(std::shared_ptr<TypedHandler> handler);
+    void add_handler(const std::shared_ptr<TypedHandler>& handler);


See my comment above - I think this might not be a good choice.

I've addressed this (by reverting the change and blocking the suggestion of adding const ref to shared_ptr args) in #227
I think this was changed by me just because I wasn't paying too close attention to the recommendation of clang-tidy here

a-w50 · 2025-01-07T08:55:19Z

src/manager.cpp

+        const std::string config_topic = fmt::format("{}/config", config.mqtt_module_prefix(module_name));
+        const Handler module_get_config_handler = [module_name, config_topic, serialized_mod_config,
+                                                   &mqtt_abstraction](const std::string&, const nlohmann::json& json) {
+            mqtt_abstraction.publish(config_topic, serialized_mod_config.dump(), QOS::QOS2);
+        };
+
+        const std::string get_config_topic = fmt::format("{}/get_config", config.mqtt_module_prefix(module_name));
+        module_it->second.get_config_token = std::make_shared<TypedHandler>(
+            HandlerType::ExternalMQTT, std::make_shared<Handler>(module_get_config_handler));
+        mqtt_abstraction.register_handler(get_config_topic, module_it->second.get_config_token, QOS::QOS2);
+


What is the reason for this rearrangement?

My thinking was that since the spawned module will first call get_config and later on ready this should also be reflected in the code here

a-w50 · 2025-01-07T08:56:59Z

everestpy/src/everest/misc.cpp

@@ -63,18 +63,14 @@ RuntimeSession::RuntimeSession(const std::string& prefix, const std::string& con
    // We extract the settings from the config file so everest-testing doesn't break
    const auto ms = Everest::ManagerSettings(prefix, config_file);

-    Everest::Logging::init(ms.runtime_settings->logging_config_file.string());


Why has this been moved? Isn't logging already used before until the module is constructed?

a-w50 · 2025-01-07T08:58:51Z

everestpy/src/everest/everestpy.cpp

+    py::class_<Everest::MQTTSettings>(m, "MQTTSettings")
+        .def(py::init<>())
+        .def_readwrite("broker_socket_path", &Everest::MQTTSettings::broker_socket_path)
+        .def_readwrite("broker_host", &Everest::MQTTSettings::broker_host)
+        .def_readwrite("broker_port", &Everest::MQTTSettings::broker_port)
+        .def_readwrite("everest_prefix", &Everest::MQTTSettings::everest_prefix)
+        .def_readwrite("external_prefix", &Everest::MQTTSettings::external_prefix)
+        .def("uses_socket", &Everest::MQTTSettings::uses_socket);
+


Why has this been added? The more we broaden the api, the harder it gets to refactor.

This allows everest-testing to use the MQTT based config distribution mechanism since it's (as far as I know) almost impossible to set the environment variables in the pytest environment. Usage is implemented here: EVerest/everest-utils#173

a-w50 · 2025-01-07T09:01:38Z

include/utils/mqtt_abstraction_impl.hpp

@@ -80,6 +80,10 @@ class MQTTAbstractionImpl {
    /// \brief unsubscribes from the given \p topic
    void unsubscribe(const std::string& topic);

+    ///
+    /// \brief clears any previously published topics that had the retain flag set
+    void clear_retained_topics();


Is there any documentation on how or when the retain-topics flag should be used?

a-w50 · 2025-01-07T09:12:18Z

src/manager.cpp

@@ -749,6 +753,9 @@ int boot(const po::variables_map& vm) {

    auto module_handles =
        start_modules(*config, mqtt_abstraction, ignored_modules, standalone_modules, ms, status_fifo, retain_topics);
+    if (module_handles.empty()) {
+        return EXIT_FAILURE;


Should this be considered a failure?

