summaryrefslogtreecommitdiffstats
path: root/fluent-bit/lib/librdkafka-2.1.0/src-cpp
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-03-09 13:19:48 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-03-09 13:20:02 +0000
commit58daab21cd043e1dc37024a7f99b396788372918 (patch)
tree96771e43bb69f7c1c2b0b4f7374cb74d7866d0cb /fluent-bit/lib/librdkafka-2.1.0/src-cpp
parentReleasing debian version 1.43.2-1. (diff)
downloadnetdata-58daab21cd043e1dc37024a7f99b396788372918.tar.xz
netdata-58daab21cd043e1dc37024a7f99b396788372918.zip
Merging upstream version 1.44.3.
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'fluent-bit/lib/librdkafka-2.1.0/src-cpp')
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/CMakeLists.txt90
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/ConfImpl.cpp84
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/ConsumerImpl.cpp244
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/HandleImpl.cpp425
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/HeadersImpl.cpp48
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/KafkaConsumerImpl.cpp296
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/Makefile55
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/MessageImpl.cpp38
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/MetadataImpl.cpp170
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/ProducerImpl.cpp197
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/QueueImpl.cpp70
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/README.md16
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/RdKafka.cpp59
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/TopicImpl.cpp124
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/TopicPartitionImpl.cpp57
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/rdkafkacpp.h3764
-rw-r--r--fluent-bit/lib/librdkafka-2.1.0/src-cpp/rdkafkacpp_int.h1628
17 files changed, 7365 insertions, 0 deletions
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/CMakeLists.txt b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/CMakeLists.txt
new file mode 100644
index 00000000..b0a6d51e
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/CMakeLists.txt
@@ -0,0 +1,90 @@
+set(LIBVER 1)
+
+set(
+ sources
+ ConfImpl.cpp
+ ConsumerImpl.cpp
+ HandleImpl.cpp
+ HeadersImpl.cpp
+ KafkaConsumerImpl.cpp
+ MessageImpl.cpp
+ MetadataImpl.cpp
+ ProducerImpl.cpp
+ QueueImpl.cpp
+ RdKafka.cpp
+ TopicImpl.cpp
+ TopicPartitionImpl.cpp
+)
+
+if(RDKAFKA_BUILD_STATIC)
+ set(CMAKE_POSITION_INDEPENDENT_CODE ON)
+ set(RDKAFKA_BUILD_MODE STATIC)
+else()
+ set(RDKAFKA_BUILD_MODE SHARED)
+endif()
+
+add_library(rdkafka++ ${RDKAFKA_BUILD_MODE} ${sources})
+if(NOT RDKAFKA_BUILD_STATIC)
+ set_property(TARGET rdkafka++ PROPERTY SOVERSION ${LIBVER})
+endif()
+
+target_link_libraries(rdkafka++ PUBLIC rdkafka)
+
+# Support '#include <rdkafcpp.h>'
+target_include_directories(rdkafka++ PUBLIC "$<BUILD_INTERFACE:${CMAKE_CURRENT_LIST_DIR}>")
+if(NOT RDKAFKA_BUILD_STATIC)
+ target_compile_definitions(rdkafka++ PRIVATE LIBRDKAFKACPP_EXPORTS)
+endif()
+
+# Generate pkg-config file
+set(PKG_CONFIG_VERSION "${PROJECT_VERSION}")
+if(NOT RDKAFKA_BUILD_STATIC)
+ set(PKG_CONFIG_NAME "librdkafka++")
+ set(PKG_CONFIG_DESCRIPTION "The Apache Kafka C/C++ library")
+ set(PKG_CONFIG_REQUIRES "rdkafka")
+ set(PKG_CONFIG_CFLAGS "-I\${includedir}")
+ set(PKG_CONFIG_LIBS "-L\${libdir} -lrdkafka++")
+ set(PKG_CONFIG_LIBS_PRIVATE "-lrdkafka")
+ configure_file(
+ "../packaging/cmake/rdkafka.pc.in"
+ "${GENERATED_DIR}/rdkafka++.pc"
+ @ONLY
+ )
+ install(
+ FILES ${GENERATED_DIR}/rdkafka++.pc
+ DESTINATION "${CMAKE_INSTALL_LIBDIR}/pkgconfig"
+ )
+else()
+ set(PKG_CONFIG_NAME "librdkafka++-static")
+ set(PKG_CONFIG_DESCRIPTION "The Apache Kafka C/C++ library (static)")
+ set(PKG_CONFIG_REQUIRES "")
+ set(PKG_CONFIG_CFLAGS "-I\${includedir} -DLIBRDKAFKA_STATICLIB")
+ set(PKG_CONFIG_LIBS "-L\${libdir} \${libdir}/librdkafka++.a")
+ if(WIN32)
+ string(APPEND PKG_CONFIG_LIBS " -lws2_32 -lsecur32 -lcrypt32")
+ endif()
+
+ configure_file(
+ "../packaging/cmake/rdkafka.pc.in"
+ "${GENERATED_DIR}/rdkafka++-static.pc"
+ @ONLY
+ )
+ install(
+ FILES ${GENERATED_DIR}/rdkafka++-static.pc
+ DESTINATION "${CMAKE_INSTALL_LIBDIR}/pkgconfig"
+ )
+endif()
+
+install(
+ TARGETS rdkafka++
+ EXPORT "${targets_export_name}"
+ LIBRARY DESTINATION "${CMAKE_INSTALL_LIBDIR}"
+ ARCHIVE DESTINATION "${CMAKE_INSTALL_LIBDIR}"
+ RUNTIME DESTINATION "${CMAKE_INSTALL_BINDIR}"
+ INCLUDES DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}"
+)
+
+install(
+ FILES "rdkafkacpp.h"
+ DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}/librdkafka"
+)
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/ConfImpl.cpp b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/ConfImpl.cpp
new file mode 100644
index 00000000..53d7b30c
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/ConfImpl.cpp
@@ -0,0 +1,84 @@
+/*
+ * librdkafka - Apache Kafka C/C++ library
+ *
+ * Copyright (c) 2014 Magnus Edenhill
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <iostream>
+#include <string>
+#include <list>
+
+#include "rdkafkacpp_int.h"
+
+
+
+RdKafka::ConfImpl::ConfResult RdKafka::ConfImpl::set(const std::string &name,
+ const std::string &value,
+ std::string &errstr) {
+ rd_kafka_conf_res_t res;
+ char errbuf[512];
+
+ if (this->conf_type_ == CONF_GLOBAL)
+ res = rd_kafka_conf_set(this->rk_conf_, name.c_str(), value.c_str(), errbuf,
+ sizeof(errbuf));
+ else
+ res = rd_kafka_topic_conf_set(this->rkt_conf_, name.c_str(), value.c_str(),
+ errbuf, sizeof(errbuf));
+
+ if (res != RD_KAFKA_CONF_OK)
+ errstr = errbuf;
+
+ return static_cast<Conf::ConfResult>(res);
+}
+
+
+std::list<std::string> *RdKafka::ConfImpl::dump() {
+ const char **arrc;
+ size_t cnt;
+ std::list<std::string> *arr;
+
+ if (rk_conf_)
+ arrc = rd_kafka_conf_dump(rk_conf_, &cnt);
+ else
+ arrc = rd_kafka_topic_conf_dump(rkt_conf_, &cnt);
+
+ arr = new std::list<std::string>();
+ for (int i = 0; i < static_cast<int>(cnt); i++)
+ arr->push_back(std::string(arrc[i]));
+
+ rd_kafka_conf_dump_free(arrc, cnt);
+ return arr;
+}
+
+RdKafka::Conf *RdKafka::Conf::create(ConfType type) {
+ ConfImpl *conf = new ConfImpl(type);
+
+ if (type == CONF_GLOBAL)
+ conf->rk_conf_ = rd_kafka_conf_new();
+ else
+ conf->rkt_conf_ = rd_kafka_topic_conf_new();
+
+ return conf;
+}
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/ConsumerImpl.cpp b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/ConsumerImpl.cpp
new file mode 100644
index 00000000..b7f5e3b2
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/ConsumerImpl.cpp
@@ -0,0 +1,244 @@
+/*
+ * librdkafka - Apache Kafka C/C++ library
+ *
+ * Copyright (c) 2014 Magnus Edenhill
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <iostream>
+#include <string>
+#include <list>
+#include <cerrno>
+
+#include "rdkafkacpp_int.h"
+
+RdKafka::Consumer::~Consumer() {
+}
+
+RdKafka::Consumer *RdKafka::Consumer::create(const RdKafka::Conf *conf,
+ std::string &errstr) {
+ char errbuf[512];
+ const RdKafka::ConfImpl *confimpl =
+ dynamic_cast<const RdKafka::ConfImpl *>(conf);
+ RdKafka::ConsumerImpl *rkc = new RdKafka::ConsumerImpl();
+ rd_kafka_conf_t *rk_conf = NULL;
+
+ if (confimpl) {
+ if (!confimpl->rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ delete rkc;
+ return NULL;
+ }
+
+ rkc->set_common_config(confimpl);
+
+ rk_conf = rd_kafka_conf_dup(confimpl->rk_conf_);
+ }
+
+ rd_kafka_t *rk;
+ if (!(rk =
+ rd_kafka_new(RD_KAFKA_CONSUMER, rk_conf, errbuf, sizeof(errbuf)))) {
+ errstr = errbuf;
+ // rd_kafka_new() takes ownership only if succeeds
+ if (rk_conf)
+ rd_kafka_conf_destroy(rk_conf);
+ delete rkc;
+ return NULL;
+ }
+
+ rkc->rk_ = rk;
+
+
+ return rkc;
+}
+
+int64_t RdKafka::Consumer::OffsetTail(int64_t offset) {
+ return RD_KAFKA_OFFSET_TAIL(offset);
+}
+
+RdKafka::ErrorCode RdKafka::ConsumerImpl::start(Topic *topic,
+ int32_t partition,
+ int64_t offset) {
+ RdKafka::TopicImpl *topicimpl = dynamic_cast<RdKafka::TopicImpl *>(topic);
+
+ if (rd_kafka_consume_start(topicimpl->rkt_, partition, offset) == -1)
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_last_error());
+
+ return RdKafka::ERR_NO_ERROR;
+}
+
+
+RdKafka::ErrorCode RdKafka::ConsumerImpl::start(Topic *topic,
+ int32_t partition,
+ int64_t offset,
+ Queue *queue) {
+ RdKafka::TopicImpl *topicimpl = dynamic_cast<RdKafka::TopicImpl *>(topic);
+ RdKafka::QueueImpl *queueimpl = dynamic_cast<RdKafka::QueueImpl *>(queue);
+
+ if (rd_kafka_consume_start_queue(topicimpl->rkt_, partition, offset,
+ queueimpl->queue_) == -1)
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_last_error());
+
+ return RdKafka::ERR_NO_ERROR;
+}
+
+
+RdKafka::ErrorCode RdKafka::ConsumerImpl::stop(Topic *topic,
+ int32_t partition) {
+ RdKafka::TopicImpl *topicimpl = dynamic_cast<RdKafka::TopicImpl *>(topic);
+
+ if (rd_kafka_consume_stop(topicimpl->rkt_, partition) == -1)
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_last_error());
+
+ return RdKafka::ERR_NO_ERROR;
+}
+
+RdKafka::ErrorCode RdKafka::ConsumerImpl::seek(Topic *topic,
+ int32_t partition,
+ int64_t offset,
+ int timeout_ms) {
+ RdKafka::TopicImpl *topicimpl = dynamic_cast<RdKafka::TopicImpl *>(topic);
+
+ if (rd_kafka_seek(topicimpl->rkt_, partition, offset, timeout_ms) == -1)
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_last_error());
+
+ return RdKafka::ERR_NO_ERROR;
+}
+
+RdKafka::Message *RdKafka::ConsumerImpl::consume(Topic *topic,
+ int32_t partition,
+ int timeout_ms) {
+ RdKafka::TopicImpl *topicimpl = dynamic_cast<RdKafka::TopicImpl *>(topic);
+ rd_kafka_message_t *rkmessage;
+
+ rkmessage = rd_kafka_consume(topicimpl->rkt_, partition, timeout_ms);
+ if (!rkmessage)
+ return new RdKafka::MessageImpl(
+ RD_KAFKA_CONSUMER, topic,
+ static_cast<RdKafka::ErrorCode>(rd_kafka_last_error()));
+
+ return new RdKafka::MessageImpl(RD_KAFKA_CONSUMER, topic, rkmessage);
+}
+
+namespace {
+/* Helper struct for `consume_callback'.
+ * Encapsulates the values we need in order to call `rd_kafka_consume_callback'
+ * and keep track of the C++ callback function and `opaque' value.
+ */
+struct ConsumerImplCallback {
+ ConsumerImplCallback(RdKafka::Topic *topic,
+ RdKafka::ConsumeCb *cb,
+ void *data) :
+ topic(topic), cb_cls(cb), cb_data(data) {
+ }
+ /* This function is the one we give to `rd_kafka_consume_callback', with
+ * the `opaque' pointer pointing to an instance of this struct, in which
+ * we can find the C++ callback and `cb_data'.
+ */
+ static void consume_cb_trampoline(rd_kafka_message_t *msg, void *opaque) {
+ ConsumerImplCallback *instance =
+ static_cast<ConsumerImplCallback *>(opaque);
+ RdKafka::MessageImpl message(RD_KAFKA_CONSUMER, instance->topic, msg,
+ false /*don't free*/);
+ instance->cb_cls->consume_cb(message, instance->cb_data);
+ }
+ RdKafka::Topic *topic;
+ RdKafka::ConsumeCb *cb_cls;
+ void *cb_data;
+};
+} // namespace
+
+int RdKafka::ConsumerImpl::consume_callback(RdKafka::Topic *topic,
+ int32_t partition,
+ int timeout_ms,
+ RdKafka::ConsumeCb *consume_cb,
+ void *opaque) {
+ RdKafka::TopicImpl *topicimpl = static_cast<RdKafka::TopicImpl *>(topic);
+ ConsumerImplCallback context(topic, consume_cb, opaque);
+ return rd_kafka_consume_callback(topicimpl->rkt_, partition, timeout_ms,
+ &ConsumerImplCallback::consume_cb_trampoline,
+ &context);
+}
+
+
+RdKafka::Message *RdKafka::ConsumerImpl::consume(Queue *queue, int timeout_ms) {
+ RdKafka::QueueImpl *queueimpl = dynamic_cast<RdKafka::QueueImpl *>(queue);
+ rd_kafka_message_t *rkmessage;
+
+ rkmessage = rd_kafka_consume_queue(queueimpl->queue_, timeout_ms);
+ if (!rkmessage)
+ return new RdKafka::MessageImpl(
+ RD_KAFKA_CONSUMER, NULL,
+ static_cast<RdKafka::ErrorCode>(rd_kafka_last_error()));
+ /*
+ * Recover our Topic * from the topic conf's opaque field, which we
+ * set in RdKafka::Topic::create() for just this kind of situation.
+ */
+ void *opaque = rd_kafka_topic_opaque(rkmessage->rkt);
+ Topic *topic = static_cast<Topic *>(opaque);
+
+ return new RdKafka::MessageImpl(RD_KAFKA_CONSUMER, topic, rkmessage);
+}
+
+namespace {
+/* Helper struct for `consume_callback' with a Queue.
+ * Encapsulates the values we need in order to call `rd_kafka_consume_callback'
+ * and keep track of the C++ callback function and `opaque' value.
+ */
+struct ConsumerImplQueueCallback {
+ ConsumerImplQueueCallback(RdKafka::ConsumeCb *cb, void *data) :
+ cb_cls(cb), cb_data(data) {
+ }
+ /* This function is the one we give to `rd_kafka_consume_callback', with
+ * the `opaque' pointer pointing to an instance of this struct, in which
+ * we can find the C++ callback and `cb_data'.
+ */
+ static void consume_cb_trampoline(rd_kafka_message_t *msg, void *opaque) {
+ ConsumerImplQueueCallback *instance =
+ static_cast<ConsumerImplQueueCallback *>(opaque);
+ /*
+ * Recover our Topic * from the topic conf's opaque field, which we
+ * set in RdKafka::Topic::create() for just this kind of situation.
+ */
+ void *topic_opaque = rd_kafka_topic_opaque(msg->rkt);
+ RdKafka::Topic *topic = static_cast<RdKafka::Topic *>(topic_opaque);
+ RdKafka::MessageImpl message(RD_KAFKA_CONSUMER, topic, msg,
+ false /*don't free*/);
+ instance->cb_cls->consume_cb(message, instance->cb_data);
+ }
+ RdKafka::ConsumeCb *cb_cls;
+ void *cb_data;
+};
+} // namespace
+
+int RdKafka::ConsumerImpl::consume_callback(Queue *queue,
+ int timeout_ms,
+ RdKafka::ConsumeCb *consume_cb,
+ void *opaque) {
+ RdKafka::QueueImpl *queueimpl = dynamic_cast<RdKafka::QueueImpl *>(queue);
+ ConsumerImplQueueCallback context(consume_cb, opaque);
+ return rd_kafka_consume_callback_queue(
+ queueimpl->queue_, timeout_ms,
+ &ConsumerImplQueueCallback::consume_cb_trampoline, &context);
+}
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/HandleImpl.cpp b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/HandleImpl.cpp
new file mode 100644
index 00000000..7aa2f293
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/HandleImpl.cpp
@@ -0,0 +1,425 @@
+/*
+ * librdkafka - Apache Kafka C/C++ library
+ *
+ * Copyright (c) 2014 Magnus Edenhill
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <iostream>
+#include <string>
+#include <list>
+
+#include "rdkafkacpp_int.h"
+
+void RdKafka::consume_cb_trampoline(rd_kafka_message_t *msg, void *opaque) {
+ RdKafka::HandleImpl *handle = static_cast<RdKafka::HandleImpl *>(opaque);
+ RdKafka::Topic *topic = static_cast<Topic *>(rd_kafka_topic_opaque(msg->rkt));
+
+ RdKafka::MessageImpl message(RD_KAFKA_CONSUMER, topic, msg,
+ false /*don't free*/);
+
+ handle->consume_cb_->consume_cb(message, opaque);
+}
+
+void RdKafka::log_cb_trampoline(const rd_kafka_t *rk,
+ int level,
+ const char *fac,
+ const char *buf) {
+ if (!rk) {
+ rd_kafka_log_print(rk, level, fac, buf);
+ return;
+ }
+
+ void *opaque = rd_kafka_opaque(rk);
+ RdKafka::HandleImpl *handle = static_cast<RdKafka::HandleImpl *>(opaque);
+
+ if (!handle->event_cb_) {
+ rd_kafka_log_print(rk, level, fac, buf);
+ return;
+ }
+
+ RdKafka::EventImpl event(RdKafka::Event::EVENT_LOG, RdKafka::ERR_NO_ERROR,
+ static_cast<RdKafka::Event::Severity>(level), fac,
+ buf);
+
+ handle->event_cb_->event_cb(event);
+}
+
+
+void RdKafka::error_cb_trampoline(rd_kafka_t *rk,
+ int err,
+ const char *reason,
+ void *opaque) {
+ RdKafka::HandleImpl *handle = static_cast<RdKafka::HandleImpl *>(opaque);
+ char errstr[512];
+ bool is_fatal = false;
+
+ if (err == RD_KAFKA_RESP_ERR__FATAL) {
+ /* Translate to underlying fatal error code and string */
+ err = rd_kafka_fatal_error(rk, errstr, sizeof(errstr));
+ if (err)
+ reason = errstr;
+ is_fatal = true;
+ }
+ RdKafka::EventImpl event(RdKafka::Event::EVENT_ERROR,
+ static_cast<RdKafka::ErrorCode>(err),
+ RdKafka::Event::EVENT_SEVERITY_ERROR, NULL, reason);
+ event.fatal_ = is_fatal;
+ handle->event_cb_->event_cb(event);
+}
+
+
+void RdKafka::throttle_cb_trampoline(rd_kafka_t *rk,
+ const char *broker_name,
+ int32_t broker_id,
+ int throttle_time_ms,
+ void *opaque) {
+ RdKafka::HandleImpl *handle = static_cast<RdKafka::HandleImpl *>(opaque);
+
+ RdKafka::EventImpl event(RdKafka::Event::EVENT_THROTTLE);
+ event.str_ = broker_name;
+ event.id_ = broker_id;
+ event.throttle_time_ = throttle_time_ms;
+
+ handle->event_cb_->event_cb(event);
+}
+
+
+int RdKafka::stats_cb_trampoline(rd_kafka_t *rk,
+ char *json,
+ size_t json_len,
+ void *opaque) {
+ RdKafka::HandleImpl *handle = static_cast<RdKafka::HandleImpl *>(opaque);
+
+ RdKafka::EventImpl event(RdKafka::Event::EVENT_STATS, RdKafka::ERR_NO_ERROR,
+ RdKafka::Event::EVENT_SEVERITY_INFO, NULL, json);
+
+ handle->event_cb_->event_cb(event);
+
+ return 0;
+}
+
+
+int RdKafka::socket_cb_trampoline(int domain,
+ int type,
+ int protocol,
+ void *opaque) {
+ RdKafka::HandleImpl *handle = static_cast<RdKafka::HandleImpl *>(opaque);
+
+ return handle->socket_cb_->socket_cb(domain, type, protocol);
+}
+
+int RdKafka::open_cb_trampoline(const char *pathname,
+ int flags,
+ mode_t mode,
+ void *opaque) {
+ RdKafka::HandleImpl *handle = static_cast<RdKafka::HandleImpl *>(opaque);
+
+ return handle->open_cb_->open_cb(pathname, flags, static_cast<int>(mode));
+}
+
+void RdKafka::oauthbearer_token_refresh_cb_trampoline(
+ rd_kafka_t *rk,
+ const char *oauthbearer_config,
+ void *opaque) {
+ RdKafka::HandleImpl *handle = static_cast<RdKafka::HandleImpl *>(opaque);
+
+ handle->oauthbearer_token_refresh_cb_->oauthbearer_token_refresh_cb(
+ handle, std::string(oauthbearer_config ? oauthbearer_config : ""));
+}
+
+
+int RdKafka::ssl_cert_verify_cb_trampoline(rd_kafka_t *rk,
+ const char *broker_name,
+ int32_t broker_id,
+ int *x509_error,
+ int depth,
+ const char *buf,
+ size_t size,
+ char *errstr,
+ size_t errstr_size,
+ void *opaque) {
+ RdKafka::HandleImpl *handle = static_cast<RdKafka::HandleImpl *>(opaque);
+ std::string errbuf;
+
+ bool res = 0 != handle->ssl_cert_verify_cb_->ssl_cert_verify_cb(
+ std::string(broker_name), broker_id, x509_error, depth,
+ buf, size, errbuf);
+
+ if (res)
+ return (int)res;
+
+ size_t errlen =
+ errbuf.size() > errstr_size - 1 ? errstr_size - 1 : errbuf.size();
+
+ memcpy(errstr, errbuf.c_str(), errlen);
+ if (errstr_size > 0)
+ errstr[errlen] = '\0';
+
+ return (int)res;
+}
+
+
+RdKafka::ErrorCode RdKafka::HandleImpl::metadata(bool all_topics,
+ const Topic *only_rkt,
+ Metadata **metadatap,
+ int timeout_ms) {
+ const rd_kafka_metadata_t *cmetadatap = NULL;
+
+ rd_kafka_topic_t *topic =
+ only_rkt ? static_cast<const TopicImpl *>(only_rkt)->rkt_ : NULL;
+
+ const rd_kafka_resp_err_t rc =
+ rd_kafka_metadata(rk_, all_topics, topic, &cmetadatap, timeout_ms);
+
+ *metadatap = (rc == RD_KAFKA_RESP_ERR_NO_ERROR)
+ ? new RdKafka::MetadataImpl(cmetadatap)
+ : NULL;
+
+ return static_cast<RdKafka::ErrorCode>(rc);
+}
+
+/**
+ * Convert a list of C partitions to C++ partitions
+ */
+static void c_parts_to_partitions(
+ const rd_kafka_topic_partition_list_t *c_parts,
+ std::vector<RdKafka::TopicPartition *> &partitions) {
+ partitions.resize(c_parts->cnt);
+ for (int i = 0; i < c_parts->cnt; i++)
+ partitions[i] = new RdKafka::TopicPartitionImpl(&c_parts->elems[i]);
+}
+
+static void free_partition_vector(std::vector<RdKafka::TopicPartition *> &v) {
+ for (unsigned int i = 0; i < v.size(); i++)
+ delete v[i];
+ v.clear();
+}
+
+void RdKafka::rebalance_cb_trampoline(
+ rd_kafka_t *rk,
+ rd_kafka_resp_err_t err,
+ rd_kafka_topic_partition_list_t *c_partitions,
+ void *opaque) {
+ RdKafka::HandleImpl *handle = static_cast<RdKafka::HandleImpl *>(opaque);
+ std::vector<RdKafka::TopicPartition *> partitions;
+
+ c_parts_to_partitions(c_partitions, partitions);
+
+ handle->rebalance_cb_->rebalance_cb(
+ dynamic_cast<RdKafka::KafkaConsumer *>(handle),
+ static_cast<RdKafka::ErrorCode>(err), partitions);
+
+ free_partition_vector(partitions);
+}
+
+
+void RdKafka::offset_commit_cb_trampoline0(
+ rd_kafka_t *rk,
+ rd_kafka_resp_err_t err,
+ rd_kafka_topic_partition_list_t *c_offsets,
+ void *opaque) {
+ OffsetCommitCb *cb = static_cast<RdKafka::OffsetCommitCb *>(opaque);
+ std::vector<RdKafka::TopicPartition *> offsets;
+
+ if (c_offsets)
+ c_parts_to_partitions(c_offsets, offsets);
+
+ cb->offset_commit_cb(static_cast<RdKafka::ErrorCode>(err), offsets);
+
+ free_partition_vector(offsets);
+}
+
+static void offset_commit_cb_trampoline(
+ rd_kafka_t *rk,
+ rd_kafka_resp_err_t err,
+ rd_kafka_topic_partition_list_t *c_offsets,
+ void *opaque) {
+ RdKafka::HandleImpl *handle = static_cast<RdKafka::HandleImpl *>(opaque);
+ RdKafka::offset_commit_cb_trampoline0(rk, err, c_offsets,
+ handle->offset_commit_cb_);
+}
+
+
+void RdKafka::HandleImpl::set_common_config(const RdKafka::ConfImpl *confimpl) {
+ rd_kafka_conf_set_opaque(confimpl->rk_conf_, this);
+
+ if (confimpl->event_cb_) {
+ rd_kafka_conf_set_log_cb(confimpl->rk_conf_, RdKafka::log_cb_trampoline);
+ rd_kafka_conf_set_error_cb(confimpl->rk_conf_,
+ RdKafka::error_cb_trampoline);
+ rd_kafka_conf_set_throttle_cb(confimpl->rk_conf_,
+ RdKafka::throttle_cb_trampoline);
+ rd_kafka_conf_set_stats_cb(confimpl->rk_conf_,
+ RdKafka::stats_cb_trampoline);
+ event_cb_ = confimpl->event_cb_;
+ }
+
+ if (confimpl->oauthbearer_token_refresh_cb_) {
+ rd_kafka_conf_set_oauthbearer_token_refresh_cb(
+ confimpl->rk_conf_, RdKafka::oauthbearer_token_refresh_cb_trampoline);
+ oauthbearer_token_refresh_cb_ = confimpl->oauthbearer_token_refresh_cb_;
+ }
+
+ if (confimpl->socket_cb_) {
+ rd_kafka_conf_set_socket_cb(confimpl->rk_conf_,
+ RdKafka::socket_cb_trampoline);
+ socket_cb_ = confimpl->socket_cb_;
+ }
+
+ if (confimpl->ssl_cert_verify_cb_) {
+ rd_kafka_conf_set_ssl_cert_verify_cb(
+ confimpl->rk_conf_, RdKafka::ssl_cert_verify_cb_trampoline);
+ ssl_cert_verify_cb_ = confimpl->ssl_cert_verify_cb_;
+ }
+
+ if (confimpl->open_cb_) {
+#ifndef _WIN32
+ rd_kafka_conf_set_open_cb(confimpl->rk_conf_, RdKafka::open_cb_trampoline);
+ open_cb_ = confimpl->open_cb_;
+#endif
+ }
+
+ if (confimpl->rebalance_cb_) {
+ rd_kafka_conf_set_rebalance_cb(confimpl->rk_conf_,
+ RdKafka::rebalance_cb_trampoline);
+ rebalance_cb_ = confimpl->rebalance_cb_;
+ }
+
+ if (confimpl->offset_commit_cb_) {
+ rd_kafka_conf_set_offset_commit_cb(confimpl->rk_conf_,
+ offset_commit_cb_trampoline);
+ offset_commit_cb_ = confimpl->offset_commit_cb_;
+ }
+
+ if (confimpl->consume_cb_) {
+ rd_kafka_conf_set_consume_cb(confimpl->rk_conf_,
+ RdKafka::consume_cb_trampoline);
+ consume_cb_ = confimpl->consume_cb_;
+ }
+}
+
+
+RdKafka::ErrorCode RdKafka::HandleImpl::pause(
+ std::vector<RdKafka::TopicPartition *> &partitions) {
+ rd_kafka_topic_partition_list_t *c_parts;
+ rd_kafka_resp_err_t err;
+
+ c_parts = partitions_to_c_parts(partitions);
+
+ err = rd_kafka_pause_partitions(rk_, c_parts);
+
+ if (!err)
+ update_partitions_from_c_parts(partitions, c_parts);
+
+ rd_kafka_topic_partition_list_destroy(c_parts);
+
+ return static_cast<RdKafka::ErrorCode>(err);
+}
+
+
+RdKafka::ErrorCode RdKafka::HandleImpl::resume(
+ std::vector<RdKafka::TopicPartition *> &partitions) {
+ rd_kafka_topic_partition_list_t *c_parts;
+ rd_kafka_resp_err_t err;
+
+ c_parts = partitions_to_c_parts(partitions);
+
+ err = rd_kafka_resume_partitions(rk_, c_parts);
+
+ if (!err)
+ update_partitions_from_c_parts(partitions, c_parts);
+
+ rd_kafka_topic_partition_list_destroy(c_parts);
+
+ return static_cast<RdKafka::ErrorCode>(err);
+}
+
+RdKafka::Queue *RdKafka::HandleImpl::get_partition_queue(
+ const TopicPartition *part) {
+ rd_kafka_queue_t *rkqu;
+ rkqu = rd_kafka_queue_get_partition(rk_, part->topic().c_str(),
+ part->partition());
+
+ if (rkqu == NULL)
+ return NULL;
+
+ return new QueueImpl(rkqu);
+}
+
+RdKafka::ErrorCode RdKafka::HandleImpl::set_log_queue(RdKafka::Queue *queue) {
+ rd_kafka_queue_t *rkqu = NULL;
+ if (queue) {
+ QueueImpl *queueimpl = dynamic_cast<QueueImpl *>(queue);
+ rkqu = queueimpl->queue_;
+ }
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_set_log_queue(rk_, rkqu));
+}
+
+namespace RdKafka {
+
+rd_kafka_topic_partition_list_t *partitions_to_c_parts(
+ const std::vector<RdKafka::TopicPartition *> &partitions) {
+ rd_kafka_topic_partition_list_t *c_parts;
+
+ c_parts = rd_kafka_topic_partition_list_new((int)partitions.size());
+
+ for (unsigned int i = 0; i < partitions.size(); i++) {
+ const RdKafka::TopicPartitionImpl *tpi =
+ dynamic_cast<const RdKafka::TopicPartitionImpl *>(partitions[i]);
+ rd_kafka_topic_partition_t *rktpar = rd_kafka_topic_partition_list_add(
+ c_parts, tpi->topic_.c_str(), tpi->partition_);
+ rktpar->offset = tpi->offset_;
+ if (tpi->leader_epoch_ != -1)
+ rd_kafka_topic_partition_set_leader_epoch(rktpar, tpi->leader_epoch_);
+ }
+
+ return c_parts;
+}
+
+
+/**
+ * @brief Update the application provided 'partitions' with info from 'c_parts'
+ */
+void update_partitions_from_c_parts(
+ std::vector<RdKafka::TopicPartition *> &partitions,
+ const rd_kafka_topic_partition_list_t *c_parts) {
+ for (int i = 0; i < c_parts->cnt; i++) {
+ rd_kafka_topic_partition_t *p = &c_parts->elems[i];
+
+ /* Find corresponding C++ entry */
+ for (unsigned int j = 0; j < partitions.size(); j++) {