That's certainly debatable. The current behavior is that everest will happily load empty configs, but it doesn't signal a completed startup via the status fifo or logs. So for a user it just "hangs" without doing anything else. I thought failing early would be better in this case, but we can also leave the current behavior in place if that's more desirable

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

… is only used there Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

hikinggrass · 2025-01-07T13:53:46Z

include/framework/everest.hpp

@@ -66,7 +66,7 @@ class Everest {
    ///
    /// \brief Allows a module to indicate that it provides the given command \p cmd
    ///
-    void provide_cmd(const std::string impl_id, const std::string cmd_name, const JsonCommand handler);
+    void provide_cmd(const std::string& impl_id, const std::string cmd_name, const JsonCommand& handler);


Addressed in #227

hikinggrass · 2025-01-07T13:53:55Z

include/framework/runtime.hpp

@@ -194,7 +194,7 @@ class ModuleLoader {
    explicit ModuleLoader(int argc, char* argv[], ModuleCallbacks callbacks) :
        ModuleLoader(argc, argv, callbacks, {"undefined project", "undefined version", "undefined git version"}){};
    explicit ModuleLoader(int argc, char* argv[], ModuleCallbacks callbacks,
-                          const VersionInformation version_information);
+                          const VersionInformation& version_information);


Addressed in #227

hikinggrass · 2025-01-07T13:57:03Z

include/utils/mqtt_abstraction_impl.hpp

@@ -20,7 +20,7 @@

 #include <utils/thread.hpp>

-constexpr std::size_t MQTT_BUF_SIZE = 500 * 1024;
+constexpr std::size_t MQTT_BUF_SIZE = std::size_t{500} * std::size_t{1024};


std::size_t{500 * 1024} still produces a widening conversion warning since 500 and 1024 are interpreted as string. In C++23 we could postfix it with uz but I think this way is fine for now (it only fixes a warning anyway). Addressed in #227

hikinggrass · 2025-01-07T13:58:00Z

lib/yaml_loader.cpp

+static std::streamsize clamp(std::size_t len) {
+    return (len <= std::numeric_limits<std::streamsize>::max()) ? static_cast<std::streamsize>(len)
+                                                                : std::numeric_limits<std::streamsize>::max();
+}
+


Addressed in #227

hikinggrass · 2025-01-07T14:19:22Z

lib/yaml_loader.cpp

+    std::ifstream ifs(path.string());
+    ifs.seekg(0, std::ios::end);
+    std::string content;
+    content.reserve(ifs.tellg());
+    ifs.seekg(0);
+    content.assign(std::istreambuf_iterator<char>(ifs), std::istreambuf_iterator<char>());
+    return content;


Agree, I did some quick comparisons between the two and didn't really see much different one way or the other so I reverted this back to the original code

hikinggrass · 2025-01-07T14:20:33Z

src/manager.cpp

@@ -638,7 +638,7 @@ int boot(const po::variables_map& vm) {
    EVLOG_info << ms.version_information;
    EVLOG_info << "";

-    if (ms.mqtt_settings.uses_socket()) {
+    if (not ms.mqtt_settings.uses_socket()) {


Addressed in #228

hikinggrass · 2025-01-07T14:22:45Z

lib/config.cpp

+SchemaValidation Config::load_schemas(const fs::path& schemas_dir) {
    BOOST_LOG_FUNCTION();
-    schemas schemas;
+    SchemaValidation schema_validation;