+ RdKafka::TopicPartitionImpl *pp =
+ dynamic_cast<RdKafka::TopicPartitionImpl *>(partitions[j]);
+ if (!strcmp(p->topic, pp->topic_.c_str()) &&
+ p->partition == pp->partition_) {
+ pp->offset_ = p->offset;
+ pp->err_ = static_cast<RdKafka::ErrorCode>(p->err);
+ pp->leader_epoch_ = rd_kafka_topic_partition_get_leader_epoch(p);
+ }
+ }
+ }
+}
+
+} // namespace RdKafka
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/HeadersImpl.cpp b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/HeadersImpl.cpp
new file mode 100644
index 00000000..b567ef36
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/HeadersImpl.cpp
@@ -0,0 +1,48 @@
+/*
+ * librdkafka - Apache Kafka C/C++ library
+ *
+ * Copyright (c) 2014 Magnus Edenhill
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <iostream>
+#include <string>
+#include <list>
+#include <cerrno>
+
+#include "rdkafkacpp_int.h"
+
+RdKafka::Headers *RdKafka::Headers::create() {
+ return new RdKafka::HeadersImpl();
+}
+
+RdKafka::Headers *RdKafka::Headers::create(const std::vector<Header> &headers) {
+ if (headers.size() > 0)
+ return new RdKafka::HeadersImpl(headers);
+ else
+ return new RdKafka::HeadersImpl();
+}
+
+RdKafka::Headers::~Headers() {
+}
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/KafkaConsumerImpl.cpp b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/KafkaConsumerImpl.cpp
new file mode 100644
index 00000000..6f3b81c7
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/KafkaConsumerImpl.cpp
@@ -0,0 +1,296 @@
+/*
+ * librdkafka - Apache Kafka C/C++ library
+ *
+ * Copyright (c) 2015 Magnus Edenhill
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <string>
+#include <vector>
+
+#include "rdkafkacpp_int.h"
+
+RdKafka::KafkaConsumer::~KafkaConsumer() {
+}
+
+RdKafka::KafkaConsumer *RdKafka::KafkaConsumer::create(
+ const RdKafka::Conf *conf,
+ std::string &errstr) {
+ char errbuf[512];
+ const RdKafka::ConfImpl *confimpl =
+ dynamic_cast<const RdKafka::ConfImpl *>(conf);
+ RdKafka::KafkaConsumerImpl *rkc = new RdKafka::KafkaConsumerImpl();
+ rd_kafka_conf_t *rk_conf = NULL;
+ size_t grlen;
+
+ if (!confimpl || !confimpl->rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ delete rkc;
+ return NULL;
+ }
+
+ if (rd_kafka_conf_get(confimpl->rk_conf_, "group.id", NULL, &grlen) !=
+ RD_KAFKA_CONF_OK ||
+ grlen <= 1 /* terminating null only */) {
+ errstr = "\"group.id\" must be configured";
+ delete rkc;
+ return NULL;
+ }
+
+ rkc->set_common_config(confimpl);
+
+ rk_conf = rd_kafka_conf_dup(confimpl->rk_conf_);
+
+ rd_kafka_t *rk;
+ if (!(rk =
+ rd_kafka_new(RD_KAFKA_CONSUMER, rk_conf, errbuf, sizeof(errbuf)))) {
+ errstr = errbuf;
+ // rd_kafka_new() takes ownership only if succeeds
+ rd_kafka_conf_destroy(rk_conf);
+ delete rkc;
+ return NULL;
+ }
+
+ rkc->rk_ = rk;
+
+ /* Redirect handle queue to cgrp's queue to provide a single queue point */
+ rd_kafka_poll_set_consumer(rk);
+
+ return rkc;
+}
+
+
+
+RdKafka::ErrorCode RdKafka::KafkaConsumerImpl::subscribe(
+ const std::vector<std::string> &topics) {
+ rd_kafka_topic_partition_list_t *c_topics;
+ rd_kafka_resp_err_t err;
+
+ c_topics = rd_kafka_topic_partition_list_new((int)topics.size());
+
+ for (unsigned int i = 0; i < topics.size(); i++)
+ rd_kafka_topic_partition_list_add(c_topics, topics[i].c_str(),
+ RD_KAFKA_PARTITION_UA);
+
+ err = rd_kafka_subscribe(rk_, c_topics);
+
+ rd_kafka_topic_partition_list_destroy(c_topics);
+
+ return static_cast<RdKafka::ErrorCode>(err);
+}
+
+
+
+RdKafka::ErrorCode RdKafka::KafkaConsumerImpl::unsubscribe() {
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_unsubscribe(this->rk_));
+}
+
+RdKafka::Message *RdKafka::KafkaConsumerImpl::consume(int timeout_ms) {
+ rd_kafka_message_t *rkmessage;
+
+ rkmessage = rd_kafka_consumer_poll(this->rk_, timeout_ms);
+
+ if (!rkmessage)
+ return new RdKafka::MessageImpl(RD_KAFKA_CONSUMER, NULL,
+ RdKafka::ERR__TIMED_OUT);
+
+ return new RdKafka::MessageImpl(RD_KAFKA_CONSUMER, rkmessage);
+}
+
+
+
+RdKafka::ErrorCode RdKafka::KafkaConsumerImpl::assignment(
+ std::vector<RdKafka::TopicPartition *> &partitions) {
+ rd_kafka_topic_partition_list_t *c_parts;
+ rd_kafka_resp_err_t err;
+
+ if ((err = rd_kafka_assignment(rk_, &c_parts)))
+ return static_cast<RdKafka::ErrorCode>(err);
+
+ partitions.resize(c_parts->cnt);
+
+ for (int i = 0; i < c_parts->cnt; i++)
+ partitions[i] = new RdKafka::TopicPartitionImpl(&c_parts->elems[i]);
+
+ rd_kafka_topic_partition_list_destroy(c_parts);
+
+ return RdKafka::ERR_NO_ERROR;
+}
+
+
+
+bool RdKafka::KafkaConsumerImpl::assignment_lost() {
+ return rd_kafka_assignment_lost(rk_) ? true : false;
+}
+
+
+
+RdKafka::ErrorCode RdKafka::KafkaConsumerImpl::subscription(
+ std::vector<std::string> &topics) {
+ rd_kafka_topic_partition_list_t *c_topics;
+ rd_kafka_resp_err_t err;
+
+ if ((err = rd_kafka_subscription(rk_, &c_topics)))
+ return static_cast<RdKafka::ErrorCode>(err);
+
+ topics.resize(c_topics->cnt);
+ for (int i = 0; i < c_topics->cnt; i++)
+ topics[i] = std::string(c_topics->elems[i].topic);
+
+ rd_kafka_topic_partition_list_destroy(c_topics);
+
+ return RdKafka::ERR_NO_ERROR;
+}
+
+
+RdKafka::ErrorCode RdKafka::KafkaConsumerImpl::assign(
+ const std::vector<TopicPartition *> &partitions) {
+ rd_kafka_topic_partition_list_t *c_parts;
+ rd_kafka_resp_err_t err;
+
+ c_parts = partitions_to_c_parts(partitions);
+
+ err = rd_kafka_assign(rk_, c_parts);
+
+ rd_kafka_topic_partition_list_destroy(c_parts);
+ return static_cast<RdKafka::ErrorCode>(err);
+}
+
+
+RdKafka::ErrorCode RdKafka::KafkaConsumerImpl::unassign() {
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_assign(rk_, NULL));
+}
+
+
+RdKafka::Error *RdKafka::KafkaConsumerImpl::incremental_assign(
+ const std::vector<TopicPartition *> &partitions) {
+ rd_kafka_topic_partition_list_t *c_parts;
+ rd_kafka_error_t *c_error;
+
+ c_parts = partitions_to_c_parts(partitions);
+ c_error = rd_kafka_incremental_assign(rk_, c_parts);
+ rd_kafka_topic_partition_list_destroy(c_parts);
+
+ if (c_error)
+ return new ErrorImpl(c_error);
+
+ return NULL;
+}
+
+
+RdKafka::Error *RdKafka::KafkaConsumerImpl::incremental_unassign(
+ const std::vector<TopicPartition *> &partitions) {
+ rd_kafka_topic_partition_list_t *c_parts;
+ rd_kafka_error_t *c_error;
+
+ c_parts = partitions_to_c_parts(partitions);
+ c_error = rd_kafka_incremental_unassign(rk_, c_parts);
+ rd_kafka_topic_partition_list_destroy(c_parts);
+
+ if (c_error)
+ return new ErrorImpl(c_error);
+
+ return NULL;
+}
+
+
+RdKafka::ErrorCode RdKafka::KafkaConsumerImpl::committed(
+ std::vector<RdKafka::TopicPartition *> &partitions,
+ int timeout_ms) {
+ rd_kafka_topic_partition_list_t *c_parts;
+ rd_kafka_resp_err_t err;
+
+ c_parts = partitions_to_c_parts(partitions);
+
+ err = rd_kafka_committed(rk_, c_parts, timeout_ms);
+
+ if (!err) {
+ update_partitions_from_c_parts(partitions, c_parts);
+ }
+
+ rd_kafka_topic_partition_list_destroy(c_parts);
+
+ return static_cast<RdKafka::ErrorCode>(err);
+}
+
+
+RdKafka::ErrorCode RdKafka::KafkaConsumerImpl::position(
+ std::vector<RdKafka::TopicPartition *> &partitions) {
+ rd_kafka_topic_partition_list_t *c_parts;
+ rd_kafka_resp_err_t err;
+
+ c_parts = partitions_to_c_parts(partitions);
+
+ err = rd_kafka_position(rk_, c_parts);
+
+ if (!err) {
+ update_partitions_from_c_parts(partitions, c_parts);
+ }
+
+ rd_kafka_topic_partition_list_destroy(c_parts);
+
+ return static_cast<RdKafka::ErrorCode>(err);
+}
+
+
+RdKafka::ErrorCode RdKafka::KafkaConsumerImpl::seek(
+ const RdKafka::TopicPartition &partition,
+ int timeout_ms) {
+ const RdKafka::TopicPartitionImpl *p =
+ dynamic_cast<const RdKafka::TopicPartitionImpl *>(&partition);
+ rd_kafka_topic_t *rkt;
+
+ if (!(rkt = rd_kafka_topic_new(rk_, p->topic_.c_str(), NULL)))
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_last_error());
+
+ /* FIXME: Use a C API that takes a topic_partition_list_t instead */
+ RdKafka::ErrorCode err = static_cast<RdKafka::ErrorCode>(
+ rd_kafka_seek(rkt, p->partition_, p->offset_, timeout_ms));
+
+ rd_kafka_topic_destroy(rkt);
+
+ return err;
+}
+
+
+
+RdKafka::ErrorCode RdKafka::KafkaConsumerImpl::close() {
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_consumer_close(rk_));
+}
+
+
+RdKafka::Error *RdKafka::KafkaConsumerImpl::close(Queue *queue) {
+ QueueImpl *queueimpl = dynamic_cast<QueueImpl *>(queue);
+ rd_kafka_error_t *c_error;
+
+ c_error = rd_kafka_consumer_close_queue(rk_, queueimpl->queue_);
+ if (c_error)
+ return new ErrorImpl(c_error);
+
+ return NULL;
+}
+
+
+RdKafka::ConsumerGroupMetadata::~ConsumerGroupMetadata() {
+}
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/Makefile b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/Makefile
new file mode 100644
index 00000000..78ecb31f
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/Makefile
@@ -0,0 +1,55 @@
+PKGNAME= librdkafka
+LIBNAME= librdkafka++
+LIBVER= 1
+
+CXXSRCS= RdKafka.cpp ConfImpl.cpp HandleImpl.cpp \
+ ConsumerImpl.cpp ProducerImpl.cpp KafkaConsumerImpl.cpp \
+ TopicImpl.cpp TopicPartitionImpl.cpp MessageImpl.cpp \
+ HeadersImpl.cpp QueueImpl.cpp MetadataImpl.cpp
+
+HDRS= rdkafkacpp.h
+
+OBJS= $(CXXSRCS:%.cpp=%.o)
+
+
+
+all: lib check
+
+# No linker script/symbol hiding for C++ library
+DISABLE_LDS=y
+
+MKL_NO_SELFCONTAINED_STATIC_LIB=y
+include ../mklove/Makefile.base
+
+# Use C++ compiler as linker rather than the default C compiler
+CC_LD=$(CXX)
+
+# OSX and Cygwin requires linking required libraries
+ifeq ($(_UNAME_S),Darwin)
+ FWD_LINKING_REQ=y
+endif
+ifeq ($(_UNAME_S),AIX)
+ FWD_LINKING_REQ=y
+endif
+ifeq ($(shell uname -o 2>/dev/null),Cygwin)
+ FWD_LINKING_REQ=y
+endif
+
+# Ignore previously defined library dependencies for the C library,
+# we'll get those dependencies through the C library linkage.
+LIBS := -L../src -lrdkafka
+MKL_PKGCONFIG_REQUIRES_PRIVATE := rdkafka
+MKL_PKGCONFIG_REQUIRES := rdkafka
+
+CHECK_FILES+= $(LIBFILENAME) $(LIBNAME).a
+
+
+file-check: lib
+check: file-check
+
+install: lib-install
+uninstall: lib-uninstall
+
+clean: lib-clean
+
+-include $(DEPS)
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/MessageImpl.cpp b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/MessageImpl.cpp
new file mode 100644
index 00000000..c6d83150
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/MessageImpl.cpp
@@ -0,0 +1,38 @@
+/*
+ * librdkafka - Apache Kafka C/C++ library
+ *
+ * Copyright (c) 2014 Magnus Edenhill
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <iostream>
+#include <string>
+#include <list>
+#include <cerrno>
+
+#include "rdkafkacpp_int.h"
+
+
+RdKafka::Message::~Message() {
+}
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/MetadataImpl.cpp b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/MetadataImpl.cpp
new file mode 100644
index 00000000..62cbf904
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/MetadataImpl.cpp
@@ -0,0 +1,170 @@
+/*
+ * librdkafka - Apache Kafka C/C++ library
+ *
+ * Copyright (c) 2014 Magnus Edenhill
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include "rdkafkacpp_int.h"
+
+using namespace RdKafka;
+
+BrokerMetadata::~BrokerMetadata() {
+}
+PartitionMetadata::~PartitionMetadata() {
+}
+TopicMetadata::~TopicMetadata() {
+}
+Metadata::~Metadata() {
+}
+
+
+/**
+ * Metadata: Broker information handler implementation
+ */
+class BrokerMetadataImpl : public BrokerMetadata {
+ public:
+ BrokerMetadataImpl(const rd_kafka_metadata_broker_t *broker_metadata) :
+ broker_metadata_(broker_metadata), host_(broker_metadata->host) {
+ }
+
+ int32_t id() const {
+ return broker_metadata_->id;
+ }
+
+ std::string host() const {
+ return host_;
+ }
+ int port() const {
+ return broker_metadata_->port;
+ }
+
+ virtual ~BrokerMetadataImpl() {
+ }
+
+ private:
+ const rd_kafka_metadata_broker_t *broker_metadata_;
+ const std::string host_;
+};
+
+/**
+ * Metadata: Partition information handler
+ */
+class PartitionMetadataImpl : public PartitionMetadata {
+ public:
+ // @TODO too much memory copy? maybe we should create a new vector class that
+ // read directly from C arrays?
+ // @TODO use auto_ptr?
+ PartitionMetadataImpl(
+ const rd_kafka_metadata_partition_t *partition_metadata) :
+ partition_metadata_(partition_metadata) {
+ replicas_.reserve(partition_metadata->replica_cnt);
+ for (int i = 0; i < partition_metadata->replica_cnt; ++i)
+ replicas_.push_back(partition_metadata->replicas[i]);
+
+ isrs_.reserve(partition_metadata->isr_cnt);
+ for (int i = 0; i < partition_metadata->isr_cnt; ++i)
+ isrs_.push_back(partition_metadata->isrs[i]);
+ }
+
+ int32_t id() const {
+ return partition_metadata_->id;
+ }
+ int32_t leader() const {
+ return partition_metadata_->leader;
+ }
+ ErrorCode err() const {
+ return static_cast<ErrorCode>(partition_metadata_->err);
+ }
+
+ const std::vector<int32_t> *replicas() const {
+ return &replicas_;
+ }
+ const std::vector<int32_t> *isrs() const {
+ return &isrs_;
+ }
+
+ ~PartitionMetadataImpl() {
+ }
+
+ private:
+ const rd_kafka_metadata_partition_t *partition_metadata_;
+ std::vector<int32_t> replicas_, isrs_;
+};
+
+/**
+ * Metadata: Topic information handler
+ */
+class TopicMetadataImpl : public TopicMetadata {
+ public:
+ TopicMetadataImpl(const rd_kafka_metadata_topic_t *topic_metadata) :
+ topic_metadata_(topic_metadata), topic_(topic_metadata->topic) {
+ partitions_.reserve(topic_metadata->partition_cnt);
+ for (int i = 0; i < topic_metadata->partition_cnt; ++i)
+ partitions_.push_back(
+ new PartitionMetadataImpl(&topic_metadata->partitions[i]));
+ }
+
+ ~TopicMetadataImpl() {
+ for (size_t i = 0; i < partitions_.size(); ++i)
+ delete partitions_[i];
+ }
+
+ std::string topic() const {
+ return topic_;
+ }
+ const std::vector<const PartitionMetadata *> *partitions() const {
+ return &partitions_;
+ }
+ ErrorCode err() const {
+ return static_cast<ErrorCode>(topic_metadata_->err);
+ }
+
+ private:
+ const rd_kafka_metadata_topic_t *topic_metadata_;
+ const std::string topic_;
+ std::vector<const PartitionMetadata *> partitions_;
+};
+
+MetadataImpl::MetadataImpl(const rd_kafka_metadata_t *metadata) :
+ metadata_(metadata) {
+ brokers_.reserve(metadata->broker_cnt);
+ for (int i = 0; i < metadata->broker_cnt; ++i)
+ brokers_.push_back(new BrokerMetadataImpl(&metadata->brokers[i]));
+
+ topics_.reserve(metadata->topic_cnt);
+ for (int i = 0; i < metadata->topic_cnt; ++i)
+ topics_.push_back(new TopicMetadataImpl(&metadata->topics[i]));
+}
+
+MetadataImpl::~MetadataImpl() {
+ for (size_t i = 0; i < brokers_.size(); ++i)
+ delete brokers_[i];
+ for (size_t i = 0; i < topics_.size(); ++i)
+ delete topics_[i];
+
+
+ if (metadata_)
+ rd_kafka_metadata_destroy(metadata_);
+}
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/ProducerImpl.cpp b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/ProducerImpl.cpp
new file mode 100644
index 00000000..8300dfb3
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/ProducerImpl.cpp
@@ -0,0 +1,197 @@
+/*
+ * librdkafka - Apache Kafka C/C++ library
+ *
+ * Copyright (c) 2014 Magnus Edenhill
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <iostream>
+#include <string>
+#include <list>
+#include <cerrno>
+
+#include "rdkafkacpp_int.h"
+
+
+RdKafka::Producer::~Producer() {
+}
+
+static void dr_msg_cb_trampoline(rd_kafka_t *rk,
+ const rd_kafka_message_t *rkmessage,
+ void *opaque) {
+ RdKafka::HandleImpl *handle = static_cast<RdKafka::HandleImpl *>(opaque);
+ RdKafka::MessageImpl message(RD_KAFKA_PRODUCER, NULL,
+ (rd_kafka_message_t *)rkmessage, false);
+ handle->dr_cb_->dr_cb(message);
+}
+
+
+
+RdKafka::Producer *RdKafka::Producer::create(const RdKafka::Conf *conf,
+ std::string &errstr) {
+ char errbuf[512];
+ const RdKafka::ConfImpl *confimpl =
+ dynamic_cast<const RdKafka::ConfImpl *>(conf);
+ RdKafka::ProducerImpl *rkp = new RdKafka::ProducerImpl();
+ rd_kafka_conf_t *rk_conf = NULL;
+
+ if (confimpl) {
+ if (!confimpl->rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ delete rkp;
+ return NULL;
+ }
+
+ rkp->set_common_config(confimpl);
+
+ rk_conf = rd_kafka_conf_dup(confimpl->rk_conf_);
+
+ if (confimpl->dr_cb_) {
+ rd_kafka_conf_set_dr_msg_cb(rk_conf, dr_msg_cb_trampoline);
+ rkp->dr_cb_ = confimpl->dr_cb_;
+ }
+ }
+
+
+ rd_kafka_t *rk;
+ if (!(rk =
+ rd_kafka_new(RD_KAFKA_PRODUCER, rk_conf, errbuf, sizeof(errbuf)))) {
+ errstr = errbuf;
+ // rd_kafka_new() takes ownership only if succeeds
+ if (rk_conf)
+ rd_kafka_conf_destroy(rk_conf);
+ delete rkp;
+ return NULL;
+ }
+
+ rkp->rk_ = rk;
+
+ return rkp;
+}
+
+
+RdKafka::ErrorCode RdKafka::ProducerImpl::produce(RdKafka::Topic *topic,
+ int32_t partition,
+ int msgflags,
+ void *payload,
+ size_t len,
+ const std::string *key,
+ void *msg_opaque) {
+ RdKafka::TopicImpl *topicimpl = dynamic_cast<RdKafka::TopicImpl *>(topic);
+
+ if (rd_kafka_produce(topicimpl->rkt_, partition, msgflags, payload, len,
+ key ? key->c_str() : NULL, key ? key->size() : 0,
+ msg_opaque) == -1)
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_last_error());
+
+ return RdKafka::ERR_NO_ERROR;
+}
+
+
+RdKafka::ErrorCode RdKafka::ProducerImpl::produce(RdKafka::Topic *topic,
+ int32_t partition,
+ int msgflags,
+ void *payload,
+ size_t len,
+ const void *key,
+ size_t key_len,
+ void *msg_opaque) {
+ RdKafka::TopicImpl *topicimpl = dynamic_cast<RdKafka::TopicImpl *>(topic);
+
+ if (rd_kafka_produce(topicimpl->rkt_, partition, msgflags, payload, len, key,
+ key_len, msg_opaque) == -1)
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_last_error());
+
+ return RdKafka::ERR_NO_ERROR;
+}
+
+
+RdKafka::ErrorCode RdKafka::ProducerImpl::produce(
+ RdKafka::Topic *topic,
+ int32_t partition,
+ const std::vector<char> *payload,
+ const std::vector<char> *key,
+ void *msg_opaque) {
+ RdKafka::TopicImpl *topicimpl = dynamic_cast<RdKafka::TopicImpl *>(topic);
+
+ if (rd_kafka_produce(topicimpl->rkt_, partition, RD_KAFKA_MSG_F_COPY,
+ payload ? (void *)&(*payload)[0] : NULL,
+ payload ? payload->size() : 0, key ? &(*key)[0] : NULL,
+ key ? key->size() : 0, msg_opaque) == -1)
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_last_error());
+
+ return RdKafka::ERR_NO_ERROR;
+}
+
+RdKafka::ErrorCode RdKafka::ProducerImpl::produce(const std::string topic_name,
+ int32_t partition,
+ int msgflags,
+ void *payload,
+ size_t len,
+ const void *key,
+ size_t key_len,
+ int64_t timestamp,
+ void *msg_opaque) {
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_producev(
+ rk_, RD_KAFKA_V_TOPIC(topic_name.c_str()),
+ RD_KAFKA_V_PARTITION(partition), RD_KAFKA_V_MSGFLAGS(msgflags),
+ RD_KAFKA_V_VALUE(payload, len), RD_KAFKA_V_KEY(key, key_len),
+ RD_KAFKA_V_TIMESTAMP(timestamp), RD_KAFKA_V_OPAQUE(msg_opaque),
+ RD_KAFKA_V_END));
+}
+
+RdKafka::ErrorCode RdKafka::ProducerImpl::produce(const std::string topic_name,
+ int32_t partition,
+ int msgflags,
+ void *payload,
+ size_t len,
+ const void *key,
+ size_t key_len,
+ int64_t timestamp,
+ RdKafka::Headers *headers,
+ void *msg_opaque) {
+ rd_kafka_headers_t *hdrs = NULL;
+ RdKafka::HeadersImpl *headersimpl = NULL;
+ rd_kafka_resp_err_t err;
+
+ if (headers) {
+ headersimpl = static_cast<RdKafka::HeadersImpl *>(headers);
+ hdrs = headersimpl->c_ptr();
+ }
+
+ err = rd_kafka_producev(
+ rk_, RD_KAFKA_V_TOPIC(topic_name.c_str()),
+ RD_KAFKA_V_PARTITION(partition), RD_KAFKA_V_MSGFLAGS(msgflags),
+ RD_KAFKA_V_VALUE(payload, len), RD_KAFKA_V_KEY(key, key_len),
+ RD_KAFKA_V_TIMESTAMP(timestamp), RD_KAFKA_V_OPAQUE(msg_opaque),
+ RD_KAFKA_V_HEADERS(hdrs), RD_KAFKA_V_END);
+
+ if (!err && headersimpl) {
+ /* A successful producev() call will destroy the C headers. */
+ headersimpl->c_headers_destroyed();
+ delete headers;
+ }
+
+ return static_cast<RdKafka::ErrorCode>(err);
+}
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/QueueImpl.cpp b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/QueueImpl.cpp
new file mode 100644
index 00000000..19ebce9d
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/QueueImpl.cpp
@@ -0,0 +1,70 @@
+/*
+ * librdkafka - Apache Kafka C/C++ library
+ *
+ * Copyright (c) 2014 Magnus Edenhill
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <cerrno>
+
+#include "rdkafkacpp_int.h"
+
+RdKafka::Queue::~Queue() {
+}
+
+RdKafka::Queue *RdKafka::Queue::create(Handle *base) {
+ return new RdKafka::QueueImpl(
+ rd_kafka_queue_new(dynamic_cast<HandleImpl *>(base)->rk_));
+}
+
+RdKafka::ErrorCode RdKafka::QueueImpl::forward(Queue *queue) {
+ if (!queue) {
+ rd_kafka_queue_forward(queue_, NULL);
+ } else {
+ QueueImpl *queueimpl = dynamic_cast<QueueImpl *>(queue);
+ rd_kafka_queue_forward(queue_, queueimpl->queue_);
+ }
+ return RdKafka::ERR_NO_ERROR;
+}
+
+RdKafka::Message *RdKafka::QueueImpl::consume(int timeout_ms) {
+ rd_kafka_message_t *rkmessage;
+ rkmessage = rd_kafka_consume_queue(queue_, timeout_ms);
+
+ if (!rkmessage)
+ return new RdKafka::MessageImpl(RD_KAFKA_CONSUMER, NULL,
+ RdKafka::ERR__TIMED_OUT);
+
+ return new RdKafka::MessageImpl(RD_KAFKA_CONSUMER, rkmessage);
+}
+
+int RdKafka::QueueImpl::poll(int timeout_ms) {
+ return rd_kafka_queue_poll_callback(queue_, timeout_ms);
+}
+
+void RdKafka::QueueImpl::io_event_enable(int fd,
+ const void *payload,
+ size_t size) {
+ rd_kafka_queue_io_event_enable(queue_, fd, payload, size);
+}
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/README.md b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/README.md
new file mode 100644
index 00000000..a4845894
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/README.md
@@ -0,0 +1,16 @@
+librdkafka C++ interface
+========================
+
+**See rdkafkacpp.h for the public C++ API**
+
+
+
+Maintainer notes for the C++ interface:
+
+ * The public C++ interface (rdkafkacpp.h) does not include the
+ public C interface (rdkafka.h) in any way, this means that all
+ constants, flags, etc, must be kept in sync manually between the two
+ header files.
+ A regression test should be implemented that checks this is true.
+
+ * The public C++ interface is provided using pure virtual abstract classes.
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/RdKafka.cpp b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/RdKafka.cpp
new file mode 100644
index 00000000..b6cb33c2
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/RdKafka.cpp
@@ -0,0 +1,59 @@
+/*
+ * librdkafka - Apache Kafka C/C++ library
+ *
+ * Copyright (c) 2014 Magnus Edenhill
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <string>
+
+#include "rdkafkacpp_int.h"
+
+int RdKafka::version() {
+ return rd_kafka_version();
+}
+
+std::string RdKafka::version_str() {
+ return std::string(rd_kafka_version_str());
+}
+
+std::string RdKafka::get_debug_contexts() {
+ return std::string(RD_KAFKA_DEBUG_CONTEXTS);
+}
+
+std::string RdKafka::err2str(RdKafka::ErrorCode err) {
+ return std::string(rd_kafka_err2str(static_cast<rd_kafka_resp_err_t>(err)));
+}
+
+int RdKafka::wait_destroyed(int timeout_ms) {
+ return rd_kafka_wait_destroyed(timeout_ms);
+}
+
+void *RdKafka::mem_malloc(size_t size) {
+ return rd_kafka_mem_malloc(NULL, size);
+}
+
+void RdKafka::mem_free(void *ptr) {
+ rd_kafka_mem_free(NULL, ptr);
+}
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/TopicImpl.cpp b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/TopicImpl.cpp
new file mode 100644
index 00000000..bf9734df
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/TopicImpl.cpp
@@ -0,0 +1,124 @@
+/*
+ * librdkafka - Apache Kafka C/C++ library
+ *
+ * Copyright (c) 2014 Magnus Edenhill
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <iostream>
+#include <string>
+#include <list>
+#include <cerrno>
+
+#include "rdkafkacpp_int.h"
+
+const int32_t RdKafka::Topic::PARTITION_UA = RD_KAFKA_PARTITION_UA;
+
+const int64_t RdKafka::Topic::OFFSET_BEGINNING = RD_KAFKA_OFFSET_BEGINNING;
+
+const int64_t RdKafka::Topic::OFFSET_END = RD_KAFKA_OFFSET_END;
+
+const int64_t RdKafka::Topic::OFFSET_STORED = RD_KAFKA_OFFSET_STORED;
+
+const int64_t RdKafka::Topic::OFFSET_INVALID = RD_KAFKA_OFFSET_INVALID;
+
+RdKafka::Topic::~Topic() {
+}
+
+static int32_t partitioner_cb_trampoline(const rd_kafka_topic_t *rkt,
+ const void *keydata,
+ size_t keylen,
+ int32_t partition_cnt,
+ void *rkt_opaque,
+ void *msg_opaque) {
+ RdKafka::TopicImpl *topicimpl = static_cast<RdKafka::TopicImpl *>(rkt_opaque);
+ std::string key(static_cast<const char *>(keydata), keylen);
+ return topicimpl->partitioner_cb_->partitioner_cb(topicimpl, &key,
+ partition_cnt, msg_opaque);
+}
+
+static int32_t partitioner_kp_cb_trampoline(const rd_kafka_topic_t *rkt,
+ const void *keydata,
+ size_t keylen,
+ int32_t partition_cnt,
+ void *rkt_opaque,
+ void *msg_opaque) {
+ RdKafka::TopicImpl *topicimpl = static_cast<RdKafka::TopicImpl *>(rkt_opaque);
+ return topicimpl->partitioner_kp_cb_->partitioner_cb(
+ topicimpl, keydata, keylen, partition_cnt, msg_opaque);
+}
+
+
+
+RdKafka::Topic *RdKafka::Topic::create(Handle *base,
+ const std::string &topic_str,
+ const Conf *conf,
+ std::string &errstr) {
+ const RdKafka::ConfImpl *confimpl =
+ static_cast<const RdKafka::ConfImpl *>(conf);
+ rd_kafka_topic_t *rkt;
+ rd_kafka_topic_conf_t *rkt_conf;
+ rd_kafka_t *rk = dynamic_cast<HandleImpl *>(base)->rk_;
+
+ RdKafka::TopicImpl *topic = new RdKafka::TopicImpl();
+
+ if (!confimpl) {
+ /* Reuse default topic config, but we need our own copy to
+ * set the topic opaque. */
+ rkt_conf = rd_kafka_default_topic_conf_dup(rk);
+ } else {
+ /* Make a copy of conf struct to allow Conf reuse. */
+ rkt_conf = rd_kafka_topic_conf_dup(confimpl->rkt_conf_);
+ }
+
+ /* Set topic opaque to the topic so that we can reach our topic object
+ * from whatever callbacks get registered.
+ * The application itself will not need these opaques since their
+ * callbacks are class based. */
+ rd_kafka_topic_conf_set_opaque(rkt_conf, static_cast<void *>(topic));
+
+ if (confimpl) {
+ if (confimpl->partitioner_cb_) {
+ rd_kafka_topic_conf_set_partitioner_cb(rkt_conf,
+ partitioner_cb_trampoline);
+ topic->partitioner_cb_ = confimpl->partitioner_cb_;
+ } else if (confimpl->partitioner_kp_cb_) {
+ rd_kafka_topic_conf_set_partitioner_cb(rkt_conf,
+ partitioner_kp_cb_trampoline);
+ topic->partitioner_kp_cb_ = confimpl->partitioner_kp_cb_;
+ }
+ }
+
+
+ if (!(rkt = rd_kafka_topic_new(rk, topic_str.c_str(), rkt_conf))) {
+ errstr = rd_kafka_err2str(rd_kafka_last_error());
+ delete topic;
+ rd_kafka_topic_conf_destroy(rkt_conf);
+ return NULL;
+ }
+
+ topic->rkt_ = rkt;
+
+ return topic;
+}
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/TopicPartitionImpl.cpp b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/TopicPartitionImpl.cpp
new file mode 100644
index 00000000..90ef820b
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/TopicPartitionImpl.cpp
@@ -0,0 +1,57 @@
+/*
+ * librdkafka - Apache Kafka C/C++ library
+ *
+ * Copyright (c) 2015 Magnus Edenhill
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <iostream>
+#include <string>
+#include <vector>
+
+#include "rdkafkacpp_int.h"
+
+RdKafka::TopicPartition::~TopicPartition() {
+}
+
+RdKafka::TopicPartition *RdKafka::TopicPartition::create(
+ const std::string &topic,
+ int partition) {
+ return new TopicPartitionImpl(topic, partition);
+}
+
+RdKafka::TopicPartition *RdKafka::TopicPartition::create(
+ const std::string &topic,
+ int partition,
+ int64_t offset) {
+ return new TopicPartitionImpl(topic, partition, offset);
+}
+
+void RdKafka::TopicPartition::destroy(
+ std::vector<TopicPartition *> &partitions) {
+ for (std::vector<TopicPartition *>::iterator it = partitions.begin();
+ it != partitions.end(); ++it)
+ delete (*it);
+ partitions.clear();
+}
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/rdkafkacpp.h b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/rdkafkacpp.h
new file mode 100644
index 00000000..1df1043c
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/rdkafkacpp.h
@@ -0,0 +1,3764 @@
+/*
+ * librdkafka - Apache Kafka C/C++ library
+ *
+ * Copyright (c) 2014-2022 Magnus Edenhill
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _RDKAFKACPP_H_
+#define _RDKAFKACPP_H_
+
+/**
+ * @file rdkafkacpp.h
+ * @brief Apache Kafka C/C++ consumer and producer client library.