    EVLOG_debug << fmt::format("Loading base schema files for config and manifests... from: {}", schemas_dir.string());
-    schemas.config = Config::load_schema(schemas_dir / "config.yaml");
-    schemas.manifest = Config::load_schema(schemas_dir / "manifest.yaml");
-    schemas.interface = Config::load_schema(schemas_dir / "interface.yaml");
-    schemas.type = Config::load_schema(schemas_dir / "type.yaml");
-    schemas.error_declaration_list = Config::load_schema(schemas_dir / "error-declaration-list.yaml");
-
-    return schemas;
+    auto [config_schema, config_val] = Config::load_schema(schemas_dir / "config.yaml");
+    schema_validation.schemas.config = config_schema;
+    schema_validation.validators.config = std::move(config_val);
+    auto [manifest_schema, manifest_val] = Config::load_schema(schemas_dir / "manifest.yaml");
+    schema_validation.schemas.manifest = manifest_schema;
+    schema_validation.validators.manifest = std::move(manifest_val);
+    auto [interface_schema, interface_val] = Config::load_schema(schemas_dir / "interface.yaml");
+    schema_validation.schemas.interface = interface_schema;
+    schema_validation.validators.interface = std::move(interface_val);
+    auto [type_schema, type_val] = Config::load_schema(schemas_dir / "type.yaml");
+    schema_validation.schemas.type = type_schema;
+    schema_validation.validators.type = std::move(type_val);
+    auto [error_declaration_list_schema, error_declaration_list_val] =
+        Config::load_schema(schemas_dir / "error-declaration-list.yaml");
+    schema_validation.schemas.error_declaration_list = error_declaration_list_schema;
+    schema_validation.validators.error_declaration_list = std::move(error_declaration_list_val);
+
+    return schema_validation;
 }


yes, I've turned both load_schema and load_schemas into free functions

hikinggrass · 2025-01-07T14:22:56Z

include/utils/config.hpp

+    std::unique_ptr<nlohmann::json_schema::json_validator> config;
+    std::unique_ptr<nlohmann::json_schema::json_validator> manifest;
+    std::unique_ptr<nlohmann::json_schema::json_validator> type;
+    std::unique_ptr<nlohmann::json_schema::json_validator> interface;
+    std::unique_ptr<nlohmann::json_schema::json_validator> error_declaration_list;


hikinggrass · 2025-01-07T14:24:03Z

include/utils/mqtt_abstraction.hpp

@@ -64,6 +64,10 @@ class MQTTAbstraction {
    /// \copydoc MQTTAbstractionImpl::unsubscribe(const std::string&)
    void unsubscribe(const std::string& topic);

+    ///
+    /// \copydoc MQTTAbstractionImpl::get_async(const std::string&, QOS)
+    AsyncReturn get_async(const std::string& topic, QOS qos);


I moved this (and get() as well) into module_config.cpp since it is only used there

hikinggrass · 2025-01-07T14:27:09Z

lib/mqtt_abstraction_impl.cpp

@@ -484,6 +472,7 @@ void MQTTAbstractionImpl::unregister_handler(const std::string& topic, const Tok
            EVLOG_verbose << fmt::format("Unsubscribing from {}", topic);
            this->unsubscribe(topic);
        }
+        this->message_handlers.erase(topic);


Opened a new PR for this change which helps reduce the amount of idle threads quite significantly: #229
In general this design needs to be refactored, but that's probably best done in a future PR

…mprovements Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

hikinggrass added 30 commits December 10, 2024 19:43

Add performance checks to clang-tidy config

0962fe3

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Log module startup time for JavaScript and Python modules as well

3f76a88

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Use const ref for a few more variables and parameters

2b17dfb

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Remove serialize function from config

f5d4333

The same functionality can be achieved without first including the main config in the serialized config and then deleting it... Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Refactor schema loading and validation

9d4da56

Now when a schema is loaded a validator for it is instantiated as well Then validators can then be re-used instead of instantiating new ones for the same schema multiple times Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Ad some moves that clang-tidy suggested

9d5496c

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Directly pass the active_modules part of a config to the parse function

a93bdfa

Skip storing it in a temporary variable Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Improve yaml file loading performance by reserving the appropriate am…

c74055b

…ount of space in the string the file content gets read into Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Log the number of modules started

b8385de

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Fix narrowing conversion warning in yaml error handler by claming max…

0bc601b

…imum length Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Fix widening conversion warning by multiplying two std::size_t already

bb1c2de

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Bump version to 0.19.2

0acb24f

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Only publish module_names (a mapping of module type to id) once and r…

9ddb9ce

…etain Previously this was sent to every module individually which isn't necessary Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Publish manifest individually

d40f7c3

Add type to parsed_config_map and remove config entry from manifest sent via MQTT Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

async get() function for MQTT used in get_module_config

dbd4dc2

refactor interfaces, manifests and types get_async Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Fix everestjs config parsing with new config entry format