+ *
+ * rdkafkacpp.h contains the public C++ API for librdkafka.
+ * The API is documented in this file as comments prefixing the class,
+ * function, type, enum, define, etc.
+ * For more information, see the C interface in rdkafka.h and read the
+ * manual in INTRODUCTION.md.
+ * The C++ interface is STD C++ '03 compliant and adheres to the
+ * Google C++ Style Guide.
+
+ * @sa For the C interface see rdkafka.h
+ *
+ * @tableofcontents
+ */
+
+/**@cond NO_DOC*/
+#include <string>
+#include <list>
+#include <vector>
+#include <cstdlib>
+#include <cstring>
+#include <stdint.h>
+#include <sys/types.h>
+
+#ifdef _WIN32
+#ifndef ssize_t
+#ifndef _BASETSD_H_
+#include <basetsd.h>
+#endif
+#ifndef _SSIZE_T_DEFINED
+#define _SSIZE_T_DEFINED
+typedef SSIZE_T ssize_t;
+#endif
+#endif
+#undef RD_EXPORT
+#ifdef LIBRDKAFKA_STATICLIB
+#define RD_EXPORT
+#else
+#ifdef LIBRDKAFKACPP_EXPORTS
+#define RD_EXPORT __declspec(dllexport)
+#else
+#define RD_EXPORT __declspec(dllimport)
+#endif
+#endif
+#else
+#define RD_EXPORT
+#endif
+
+/**@endcond*/
+
+extern "C" {
+/* Forward declarations */
+struct rd_kafka_s;
+struct rd_kafka_topic_s;
+struct rd_kafka_message_s;
+struct rd_kafka_conf_s;
+struct rd_kafka_topic_conf_s;
+}
+
+namespace RdKafka {
+
+/**
+ * @name Miscellaneous APIs
+ * @{
+ */
+
+/**
+ * @brief librdkafka version
+ *
+ * Interpreted as hex \c MM.mm.rr.xx:
+ * - MM = Major
+ * - mm = minor
+ * - rr = revision
+ * - xx = pre-release id (0xff is the final release)
+ *
+ * E.g.: \c 0x000801ff = 0.8.1
+ *
+ * @remark This value should only be used during compile time,
+ * for runtime checks of version use RdKafka::version()
+ */
+#define RD_KAFKA_VERSION 0x020100ff
+
+/**
+ * @brief Returns the librdkafka version as integer.
+ *
+ * @sa See RD_KAFKA_VERSION for how to parse the integer format.
+ */
+RD_EXPORT
+int version();
+
+/**
+ * @brief Returns the librdkafka version as string.
+ */
+RD_EXPORT
+std::string version_str();
+
+/**
+ * @brief Returns a CSV list of the supported debug contexts
+ * for use with Conf::Set("debug", ..).
+ */
+RD_EXPORT
+std::string get_debug_contexts();
+
+/**
+ * @brief Wait for all rd_kafka_t objects to be destroyed.
+ *
+ * @returns 0 if all kafka objects are now destroyed, or -1 if the
+ * timeout was reached.
+ * Since RdKafka handle deletion is an asynch operation the
+ * \p wait_destroyed() function can be used for applications where
+ * a clean shutdown is required.
+ */
+RD_EXPORT
+int wait_destroyed(int timeout_ms);
+
+/**
+ * @brief Allocate memory using the same allocator librdkafka uses.
+ *
+ * This is typically an abstraction for the malloc(3) call and makes sure
+ * the application can use the same memory allocator as librdkafka for
+ * allocating pointers that are used by librdkafka.
+ *
+ * @remark Memory allocated by mem_malloc() must be freed using
+ * mem_free().
+ */
+RD_EXPORT
+void *mem_malloc(size_t size);
+
+/**
+ * @brief Free pointer returned by librdkafka
+ *
+ * This is typically an abstraction for the free(3) call and makes sure
+ * the application can use the same memory allocator as librdkafka for
+ * freeing pointers returned by librdkafka.
+ *
+ * In standard setups it is usually not necessary to use this interface
+ * rather than the free(3) function.
+ *
+ * @remark mem_free() must only be used for pointers returned by APIs
+ * that explicitly mention using this function for freeing.
+ */
+RD_EXPORT
+void mem_free(void *ptr);
+
+/**@}*/
+
+
+
+/**
+ * @name Constants, errors, types
+ * @{
+ *
+ *
+ */
+
+/**
+ * @brief Error codes.
+ *
+ * The negative error codes delimited by two underscores
+ * (\c _ERR__..) denotes errors internal to librdkafka and are
+ * displayed as \c \"Local: \<error string..\>\", while the error codes
+ * delimited by a single underscore (\c ERR_..) denote broker
+ * errors and are displayed as \c \"Broker: \<error string..\>\".
+ *
+ * @sa Use RdKafka::err2str() to translate an error code a human readable string
+ */
+enum ErrorCode {
+ /* Internal errors to rdkafka: */
+ /** Begin internal error codes */
+ ERR__BEGIN = -200,
+ /** Received message is incorrect */
+ ERR__BAD_MSG = -199,
+ /** Bad/unknown compression */
+ ERR__BAD_COMPRESSION = -198,
+ /** Broker is going away */
+ ERR__DESTROY = -197,
+ /** Generic failure */
+ ERR__FAIL = -196,
+ /** Broker transport failure */
+ ERR__TRANSPORT = -195,
+ /** Critical system resource */
+ ERR__CRIT_SYS_RESOURCE = -194,
+ /** Failed to resolve broker */
+ ERR__RESOLVE = -193,
+ /** Produced message timed out*/
+ ERR__MSG_TIMED_OUT = -192,
+ /** Reached the end of the topic+partition queue on
+ * the broker. Not really an error.
+ * This event is disabled by default,
+ * see the `enable.partition.eof` configuration property. */
+ ERR__PARTITION_EOF = -191,
+ /** Permanent: Partition does not exist in cluster. */
+ ERR__UNKNOWN_PARTITION = -190,
+ /** File or filesystem error */
+ ERR__FS = -189,
+ /** Permanent: Topic does not exist in cluster. */
+ ERR__UNKNOWN_TOPIC = -188,
+ /** All broker connections are down. */
+ ERR__ALL_BROKERS_DOWN = -187,
+ /** Invalid argument, or invalid configuration */
+ ERR__INVALID_ARG = -186,
+ /** Operation timed out */
+ ERR__TIMED_OUT = -185,
+ /** Queue is full */
+ ERR__QUEUE_FULL = -184,
+ /** ISR count < required.acks */
+ ERR__ISR_INSUFF = -183,
+ /** Broker node update */
+ ERR__NODE_UPDATE = -182,
+ /** SSL error */
+ ERR__SSL = -181,
+ /** Waiting for coordinator to become available. */
+ ERR__WAIT_COORD = -180,
+ /** Unknown client group */
+ ERR__UNKNOWN_GROUP = -179,
+ /** Operation in progress */
+ ERR__IN_PROGRESS = -178,
+ /** Previous operation in progress, wait for it to finish. */
+ ERR__PREV_IN_PROGRESS = -177,
+ /** This operation would interfere with an existing subscription */
+ ERR__EXISTING_SUBSCRIPTION = -176,
+ /** Assigned partitions (rebalance_cb) */
+ ERR__ASSIGN_PARTITIONS = -175,
+ /** Revoked partitions (rebalance_cb) */
+ ERR__REVOKE_PARTITIONS = -174,
+ /** Conflicting use */
+ ERR__CONFLICT = -173,
+ /** Wrong state */
+ ERR__STATE = -172,
+ /** Unknown protocol */
+ ERR__UNKNOWN_PROTOCOL = -171,
+ /** Not implemented */
+ ERR__NOT_IMPLEMENTED = -170,
+ /** Authentication failure*/
+ ERR__AUTHENTICATION = -169,
+ /** No stored offset */
+ ERR__NO_OFFSET = -168,
+ /** Outdated */
+ ERR__OUTDATED = -167,
+ /** Timed out in queue */
+ ERR__TIMED_OUT_QUEUE = -166,
+ /** Feature not supported by broker */
+ ERR__UNSUPPORTED_FEATURE = -165,
+ /** Awaiting cache update */
+ ERR__WAIT_CACHE = -164,
+ /** Operation interrupted */
+ ERR__INTR = -163,
+ /** Key serialization error */
+ ERR__KEY_SERIALIZATION = -162,
+ /** Value serialization error */
+ ERR__VALUE_SERIALIZATION = -161,
+ /** Key deserialization error */
+ ERR__KEY_DESERIALIZATION = -160,
+ /** Value deserialization error */
+ ERR__VALUE_DESERIALIZATION = -159,
+ /** Partial response */
+ ERR__PARTIAL = -158,
+ /** Modification attempted on read-only object */
+ ERR__READ_ONLY = -157,
+ /** No such entry / item not found */
+ ERR__NOENT = -156,
+ /** Read underflow */
+ ERR__UNDERFLOW = -155,
+ /** Invalid type */
+ ERR__INVALID_TYPE = -154,
+ /** Retry operation */
+ ERR__RETRY = -153,
+ /** Purged in queue */
+ ERR__PURGE_QUEUE = -152,
+ /** Purged in flight */
+ ERR__PURGE_INFLIGHT = -151,
+ /** Fatal error: see RdKafka::Handle::fatal_error() */
+ ERR__FATAL = -150,
+ /** Inconsistent state */
+ ERR__INCONSISTENT = -149,
+ /** Gap-less ordering would not be guaranteed if proceeding */
+ ERR__GAPLESS_GUARANTEE = -148,
+ /** Maximum poll interval exceeded */
+ ERR__MAX_POLL_EXCEEDED = -147,
+ /** Unknown broker */
+ ERR__UNKNOWN_BROKER = -146,
+ /** Functionality not configured */
+ ERR__NOT_CONFIGURED = -145,
+ /** Instance has been fenced */
+ ERR__FENCED = -144,
+ /** Application generated error */
+ ERR__APPLICATION = -143,
+ /** Assignment lost */
+ ERR__ASSIGNMENT_LOST = -142,
+ /** No operation performed */
+ ERR__NOOP = -141,
+ /** No offset to automatically reset to */
+ ERR__AUTO_OFFSET_RESET = -140,
+ /** Partition log truncation detected */
+ ERR__LOG_TRUNCATION = -139,
+
+ /** End internal error codes */
+ ERR__END = -100,
+
+ /* Kafka broker errors: */
+ /** Unknown broker error */
+ ERR_UNKNOWN = -1,
+ /** Success */
+ ERR_NO_ERROR = 0,
+ /** Offset out of range */
+ ERR_OFFSET_OUT_OF_RANGE = 1,
+ /** Invalid message */
+ ERR_INVALID_MSG = 2,
+ /** Unknown topic or partition */
+ ERR_UNKNOWN_TOPIC_OR_PART = 3,
+ /** Invalid message size */
+ ERR_INVALID_MSG_SIZE = 4,
+ /** Leader not available */
+ ERR_LEADER_NOT_AVAILABLE = 5,
+ /** Not leader for partition */
+ ERR_NOT_LEADER_FOR_PARTITION = 6,
+ /** Request timed out */
+ ERR_REQUEST_TIMED_OUT = 7,
+ /** Broker not available */
+ ERR_BROKER_NOT_AVAILABLE = 8,
+ /** Replica not available */
+ ERR_REPLICA_NOT_AVAILABLE = 9,
+ /** Message size too large */
+ ERR_MSG_SIZE_TOO_LARGE = 10,
+ /** StaleControllerEpochCode */
+ ERR_STALE_CTRL_EPOCH = 11,
+ /** Offset metadata string too large */
+ ERR_OFFSET_METADATA_TOO_LARGE = 12,
+ /** Broker disconnected before response received */
+ ERR_NETWORK_EXCEPTION = 13,
+ /** Coordinator load in progress */
+ ERR_COORDINATOR_LOAD_IN_PROGRESS = 14,
+/** Group coordinator load in progress */
+#define ERR_GROUP_LOAD_IN_PROGRESS ERR_COORDINATOR_LOAD_IN_PROGRESS
+ /** Coordinator not available */
+ ERR_COORDINATOR_NOT_AVAILABLE = 15,
+/** Group coordinator not available */
+#define ERR_GROUP_COORDINATOR_NOT_AVAILABLE ERR_COORDINATOR_NOT_AVAILABLE
+ /** Not coordinator */
+ ERR_NOT_COORDINATOR = 16,
+/** Not coordinator for group */
+#define ERR_NOT_COORDINATOR_FOR_GROUP ERR_NOT_COORDINATOR
+ /** Invalid topic */
+ ERR_TOPIC_EXCEPTION = 17,
+ /** Message batch larger than configured server segment size */
+ ERR_RECORD_LIST_TOO_LARGE = 18,
+ /** Not enough in-sync replicas */
+ ERR_NOT_ENOUGH_REPLICAS = 19,
+ /** Message(s) written to insufficient number of in-sync replicas */
+ ERR_NOT_ENOUGH_REPLICAS_AFTER_APPEND = 20,
+ /** Invalid required acks value */
+ ERR_INVALID_REQUIRED_ACKS = 21,
+ /** Specified group generation id is not valid */
+ ERR_ILLEGAL_GENERATION = 22,
+ /** Inconsistent group protocol */
+ ERR_INCONSISTENT_GROUP_PROTOCOL = 23,
+ /** Invalid group.id */
+ ERR_INVALID_GROUP_ID = 24,
+ /** Unknown member */
+ ERR_UNKNOWN_MEMBER_ID = 25,
+ /** Invalid session timeout */
+ ERR_INVALID_SESSION_TIMEOUT = 26,
+ /** Group rebalance in progress */
+ ERR_REBALANCE_IN_PROGRESS = 27,
+ /** Commit offset data size is not valid */
+ ERR_INVALID_COMMIT_OFFSET_SIZE = 28,
+ /** Topic authorization failed */
+ ERR_TOPIC_AUTHORIZATION_FAILED = 29,
+ /** Group authorization failed */
+ ERR_GROUP_AUTHORIZATION_FAILED = 30,
+ /** Cluster authorization failed */
+ ERR_CLUSTER_AUTHORIZATION_FAILED = 31,
+ /** Invalid timestamp */
+ ERR_INVALID_TIMESTAMP = 32,
+ /** Unsupported SASL mechanism */
+ ERR_UNSUPPORTED_SASL_MECHANISM = 33,
+ /** Illegal SASL state */
+ ERR_ILLEGAL_SASL_STATE = 34,
+ /** Unuspported version */
+ ERR_UNSUPPORTED_VERSION = 35,
+ /** Topic already exists */
+ ERR_TOPIC_ALREADY_EXISTS = 36,
+ /** Invalid number of partitions */
+ ERR_INVALID_PARTITIONS = 37,
+ /** Invalid replication factor */
+ ERR_INVALID_REPLICATION_FACTOR = 38,
+ /** Invalid replica assignment */
+ ERR_INVALID_REPLICA_ASSIGNMENT = 39,
+ /** Invalid config */
+ ERR_INVALID_CONFIG = 40,
+ /** Not controller for cluster */
+ ERR_NOT_CONTROLLER = 41,
+ /** Invalid request */
+ ERR_INVALID_REQUEST = 42,
+ /** Message format on broker does not support request */
+ ERR_UNSUPPORTED_FOR_MESSAGE_FORMAT = 43,
+ /** Policy violation */
+ ERR_POLICY_VIOLATION = 44,
+ /** Broker received an out of order sequence number */
+ ERR_OUT_OF_ORDER_SEQUENCE_NUMBER = 45,
+ /** Broker received a duplicate sequence number */
+ ERR_DUPLICATE_SEQUENCE_NUMBER = 46,
+ /** Producer attempted an operation with an old epoch */
+ ERR_INVALID_PRODUCER_EPOCH = 47,
+ /** Producer attempted a transactional operation in an invalid state */
+ ERR_INVALID_TXN_STATE = 48,
+ /** Producer attempted to use a producer id which is not
+ * currently assigned to its transactional id */
+ ERR_INVALID_PRODUCER_ID_MAPPING = 49,
+ /** Transaction timeout is larger than the maximum
+ * value allowed by the broker's max.transaction.timeout.ms */
+ ERR_INVALID_TRANSACTION_TIMEOUT = 50,
+ /** Producer attempted to update a transaction while another
+ * concurrent operation on the same transaction was ongoing */
+ ERR_CONCURRENT_TRANSACTIONS = 51,
+ /** Indicates that the transaction coordinator sending a
+ * WriteTxnMarker is no longer the current coordinator for a
+ * given producer */
+ ERR_TRANSACTION_COORDINATOR_FENCED = 52,
+ /** Transactional Id authorization failed */
+ ERR_TRANSACTIONAL_ID_AUTHORIZATION_FAILED = 53,
+ /** Security features are disabled */
+ ERR_SECURITY_DISABLED = 54,
+ /** Operation not attempted */
+ ERR_OPERATION_NOT_ATTEMPTED = 55,
+ /** Disk error when trying to access log file on the disk */
+ ERR_KAFKA_STORAGE_ERROR = 56,
+ /** The user-specified log directory is not found in the broker config */
+ ERR_LOG_DIR_NOT_FOUND = 57,
+ /** SASL Authentication failed */
+ ERR_SASL_AUTHENTICATION_FAILED = 58,
+ /** Unknown Producer Id */
+ ERR_UNKNOWN_PRODUCER_ID = 59,
+ /** Partition reassignment is in progress */
+ ERR_REASSIGNMENT_IN_PROGRESS = 60,
+ /** Delegation Token feature is not enabled */
+ ERR_DELEGATION_TOKEN_AUTH_DISABLED = 61,
+ /** Delegation Token is not found on server */
+ ERR_DELEGATION_TOKEN_NOT_FOUND = 62,
+ /** Specified Principal is not valid Owner/Renewer */
+ ERR_DELEGATION_TOKEN_OWNER_MISMATCH = 63,
+ /** Delegation Token requests are not allowed on this connection */
+ ERR_DELEGATION_TOKEN_REQUEST_NOT_ALLOWED = 64,
+ /** Delegation Token authorization failed */
+ ERR_DELEGATION_TOKEN_AUTHORIZATION_FAILED = 65,
+ /** Delegation Token is expired */
+ ERR_DELEGATION_TOKEN_EXPIRED = 66,
+ /** Supplied principalType is not supported */
+ ERR_INVALID_PRINCIPAL_TYPE = 67,
+ /** The group is not empty */
+ ERR_NON_EMPTY_GROUP = 68,
+ /** The group id does not exist */
+ ERR_GROUP_ID_NOT_FOUND = 69,
+ /** The fetch session ID was not found */
+ ERR_FETCH_SESSION_ID_NOT_FOUND = 70,
+ /** The fetch session epoch is invalid */
+ ERR_INVALID_FETCH_SESSION_EPOCH = 71,
+ /** No matching listener */
+ ERR_LISTENER_NOT_FOUND = 72,
+ /** Topic deletion is disabled */
+ ERR_TOPIC_DELETION_DISABLED = 73,
+ /** Leader epoch is older than broker epoch */
+ ERR_FENCED_LEADER_EPOCH = 74,
+ /** Leader epoch is newer than broker epoch */
+ ERR_UNKNOWN_LEADER_EPOCH = 75,
+ /** Unsupported compression type */
+ ERR_UNSUPPORTED_COMPRESSION_TYPE = 76,
+ /** Broker epoch has changed */
+ ERR_STALE_BROKER_EPOCH = 77,
+ /** Leader high watermark is not caught up */
+ ERR_OFFSET_NOT_AVAILABLE = 78,
+ /** Group member needs a valid member ID */
+ ERR_MEMBER_ID_REQUIRED = 79,
+ /** Preferred leader was not available */
+ ERR_PREFERRED_LEADER_NOT_AVAILABLE = 80,
+ /** Consumer group has reached maximum size */
+ ERR_GROUP_MAX_SIZE_REACHED = 81,
+ /** Static consumer fenced by other consumer with same
+ * group.instance.id. */
+ ERR_FENCED_INSTANCE_ID = 82,
+ /** Eligible partition leaders are not available */
+ ERR_ELIGIBLE_LEADERS_NOT_AVAILABLE = 83,
+ /** Leader election not needed for topic partition */
+ ERR_ELECTION_NOT_NEEDED = 84,
+ /** No partition reassignment is in progress */
+ ERR_NO_REASSIGNMENT_IN_PROGRESS = 85,
+ /** Deleting offsets of a topic while the consumer group is
+ * subscribed to it */
+ ERR_GROUP_SUBSCRIBED_TO_TOPIC = 86,
+ /** Broker failed to validate record */
+ ERR_INVALID_RECORD = 87,
+ /** There are unstable offsets that need to be cleared */
+ ERR_UNSTABLE_OFFSET_COMMIT = 88,
+ /** Throttling quota has been exceeded */
+ ERR_THROTTLING_QUOTA_EXCEEDED = 89,
+ /** There is a newer producer with the same transactionalId
+ * which fences the current one */
+ ERR_PRODUCER_FENCED = 90,
+ /** Request illegally referred to resource that does not exist */
+ ERR_RESOURCE_NOT_FOUND = 91,
+ /** Request illegally referred to the same resource twice */
+ ERR_DUPLICATE_RESOURCE = 92,
+ /** Requested credential would not meet criteria for acceptability */
+ ERR_UNACCEPTABLE_CREDENTIAL = 93,
+ /** Indicates that the either the sender or recipient of a
+ * voter-only request is not one of the expected voters */
+ ERR_INCONSISTENT_VOTER_SET = 94,
+ /** Invalid update version */
+ ERR_INVALID_UPDATE_VERSION = 95,
+ /** Unable to update finalized features due to server error */
+ ERR_FEATURE_UPDATE_FAILED = 96,
+ /** Request principal deserialization failed during forwarding */
+ ERR_PRINCIPAL_DESERIALIZATION_FAILURE = 97
+};
+
+
+/**
+ * @brief Returns a human readable representation of a kafka error.
+ */
+RD_EXPORT
+std::string err2str(RdKafka::ErrorCode err);
+
+
+
+/**
+ * @enum CertificateType
+ * @brief SSL certificate types
+ */
+enum CertificateType {
+ CERT_PUBLIC_KEY, /**< Client's public key */
+ CERT_PRIVATE_KEY, /**< Client's private key */
+ CERT_CA, /**< CA certificate */
+ CERT__CNT
+};
+
+/**
+ * @enum CertificateEncoding
+ * @brief SSL certificate encoding
+ */
+enum CertificateEncoding {
+ CERT_ENC_PKCS12, /**< PKCS#12 */
+ CERT_ENC_DER, /**< DER / binary X.509 ASN1 */
+ CERT_ENC_PEM, /**< PEM */
+ CERT_ENC__CNT
+};
+
+/**@} */
+
+
+
+/**@cond NO_DOC*/
+/* Forward declarations */
+class Handle;
+class Producer;
+class Message;
+class Headers;
+class Queue;
+class Event;
+class Topic;
+class TopicPartition;
+class Metadata;
+class KafkaConsumer;
+/**@endcond*/
+
+
+/**
+ * @name Error class
+ * @{
+ *
+ */
+
+/**
+ * @brief The Error class is used as a return value from APIs to propagate
+ * an error. The error consists of an error code which is to be used
+ * programatically, an error string for showing to the user,
+ * and various error flags that can be used programmatically to decide
+ * how to handle the error; e.g., should the operation be retried,
+ * was it a fatal error, etc.
+ *
+ * Error objects must be deleted explicitly to free its resources.
+ */
+class RD_EXPORT Error {
+ public:
+ /**
+ * @brief Create error object.
+ */
+ static Error *create(ErrorCode code, const std::string *errstr);
+
+ virtual ~Error() {
+ }
+
+ /*
+ * Error accessor methods
+ */
+
+ /**
+ * @returns the error code, e.g., RdKafka::ERR_UNKNOWN_MEMBER_ID.
+ */
+ virtual ErrorCode code() const = 0;
+
+ /**
+ * @returns the error code name, e.g, "ERR_UNKNOWN_MEMBER_ID".
+ */
+ virtual std::string name() const = 0;
+
+ /**
+ * @returns a human readable error string.
+ */
+ virtual std::string str() const = 0;
+
+ /**
+ * @returns true if the error is a fatal error, indicating that the client
+ * instance is no longer usable, else false.
+ */
+ virtual bool is_fatal() const = 0;
+
+ /**
+ * @returns true if the operation may be retried, else false.
+ */
+ virtual bool is_retriable() const = 0;
+
+ /**
+ * @returns true if the error is an abortable transaction error in which case
+ * the application must call RdKafka::Producer::abort_transaction()
+ * and start a new transaction with
+ * RdKafka::Producer::begin_transaction() if it wishes to proceed
+ * with transactions.
+ * Else returns false.
+ *
+ * @remark The return value of this method is only valid for errors returned
+ * by the transactional API.
+ */
+ virtual bool txn_requires_abort() const = 0;
+};
+
+/**@}*/
+
+
+/**
+ * @name Callback classes
+ * @{
+ *
+ *
+ * librdkafka uses (optional) callbacks to propagate information and
+ * delegate decisions to the application logic.
+ *
+ * An application must call RdKafka::poll() at regular intervals to
+ * serve queued callbacks.
+ */
+
+
+/**
+ * @brief Delivery Report callback class
+ *
+ * The delivery report callback will be called once for each message
+ * accepted by RdKafka::Producer::produce() (et.al) with
+ * RdKafka::Message::err() set to indicate the result of the produce request.
+ *
+ * The callback is called when a message is succesfully produced or
+ * if librdkafka encountered a permanent failure, or the retry counter for
+ * temporary errors has been exhausted.
+ *
+ * An application must call RdKafka::poll() at regular intervals to
+ * serve queued delivery report callbacks.
+
+ */
+class RD_EXPORT DeliveryReportCb {
+ public:
+ /**
+ * @brief Delivery report callback.
+ */
+ virtual void dr_cb(Message &message) = 0;
+
+ virtual ~DeliveryReportCb() {
+ }
+};
+
+
+/**
+ * @brief SASL/OAUTHBEARER token refresh callback class
+ *
+ * The SASL/OAUTHBEARER token refresh callback is triggered via RdKafka::poll()
+ * whenever OAUTHBEARER is the SASL mechanism and a token needs to be retrieved,
+ * typically based on the configuration defined in \c sasl.oauthbearer.config.
+ *
+ * The \c oauthbearer_config argument is the value of the
+ * \c sasl.oauthbearer.config configuration property.
+ *
+ * The callback should invoke RdKafka::Handle::oauthbearer_set_token() or
+ * RdKafka::Handle::oauthbearer_set_token_failure() to indicate success or
+ * failure, respectively.
+ *
+ * The refresh operation is eventable and may be received when an event
+ * callback handler is set with an event type of
+ * \c RdKafka::Event::EVENT_OAUTHBEARER_TOKEN_REFRESH.
+ *
+ * Note that before any SASL/OAUTHBEARER broker connection can succeed the
+ * application must call RdKafka::Handle::oauthbearer_set_token() once -- either
+ * directly or, more typically, by invoking RdKafka::poll() -- in order to
+ * cause retrieval of an initial token to occur.
+ *
+ * An application must call RdKafka::poll() at regular intervals to
+ * serve queued SASL/OAUTHBEARER token refresh callbacks (when
+ * OAUTHBEARER is the SASL mechanism).
+ */
+class RD_EXPORT OAuthBearerTokenRefreshCb {
+ public:
+ /**
+ * @brief SASL/OAUTHBEARER token refresh callback class.
+ *
+ * @param handle The RdKafka::Handle which requires a refreshed token.
+ * @param oauthbearer_config The value of the
+ * \p sasl.oauthbearer.config configuration property for \p handle.
+ */
+ virtual void oauthbearer_token_refresh_cb(
+ RdKafka::Handle *handle,
+ const std::string &oauthbearer_config) = 0;
+
+ virtual ~OAuthBearerTokenRefreshCb() {
+ }
+};
+
+
+/**
+ * @brief Partitioner callback class
+ *
+ * Generic partitioner callback class for implementing custom partitioners.
+ *
+ * @sa RdKafka::Conf::set() \c "partitioner_cb"
+ */
+class RD_EXPORT PartitionerCb {
+ public:
+ /**
+ * @brief Partitioner callback
+ *
+ * Return the partition to use for \p key in \p topic.
+ *
+ * The \p msg_opaque is the same \p msg_opaque provided in the
+ * RdKafka::Producer::produce() call.
+ *
+ * @remark \p key may be NULL or the empty.
+ *
+ * @returns Must return a value between 0 and \p partition_cnt
+ * (non-inclusive). May return RD_KAFKA_PARTITION_UA (-1) if partitioning
+ * failed.
+ *
+ * @sa The callback may use RdKafka::Topic::partition_available() to check
+ * if a partition has an active leader broker.
+ */
+ virtual int32_t partitioner_cb(const Topic *topic,
+ const std::string *key,
+ int32_t partition_cnt,
+ void *msg_opaque) = 0;
+
+ virtual ~PartitionerCb() {
+ }
+};
+
+/**
+ * @brief Variant partitioner with key pointer
+ *
+ */
+class PartitionerKeyPointerCb {
+ public:
+ /**
+ * @brief Variant partitioner callback that gets \p key as pointer and length
+ * instead of as a const std::string *.
+ *
+ * @remark \p key may be NULL or have \p key_len 0.
+ *
+ * @sa See RdKafka::PartitionerCb::partitioner_cb() for exact semantics
+ */
+ virtual int32_t partitioner_cb(const Topic *topic,
+ const void *key,
+ size_t key_len,
+ int32_t partition_cnt,
+ void *msg_opaque) = 0;
+
+ virtual ~PartitionerKeyPointerCb() {
+ }
+};
+
+
+
+/**
+ * @brief Event callback class
+ *
+ * Events are a generic interface for propagating errors, statistics, logs, etc
+ * from librdkafka to the application.
+ *
+ * @sa RdKafka::Event
+ */
+class RD_EXPORT EventCb {
+ public:
+ /**
+ * @brief Event callback
+ *
+ * @sa RdKafka::Event
+ */
+ virtual void event_cb(Event &event) = 0;
+
+ virtual ~EventCb() {
+ }
+};
+
+
+/**
+ * @brief Event object class as passed to the EventCb callback.
+ */
+class RD_EXPORT Event {
+ public:
+ /** @brief Event type */
+ enum Type {
+ EVENT_ERROR, /**< Event is an error condition */
+ EVENT_STATS, /**< Event is a statistics JSON document */
+ EVENT_LOG, /**< Event is a log message */
+ EVENT_THROTTLE /**< Event is a throttle level signaling from the broker */
+ };
+
+ /** @brief EVENT_LOG severities (conforms to syslog(3) severities) */
+ enum Severity {
+ EVENT_SEVERITY_EMERG = 0,
+ EVENT_SEVERITY_ALERT = 1,
+ EVENT_SEVERITY_CRITICAL = 2,
+ EVENT_SEVERITY_ERROR = 3,
+ EVENT_SEVERITY_WARNING = 4,
+ EVENT_SEVERITY_NOTICE = 5,
+ EVENT_SEVERITY_INFO = 6,
+ EVENT_SEVERITY_DEBUG = 7
+ };
+
+ virtual ~Event() {
+ }
+
+ /*
+ * Event Accessor methods
+ */
+
+ /**
+ * @returns The event type
+ * @remark Applies to all event types
+ */
+ virtual Type type() const = 0;
+
+ /**
+ * @returns Event error, if any.
+ * @remark Applies to all event types except THROTTLE
+ */
+ virtual ErrorCode err() const = 0;
+
+ /**
+ * @returns Log severity level.
+ * @remark Applies to LOG event type.
+ */
+ virtual Severity severity() const = 0;
+
+ /**
+ * @returns Log facility string.
+ * @remark Applies to LOG event type.
+ */
+ virtual std::string fac() const = 0;
+
+ /**
+ * @returns Log message string.
+ *
+ * \c EVENT_LOG: Log message string.
+ * \c EVENT_STATS: JSON object (as string).
+ *
+ * @remark Applies to LOG event type.
+ */
+ virtual std::string str() const = 0;
+
+ /**
+ * @returns Throttle time in milliseconds.
+ * @remark Applies to THROTTLE event type.
+ */
+ virtual int throttle_time() const = 0;
+
+ /**
+ * @returns Throttling broker's name.
+ * @remark Applies to THROTTLE event type.
+ */
+ virtual std::string broker_name() const = 0;
+
+ /**
+ * @returns Throttling broker's id.
+ * @remark Applies to THROTTLE event type.
+ */
+ virtual int broker_id() const = 0;
+
+
+ /**
+ * @returns true if this is a fatal error.
+ * @remark Applies to ERROR event type.
+ * @sa RdKafka::Handle::fatal_error()
+ */
+ virtual bool fatal() const = 0;
+};
+
+
+
+/**
+ * @brief Consume callback class
+ */
+class RD_EXPORT ConsumeCb {
+ public:
+ /**
+ * @brief The consume callback is used with
+ * RdKafka::Consumer::consume_callback()
+ * methods and will be called for each consumed \p message.
+ *
+ * The callback interface is optional but provides increased performance.