1077b00

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Capturing more vars as refs

250ee4a

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Log which topic cause a timeout exception in get()

5551fed

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Use get_with_timeout in get_module_config

c2f2f60

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Fix usage of MQTTSettings uses socket in manager

b96933b

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Use get_module_name instead of get_module_info if only the module nam…

e6bf11c

…e is required Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

More const ref usage

5d63f80

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Re-order config handler in manager

04ec269

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

everestpy: initialize logging in module ctor not RuntimeSession

4b6d16e

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

everestpy: add new RuntimeSession ctor that accepts MQTTSettings and …

ed5152e

…logging configuration directly Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

everestpy: add short documentation comments and deprecated RuntimeSes…

30cf7e2

…sion ctor that accepts config file Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

More constref and move usage

9a90570

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Add --retain-topics flag to manager to keep retained topics after sta…

c4737e3

…rtup Remove them by default now but allow them to be explicitly kept for inspection Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Remove old cleanup_retained_topics functions since this is now provid…

61603c9

…ed by the MQTT abstraction Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

clang-format

6c19fb4

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

hikinggrass mentioned this pull request Dec 23, 2024

EVerest crash issue observed when using v0.19.1 framework #223

Open

hikinggrass added 5 commits January 2, 2025 14:57

Erase MessageHandler for topics without any registered handlers

6405e8e

This drastically reduces the amount of lingering threads just hanging around and doing absolutely nothing Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Use .at() instead of [] for map access

7c539ac

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Exit early with EXIT_FAILURE if there are no modules to start

1e4f79e

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Merge remote-tracking branch 'origin/main' into feature/performance-i…

64e9ae2

…mprovements Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Make logged duration of get_module_config less verbose

9b68f6d

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

hikinggrass requested a review from a-w50 January 6, 2025 09:08

a-w50 reviewed Jan 7, 2025

View reviewed changes

hikinggrass added 5 commits January 7, 2025 14:08

Remove redundant unique_ptrs in Validators struct

abad2b8

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Turn load_schema and load_schemas into free functions

3814cdd

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Move get and get_async from MQTTAbstraction to module config since it…

ad1ecda

… is only used there Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

load_schema and load_schemas: clang-format

96835c0

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

Revert to old file loading code for redability reasons

b42680a

Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

hikinggrass commented Jan 7, 2025

View reviewed changes

Merge remote-tracking branch 'origin/main' into feature/performance-i…

560c5ab

…mprovements Signed-off-by: Kai-Uwe Hermann <kai-uwe.hermann@pionix.de>

hikinggrass marked this pull request as draft January 13, 2025 10:27

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Various performance improvements #224

Various performance improvements #224

hikinggrass commented Dec 21, 2024 •

edited

Loading

a-w50 left a comment

a-w50 Jan 7, 2025

a-w50 Jan 7, 2025

hikinggrass Jan 7, 2025

a-w50 Jan 7, 2025

hikinggrass Jan 7, 2025

a-w50 Jan 7, 2025

hikinggrass Jan 7, 2025

a-w50 Jan 7, 2025

hikinggrass Jan 7, 2025

a-w50 Jan 7, 2025

hikinggrass Jan 7, 2025

a-w50 Jan 7, 2025

a-w50 Jan 7, 2025

hikinggrass Jan 7, 2025

a-w50 Jan 7, 2025

a-w50 Jan 7, 2025

hikinggrass Jan 7, 2025

hikinggrass Jan 7, 2025

hikinggrass Jan 7, 2025

hikinggrass Jan 7, 2025

hikinggrass Jan 7, 2025

hikinggrass Jan 7, 2025

hikinggrass Jan 7, 2025

hikinggrass Jan 7, 2025

hikinggrass Jan 7, 2025

hikinggrass Jan 7, 2025

hikinggrass Jan 7, 2025

Various performance improvements #224

Are you sure you want to change the base?

Various performance improvements #224

Conversation

hikinggrass commented Dec 21, 2024 • edited Loading

a-w50 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hikinggrass commented Dec 21, 2024 •

edited

Loading