+ */
+ virtual void consume_cb(Message &message, void *opaque) = 0;
+
+ virtual ~ConsumeCb() {
+ }
+};
+
+
+/**
+ * @brief \b KafkaConsumer: Rebalance callback class
+ */
+class RD_EXPORT RebalanceCb {
+ public:
+ /**
+ * @brief Group rebalance callback for use with RdKafka::KafkaConsumer
+ *
+ * Registering a \p rebalance_cb turns off librdkafka's automatic
+ * partition assignment/revocation and instead delegates that responsibility
+ * to the application's \p rebalance_cb.
+ *
+ * The rebalance callback is responsible for updating librdkafka's
+ * assignment set based on the two events: RdKafka::ERR__ASSIGN_PARTITIONS
+ * and RdKafka::ERR__REVOKE_PARTITIONS but should also be able to handle
+ * arbitrary rebalancing failures where \p err is neither of those.
+ * @remark In this latter case (arbitrary error), the application must
+ * call unassign() to synchronize state.
+ *
+ * For eager/non-cooperative `partition.assignment.strategy` assignors,
+ * such as `range` and `roundrobin`, the application must use
+ * assign assign() to set and unassign() to clear the entire assignment.
+ * For the cooperative assignors, such as `cooperative-sticky`, the
+ * application must use incremental_assign() for ERR__ASSIGN_PARTITIONS and
+ * incremental_unassign() for ERR__REVOKE_PARTITIONS.
+ *
+ * Without a rebalance callback this is done automatically by librdkafka
+ * but registering a rebalance callback gives the application flexibility
+ * in performing other operations along with the assinging/revocation,
+ * such as fetching offsets from an alternate location (on assign)
+ * or manually committing offsets (on revoke).
+ *
+ * @sa RdKafka::KafkaConsumer::assign()
+ * @sa RdKafka::KafkaConsumer::incremental_assign()
+ * @sa RdKafka::KafkaConsumer::incremental_unassign()
+ * @sa RdKafka::KafkaConsumer::assignment_lost()
+ * @sa RdKafka::KafkaConsumer::rebalance_protocol()
+ *
+ * The following example show's the application's responsibilities:
+ * @code
+ * class MyRebalanceCb : public RdKafka::RebalanceCb {
+ * public:
+ * void rebalance_cb (RdKafka::KafkaConsumer *consumer,
+ * RdKafka::ErrorCode err,
+ * std::vector<RdKafka::TopicPartition*> &partitions) {
+ * if (err == RdKafka::ERR__ASSIGN_PARTITIONS) {
+ * // application may load offets from arbitrary external
+ * // storage here and update \p partitions
+ * if (consumer->rebalance_protocol() == "COOPERATIVE")
+ * consumer->incremental_assign(partitions);
+ * else
+ * consumer->assign(partitions);
+ *
+ * } else if (err == RdKafka::ERR__REVOKE_PARTITIONS) {
+ * // Application may commit offsets manually here
+ * // if auto.commit.enable=false
+ * if (consumer->rebalance_protocol() == "COOPERATIVE")
+ * consumer->incremental_unassign(partitions);
+ * else
+ * consumer->unassign();
+ *
+ * } else {
+ * std::cerr << "Rebalancing error: " <<
+ * RdKafka::err2str(err) << std::endl;
+ * consumer->unassign();
+ * }
+ * }
+ * }
+ * @endcode
+ *
+ * @remark The above example lacks error handling for assign calls, see
+ * the examples/ directory.
+ */
+ virtual void rebalance_cb(RdKafka::KafkaConsumer *consumer,
+ RdKafka::ErrorCode err,
+ std::vector<TopicPartition *> &partitions) = 0;
+
+ virtual ~RebalanceCb() {
+ }
+};
+
+
+/**
+ * @brief Offset Commit callback class
+ */
+class RD_EXPORT OffsetCommitCb {
+ public:
+ /**
+ * @brief Set offset commit callback for use with consumer groups
+ *
+ * The results of automatic or manual offset commits will be scheduled
+ * for this callback and is served by RdKafka::KafkaConsumer::consume().
+ *
+ * If no partitions had valid offsets to commit this callback will be called
+ * with \p err == ERR__NO_OFFSET which is not to be considered an error.
+ *
+ * The \p offsets list contains per-partition information:
+ * - \c topic The topic committed
+ * - \c partition The partition committed
+ * - \c offset: Committed offset (attempted)
+ * - \c err: Commit error
+ */
+ virtual void offset_commit_cb(RdKafka::ErrorCode err,
+ std::vector<TopicPartition *> &offsets) = 0;
+
+ virtual ~OffsetCommitCb() {
+ }
+};
+
+
+
+/**
+ * @brief SSL broker certificate verification class.
+ *
+ * @remark Class instance must outlive the RdKafka client instance.
+ */
+class RD_EXPORT SslCertificateVerifyCb {
+ public:
+ /**
+ * @brief SSL broker certificate verification callback.
+ *
+ * The verification callback is triggered from internal librdkafka threads
+ * upon connecting to a broker. On each connection attempt the callback
+ * will be called for each certificate in the broker's certificate chain,
+ * starting at the root certification, as long as the application callback
+ * returns 1 (valid certificate).
+ *
+ * \p broker_name and \p broker_id correspond to the broker the connection
+ * is being made to.
+ * The \c x509_error argument indicates if OpenSSL's verification of
+ * the certificate succeed (0) or failed (an OpenSSL error code).
+ * The application may set the SSL context error code by returning 0
+ * from the verify callback and providing a non-zero SSL context error code
+ * in \p x509_error.
+ * If the verify callback sets \p x509_error to 0, returns 1, and the
+ * original \p x509_error was non-zero, the error on the SSL context will
+ * be cleared.
+ * \p x509_error is always a valid pointer to an int.
+ *
+ * \p depth is the depth of the current certificate in the chain, starting
+ * at the root certificate.
+ *
+ * The certificate itself is passed in binary DER format in \p buf of
+ * size \p size.
+ *
+ * The callback must 1 if verification succeeds, or 0 if verification fails
+ * and write a human-readable error message
+ * to \p errstr.
+ *
+ * @warning This callback will be called from internal librdkafka threads.
+ *
+ * @remark See <openssl/x509_vfy.h> in the OpenSSL source distribution
+ * for a list of \p x509_error codes.
+ */
+ virtual bool ssl_cert_verify_cb(const std::string &broker_name,
+ int32_t broker_id,
+ int *x509_error,
+ int depth,
+ const char *buf,
+ size_t size,
+ std::string &errstr) = 0;
+
+ virtual ~SslCertificateVerifyCb() {
+ }
+};
+
+
+/**
+ * @brief \b Portability: SocketCb callback class
+ *
+ */
+class RD_EXPORT SocketCb {
+ public:
+ /**
+ * @brief Socket callback
+ *
+ * The socket callback is responsible for opening a socket
+ * according to the supplied \p domain, \p type and \p protocol.
+ * The socket shall be created with \c CLOEXEC set in a racefree fashion, if
+ * possible.
+ *
+ * It is typically not required to register an alternative socket
+ * implementation
+ *
+ * @returns The socket file descriptor or -1 on error (\c errno must be set)
+ */
+ virtual int socket_cb(int domain, int type, int protocol) = 0;
+
+ virtual ~SocketCb() {
+ }
+};
+
+
+/**
+ * @brief \b Portability: OpenCb callback class
+ *
+ */
+class RD_EXPORT OpenCb {
+ public:
+ /**
+ * @brief Open callback
+ * The open callback is responsible for opening the file specified by
+ * \p pathname, using \p flags and \p mode.
+ * The file shall be opened with \c CLOEXEC set in a racefree fashion, if
+ * possible.
+ *
+ * It is typically not required to register an alternative open implementation
+ *
+ * @remark Not currently available on native Win32
+ */
+ virtual int open_cb(const std::string &path, int flags, int mode) = 0;
+
+ virtual ~OpenCb() {
+ }
+};
+
+
+/**@}*/
+
+
+
+/**
+ * @name Configuration interface
+ * @{
+ *
+ */
+
+/**
+ * @brief Configuration interface
+ *
+ * Holds either global or topic configuration that are passed to
+ * RdKafka::Consumer::create(), RdKafka::Producer::create(),
+ * RdKafka::KafkaConsumer::create(), etc.
+ *
+ * @sa CONFIGURATION.md for the full list of supported properties.
+ */
+class RD_EXPORT Conf {
+ public:
+ /**
+ * @brief Configuration object type
+ */
+ enum ConfType {
+ CONF_GLOBAL, /**< Global configuration */
+ CONF_TOPIC /**< Topic specific configuration */
+ };
+
+ /**
+ * @brief RdKafka::Conf::Set() result code
+ */
+ enum ConfResult {
+ CONF_UNKNOWN = -2, /**< Unknown configuration property */
+ CONF_INVALID = -1, /**< Invalid configuration value */
+ CONF_OK = 0 /**< Configuration property was succesfully set */
+ };
+
+
+ /**
+ * @brief Create configuration object
+ */
+ static Conf *create(ConfType type);
+
+ virtual ~Conf() {
+ }
+
+ /**
+ * @brief Set configuration property \p name to value \p value.
+ *
+ * Fallthrough:
+ * Topic-level configuration properties may be set using this interface
+ * in which case they are applied on the \c default_topic_conf.
+ * If no \c default_topic_conf has been set one will be created.
+ * Any sub-sequent set("default_topic_conf", ..) calls will
+ * replace the current default topic configuration.
+
+ * @returns CONF_OK on success, else writes a human readable error
+ * description to \p errstr on error.
+ */
+ virtual Conf::ConfResult set(const std::string &name,
+ const std::string &value,
+ std::string &errstr) = 0;
+
+ /** @brief Use with \p name = \c \"dr_cb\" */
+ virtual Conf::ConfResult set(const std::string &name,
+ DeliveryReportCb *dr_cb,
+ std::string &errstr) = 0;
+
+ /** @brief Use with \p name = \c \"oauthbearer_token_refresh_cb\" */
+ virtual Conf::ConfResult set(
+ const std::string &name,
+ OAuthBearerTokenRefreshCb *oauthbearer_token_refresh_cb,
+ std::string &errstr) = 0;
+
+ /** @brief Use with \p name = \c \"event_cb\" */
+ virtual Conf::ConfResult set(const std::string &name,
+ EventCb *event_cb,
+ std::string &errstr) = 0;
+
+ /** @brief Use with \p name = \c \"default_topic_conf\"
+ *
+ * Sets the default topic configuration to use for for automatically
+ * subscribed topics.
+ *
+ * @sa RdKafka::KafkaConsumer::subscribe()
+ */
+ virtual Conf::ConfResult set(const std::string &name,
+ const Conf *topic_conf,
+ std::string &errstr) = 0;
+
+ /** @brief Use with \p name = \c \"partitioner_cb\" */
+ virtual Conf::ConfResult set(const std::string &name,
+ PartitionerCb *partitioner_cb,
+ std::string &errstr) = 0;
+
+ /** @brief Use with \p name = \c \"partitioner_key_pointer_cb\" */
+ virtual Conf::ConfResult set(const std::string &name,
+ PartitionerKeyPointerCb *partitioner_kp_cb,
+ std::string &errstr) = 0;
+
+ /** @brief Use with \p name = \c \"socket_cb\" */
+ virtual Conf::ConfResult set(const std::string &name,
+ SocketCb *socket_cb,
+ std::string &errstr) = 0;
+
+ /** @brief Use with \p name = \c \"open_cb\" */
+ virtual Conf::ConfResult set(const std::string &name,
+ OpenCb *open_cb,
+ std::string &errstr) = 0;
+
+ /** @brief Use with \p name = \c \"rebalance_cb\" */
+ virtual Conf::ConfResult set(const std::string &name,
+ RebalanceCb *rebalance_cb,
+ std::string &errstr) = 0;
+
+ /** @brief Use with \p name = \c \"offset_commit_cb\" */
+ virtual Conf::ConfResult set(const std::string &name,
+ OffsetCommitCb *offset_commit_cb,
+ std::string &errstr) = 0;
+
+ /** @brief Use with \p name = \c \"ssl_cert_verify_cb\".
+ * @returns CONF_OK on success or CONF_INVALID if SSL is
+ * not supported in this build.
+ */
+ virtual Conf::ConfResult set(const std::string &name,
+ SslCertificateVerifyCb *ssl_cert_verify_cb,
+ std::string &errstr) = 0;
+
+ /**
+ * @brief Set certificate/key \p cert_type from the \p cert_enc encoded
+ * memory at \p buffer of \p size bytes.
+ *
+ * @param cert_type Certificate or key type to configure.
+ * @param cert_enc Buffer \p encoding type.
+ * @param buffer Memory pointer to encoded certificate or key.
+ * The memory is not referenced after this function returns.
+ * @param size Size of memory at \p buffer.
+ * @param errstr A human-readable error string will be written to this string
+ * on failure.
+ *
+ * @returns CONF_OK on success or CONF_INVALID if the memory in
+ * \p buffer is of incorrect encoding, or if librdkafka
+ * was not built with SSL support.
+ *
+ * @remark Calling this method multiple times with the same \p cert_type
+ * will replace the previous value.
+ *
+ * @remark Calling this method with \p buffer set to NULL will clear the
+ * configuration for \p cert_type.
+ *
+ * @remark The private key may require a password, which must be specified
+ * with the `ssl.key.password` configuration property prior to
+ * calling this function.
+ *
+ * @remark Private and public keys in PEM format may also be set with the
+ * `ssl.key.pem` and `ssl.certificate.pem` configuration properties.
+ *
+ * @remark CA certificate in PEM format may also be set with the
+ * `ssl.ca.pem` configuration property.
+ *
+ * @remark When librdkafka is linked to OpenSSL 3.0 and the certificate is
+ * encoded using an obsolete cipher, it might be necessary to set up
+ * an OpenSSL configuration file to load the "legacy" provider and
+ * set the OPENSSL_CONF environment variable.
+ * See
+ * https://github.com/openssl/openssl/blob/master/README-PROVIDERS.md for more
+ * information.
+ */
+ virtual Conf::ConfResult set_ssl_cert(RdKafka::CertificateType cert_type,
+ RdKafka::CertificateEncoding cert_enc,
+ const void *buffer,
+ size_t size,
+ std::string &errstr) = 0;
+
+ /** @brief Query single configuration value
+ *
+ * Do not use this method to get callbacks registered by the configuration
+ * file. Instead use the specific get() methods with the specific callback
+ * parameter in the signature.
+ *
+ * Fallthrough:
+ * Topic-level configuration properties from the \c default_topic_conf
+ * may be retrieved using this interface.
+ *
+ * @returns CONF_OK if the property was set previously set and
+ * returns the value in \p value. */
+ virtual Conf::ConfResult get(const std::string &name,
+ std::string &value) const = 0;
+
+ /** @brief Query single configuration value
+ * @returns CONF_OK if the property was set previously set and
+ * returns the value in \p dr_cb. */
+ virtual Conf::ConfResult get(DeliveryReportCb *&dr_cb) const = 0;
+
+ /** @brief Query single configuration value
+ * @returns CONF_OK if the property was set previously set and
+ * returns the value in \p oauthbearer_token_refresh_cb. */
+ virtual Conf::ConfResult get(
+ OAuthBearerTokenRefreshCb *&oauthbearer_token_refresh_cb) const = 0;
+
+ /** @brief Query single configuration value
+ * @returns CONF_OK if the property was set previously set and
+ * returns the value in \p event_cb. */
+ virtual Conf::ConfResult get(EventCb *&event_cb) const = 0;
+
+ /** @brief Query single configuration value
+ * @returns CONF_OK if the property was set previously set and
+ * returns the value in \p partitioner_cb. */
+ virtual Conf::ConfResult get(PartitionerCb *&partitioner_cb) const = 0;
+
+ /** @brief Query single configuration value
+ * @returns CONF_OK if the property was set previously set and
+ * returns the value in \p partitioner_kp_cb. */
+ virtual Conf::ConfResult get(
+ PartitionerKeyPointerCb *&partitioner_kp_cb) const = 0;
+
+ /** @brief Query single configuration value
+ * @returns CONF_OK if the property was set previously set and
+ * returns the value in \p socket_cb. */
+ virtual Conf::ConfResult get(SocketCb *&socket_cb) const = 0;
+
+ /** @brief Query single configuration value
+ * @returns CONF_OK if the property was set previously set and
+ * returns the value in \p open_cb. */
+ virtual Conf::ConfResult get(OpenCb *&open_cb) const = 0;
+
+ /** @brief Query single configuration value
+ * @returns CONF_OK if the property was set previously set and
+ * returns the value in \p rebalance_cb. */
+ virtual Conf::ConfResult get(RebalanceCb *&rebalance_cb) const = 0;
+
+ /** @brief Query single configuration value
+ * @returns CONF_OK if the property was set previously set and
+ * returns the value in \p offset_commit_cb. */
+ virtual Conf::ConfResult get(OffsetCommitCb *&offset_commit_cb) const = 0;
+
+ /** @brief Use with \p name = \c \"ssl_cert_verify_cb\" */
+ virtual Conf::ConfResult get(
+ SslCertificateVerifyCb *&ssl_cert_verify_cb) const = 0;
+
+ /** @brief Dump configuration names and values to list containing
+ * name,value tuples */
+ virtual std::list<std::string> *dump() = 0;
+
+ /** @brief Use with \p name = \c \"consume_cb\" */
+ virtual Conf::ConfResult set(const std::string &name,
+ ConsumeCb *consume_cb,
+ std::string &errstr) = 0;
+
+ /**
+ * @brief Returns the underlying librdkafka C rd_kafka_conf_t handle.
+ *
+ * @warning Calling the C API on this handle is not recommended and there
+ * is no official support for it, but for cases where the C++
+ * does not provide the proper functionality this C handle can be
+ * used to interact directly with the core librdkafka API.
+ *
+ * @remark The lifetime of the returned pointer is the same as the Conf
+ * object this method is called on.
+ *
+ * @remark Include <rdkafka/rdkafka.h> prior to including
+ * <rdkafka/rdkafkacpp.h>
+ *
+ * @returns \c rd_kafka_conf_t* if this is a CONF_GLOBAL object, else NULL.
+ */
+ virtual struct rd_kafka_conf_s *c_ptr_global() = 0;
+
+ /**
+ * @brief Returns the underlying librdkafka C rd_kafka_topic_conf_t handle.
+ *
+ * @warning Calling the C API on this handle is not recommended and there
+ * is no official support for it, but for cases where the C++
+ * does not provide the proper functionality this C handle can be
+ * used to interact directly with the core librdkafka API.
+ *
+ * @remark The lifetime of the returned pointer is the same as the Conf
+ * object this method is called on.
+ *
+ * @remark Include <rdkafka/rdkafka.h> prior to including
+ * <rdkafka/rdkafkacpp.h>
+ *
+ * @returns \c rd_kafka_topic_conf_t* if this is a CONF_TOPIC object,
+ * else NULL.
+ */
+ virtual struct rd_kafka_topic_conf_s *c_ptr_topic() = 0;
+
+ /**
+ * @brief Set callback_data for ssl engine.
+ *
+ * @remark The \c ssl.engine.location configuration must be set for this
+ * to have affect.
+ *
+ * @remark The memory pointed to by \p value must remain valid for the
+ * lifetime of the configuration object and any Kafka clients that
+ * use it.
+ *
+ * @returns CONF_OK on success, else CONF_INVALID.
+ */
+ virtual Conf::ConfResult set_engine_callback_data(void *value,
+ std::string &errstr) = 0;
+
+
+ /** @brief Enable/disable creation of a queue specific to SASL events
+ * and callbacks.
+ *
+ * For SASL mechanisms that trigger callbacks (currently OAUTHBEARER) this
+ * configuration API allows an application to get a dedicated
+ * queue for the SASL events/callbacks. After enabling the queue with this API
+ * the application can retrieve the queue by calling
+ * RdKafka::Handle::get_sasl_queue() on the client instance.
+ * This queue may then be served directly by the application
+ * (RdKafka::Queue::poll()) or forwarded to another queue, such as
+ * the background queue.
+ *
+ * A convenience function is available to automatically forward the SASL queue
+ * to librdkafka's background thread, see
+ * RdKafka::Handle::sasl_background_callbacks_enable().
+ *
+ * By default (\p enable = false) the main queue (as served by
+ * RdKafka::Handle::poll(), et.al.) is used for SASL callbacks.
+ *
+ * @remark The SASL queue is currently only used by the SASL OAUTHBEARER "
+ * mechanism's token refresh callback.
+ */
+ virtual Conf::ConfResult enable_sasl_queue(bool enable,
+ std::string &errstr) = 0;
+};
+
+/**@}*/
+
+
+/**
+ * @name Kafka base client handle
+ * @{
+ *
+ */
+
+/**
+ * @brief Base handle, super class for specific clients.
+ */
+class RD_EXPORT Handle {
+ public:
+ virtual ~Handle() {
+ }
+
+ /** @returns the name of the handle */
+ virtual std::string name() const = 0;
+
+ /**
+ * @brief Returns the client's broker-assigned group member id
+ *
+ * @remark This currently requires the high-level KafkaConsumer
+ *
+ * @returns Last assigned member id, or empty string if not currently
+ * a group member.
+ */
+ virtual std::string memberid() const = 0;
+
+
+ /**
+ * @brief Polls the provided kafka handle for events.
+ *
+ * Events will trigger application provided callbacks to be called.
+ *
+ * The \p timeout_ms argument specifies the maximum amount of time
+ * (in milliseconds) that the call will block waiting for events.
+ * For non-blocking calls, provide 0 as \p timeout_ms.
+ * To wait indefinately for events, provide -1.
+ *
+ * Events:
+ * - delivery report callbacks (if an RdKafka::DeliveryCb is configured)
+ * [producer]
+ * - event callbacks (if an RdKafka::EventCb is configured) [producer &
+ * consumer]
+ *
+ * @remark An application should make sure to call poll() at regular
+ * intervals to serve any queued callbacks waiting to be called.
+ *
+ * @warning This method MUST NOT be used with the RdKafka::KafkaConsumer,
+ * use its RdKafka::KafkaConsumer::consume() instead.
+ *
+ * @returns the number of events served.
+ */
+ virtual int poll(int timeout_ms) = 0;
+
+ /**
+ * @brief Returns the current out queue length
+ *
+ * The out queue contains messages and requests waiting to be sent to,
+ * or acknowledged by, the broker.
+ */
+ virtual int outq_len() = 0;
+
+ /**
+ * @brief Request Metadata from broker.
+ *
+ * Parameters:
+ * \p all_topics - if non-zero: request info about all topics in cluster,
+ * if zero: only request info about locally known topics.
+ * \p only_rkt - only request info about this topic
+ * \p metadatap - pointer to hold metadata result.
+ * The \p *metadatap pointer must be released with \c
+ * delete. \p timeout_ms - maximum response time before failing.
+ *
+ * @returns RdKafka::ERR_NO_ERROR on success (in which case \p *metadatap
+ * will be set), else RdKafka::ERR__TIMED_OUT on timeout or
+ * other error code on error.
+ */
+ virtual ErrorCode metadata(bool all_topics,
+ const Topic *only_rkt,
+ Metadata **metadatap,
+ int timeout_ms) = 0;
+
+
+ /**
+ * @brief Pause producing or consumption for the provided list of partitions.
+ *
+ * Success or error is returned per-partition in the \p partitions list.
+ *
+ * @returns ErrorCode::NO_ERROR
+ *
+ * @sa resume()
+ */
+ virtual ErrorCode pause(std::vector<TopicPartition *> &partitions) = 0;
+
+
+ /**
+ * @brief Resume producing or consumption for the provided list of partitions.
+ *
+ * Success or error is returned per-partition in the \p partitions list.
+ *
+ * @returns ErrorCode::NO_ERROR
+ *
+ * @sa pause()
+ */
+ virtual ErrorCode resume(std::vector<TopicPartition *> &partitions) = 0;
+
+
+ /**
+ * @brief Query broker for low (oldest/beginning)
+ * and high (newest/end) offsets for partition.
+ *
+ * Offsets are returned in \p *low and \p *high respectively.
+ *
+ * @returns RdKafka::ERR_NO_ERROR on success or an error code on failure.
+ */
+ virtual ErrorCode query_watermark_offsets(const std::string &topic,
+ int32_t partition,
+ int64_t *low,
+ int64_t *high,
+ int timeout_ms) = 0;
+
+ /**
+ * @brief Get last known low (oldest/beginning)
+ * and high (newest/end) offsets for partition.
+ *
+ * The low offset is updated periodically (if statistics.interval.ms is set)
+ * while the high offset is updated on each fetched message set from the
+ * broker.
+ *
+ * If there is no cached offset (either low or high, or both) then
+ * OFFSET_INVALID will be returned for the respective offset.
+ *
+ * Offsets are returned in \p *low and \p *high respectively.
+ *
+ * @returns RdKafka::ERR_NO_ERROR on success or an error code on failure.
+ *
+ * @remark Shall only be used with an active consumer instance.
+ */
+ virtual ErrorCode get_watermark_offsets(const std::string &topic,
+ int32_t partition,
+ int64_t *low,
+ int64_t *high) = 0;
+
+
+ /**
+ * @brief Look up the offsets for the given partitions by timestamp.
+ *
+ * The returned offset for each partition is the earliest offset whose
+ * timestamp is greater than or equal to the given timestamp in the
+ * corresponding partition.
+ *
+ * The timestamps to query are represented as \c offset in \p offsets
+ * on input, and \c offset() will return the closest earlier offset
+ * for the timestamp on output.
+ *
+ * Timestamps are expressed as milliseconds since epoch (UTC).
+ *
+ * The function will block for at most \p timeout_ms milliseconds.
+ *
+ * @remark Duplicate Topic+Partitions are not supported.
+ * @remark Errors are also returned per TopicPartition, see \c err()
+ *
+ * @returns an error code for general errors, else RdKafka::ERR_NO_ERROR
+ * in which case per-partition errors might be set.
+ */
+ virtual ErrorCode offsetsForTimes(std::vector<TopicPartition *> &offsets,
+ int timeout_ms) = 0;
+
+
+ /**
+ * @brief Retrieve queue for a given partition.
+ *
+ * @returns The fetch queue for the given partition if successful. Else,
+ * NULL is returned.
+ *
+ * @remark This function only works on consumers.
+ */
+ virtual Queue *get_partition_queue(const TopicPartition *partition) = 0;
+
+ /**
+ * @brief Forward librdkafka logs (and debug) to the specified queue
+ * for serving with one of the ..poll() calls.
+ *
+ * This allows an application to serve log callbacks (\c log_cb)
+ * in its thread of choice.
+ *
+ * @param queue Queue to forward logs to. If the value is NULL the logs
+ * are forwarded to the main queue.
+ *
+ * @remark The configuration property \c log.queue MUST also be set to true.
+ *
+ * @remark librdkafka maintains its own reference to the provided queue.
+ *
+ * @returns ERR_NO_ERROR on success or an error code on error.
+ */
+ virtual ErrorCode set_log_queue(Queue *queue) = 0;
+
+ /**
+ * @brief Cancels the current callback dispatcher (Handle::poll(),
+ * KafkaConsumer::consume(), etc).
+ *
+ * A callback may use this to force an immediate return to the calling
+ * code (caller of e.g. Handle::poll()) without processing any further
+ * events.
+ *
+ * @remark This function MUST ONLY be called from within a
+ * librdkafka callback.
+ */
+ virtual void yield() = 0;
+
+ /**
+ * @brief Returns the ClusterId as reported in broker metadata.
+ *
+ * @param timeout_ms If there is no cached value from metadata retrieval
+ * then this specifies the maximum amount of time
+ * (in milliseconds) the call will block waiting
+ * for metadata to be retrieved.
+ * Use 0 for non-blocking calls.
+ *
+ * @remark Requires broker version >=0.10.0 and api.version.request=true.
+ *
+ * @returns Last cached ClusterId, or empty string if no ClusterId could be
+ * retrieved in the allotted timespan.
+ */
+ virtual std::string clusterid(int timeout_ms) = 0;
+
+ /**
+ * @brief Returns the underlying librdkafka C rd_kafka_t handle.
+ *
+ * @warning Calling the C API on this handle is not recommended and there
+ * is no official support for it, but for cases where the C++
+ * does not provide the proper functionality this C handle can be
+ * used to interact directly with the core librdkafka API.
+ *
+ * @remark The lifetime of the returned pointer is the same as the Topic
+ * object this method is called on.
+ *
+ * @remark Include <rdkafka/rdkafka.h> prior to including
+ * <rdkafka/rdkafkacpp.h>
+ *
+ * @returns \c rd_kafka_t*
+ */
+ virtual struct rd_kafka_s *c_ptr() = 0;
+
+ /**
+ * @brief Returns the current ControllerId (controller broker id)
+ * as reported in broker metadata.
+ *
+ * @param timeout_ms If there is no cached value from metadata retrieval
+ * then this specifies the maximum amount of time
+ * (in milliseconds) the call will block waiting
+ * for metadata to be retrieved.
+ * Use 0 for non-blocking calls.
+ *
+ * @remark Requires broker version >=0.10.0 and api.version.request=true.
+ *
+ * @returns Last cached ControllerId, or -1 if no ControllerId could be
+ * retrieved in the allotted timespan.
+ */
+ virtual int32_t controllerid(int timeout_ms) = 0;
+
+
+ /**
+ * @brief Returns the first fatal error set on this client instance,
+ * or ERR_NO_ERROR if no fatal error has occurred.
+ *
+ * This function is to be used with the Idempotent Producer and
+ * the Event class for \c EVENT_ERROR events to detect fatal errors.
+ *
+ * Generally all errors raised by the error event are to be considered
+ * informational and temporary, the client will try to recover from all
+ * errors in a graceful fashion (by retrying, etc).
+ *
+ * However, some errors should logically be considered fatal to retain
+ * consistency; in particular a set of errors that may occur when using the
+ * Idempotent Producer and the in-order or exactly-once producer guarantees
+ * can't be satisfied.
+ *
+ * @param errstr A human readable error string if a fatal error was set.
+ *
+ * @returns ERR_NO_ERROR if no fatal error has been raised, else
+ * any other error code.
+ */
+ virtual ErrorCode fatal_error(std::string &errstr) const = 0;
+
+ /**
+ * @brief Set SASL/OAUTHBEARER token and metadata
+ *
+ * @param token_value the mandatory token value to set, often (but not
+ * necessarily) a JWS compact serialization as per
+ * https://tools.ietf.org/html/rfc7515#section-3.1.
+ * @param md_lifetime_ms when the token expires, in terms of the number of
+ * milliseconds since the epoch.
+ * @param md_principal_name the Kafka principal name associated with the
+ * token.
+ * @param extensions potentially empty SASL extension keys and values where
+ * element [i] is the key and [i+1] is the key's value, to be communicated
+ * to the broker as additional key-value pairs during the initial client
+ * response as per https://tools.ietf.org/html/rfc7628#section-3.1. The
+ * number of SASL extension keys plus values must be a non-negative multiple
+ * of 2. Any provided keys and values are copied.
+ * @param errstr A human readable error string is written here, only if
+ * there is an error.
+ *
+ * The SASL/OAUTHBEARER token refresh callback should invoke
+ * this method upon success. The extension keys must not include the reserved
+ * key "`auth`", and all extension keys and values must conform to the
+ * required format as per https://tools.ietf.org/html/rfc7628#section-3.1:
+ *
+ * key = 1*(ALPHA)
+ * value = *(VCHAR / SP / HTAB / CR / LF )
+ *
+ * @returns \c RdKafka::ERR_NO_ERROR on success, otherwise \p errstr set
+ * and:<br>
+ * \c RdKafka::ERR__INVALID_ARG if any of the arguments are
+ * invalid;<br>
+ * \c RdKafka::ERR__NOT_IMPLEMENTED if SASL/OAUTHBEARER is not
+ * supported by this build;<br>
+ * \c RdKafka::ERR__STATE if SASL/OAUTHBEARER is supported but is
+ * not configured as the client's authentication mechanism.<br>
+ *
+ * @sa RdKafka::oauthbearer_set_token_failure
+ * @sa RdKafka::Conf::set() \c "oauthbearer_token_refresh_cb"
+ */
+ virtual ErrorCode oauthbearer_set_token(
+ const std::string &token_value,
+ int64_t md_lifetime_ms,
+ const std::string &md_principal_name,
+ const std::list<std::string> &extensions,
+ std::string &errstr) = 0;
+
+ /**
+ * @brief SASL/OAUTHBEARER token refresh failure indicator.
+ *
+ * @param errstr human readable error reason for failing to acquire a token.
+ *
+ * The SASL/OAUTHBEARER token refresh callback should
+ * invoke this method upon failure to refresh the token.
+ *
+ * @returns \c RdKafka::ERR_NO_ERROR on success, otherwise:<br>
+ * \c RdKafka::ERR__NOT_IMPLEMENTED if SASL/OAUTHBEARER is not
+ * supported by this build;<br>
+ * \c RdKafka::ERR__STATE if SASL/OAUTHBEARER is supported but is
+ * not configured as the client's authentication mechanism.
+ *
+ * @sa RdKafka::oauthbearer_set_token
+ * @sa RdKafka::Conf::set() \c "oauthbearer_token_refresh_cb"
+ */
+ virtual ErrorCode oauthbearer_set_token_failure(
+ const std::string &errstr) = 0;
+
+ /**
+ * @brief Enable SASL OAUTHBEARER refresh callbacks on the librdkafka
+ * background thread.
+ *
+ * This serves as an alternative for applications that do not
+ * call RdKafka::Handle::poll() (et.al.) at regular intervals.
+ */
+ virtual Error *sasl_background_callbacks_enable() = 0;
+
+
+ /**
+ * @returns the SASL callback queue, if enabled, else NULL.
+ *
+ * @sa RdKafka::Conf::enable_sasl_queue()
+ */
+ virtual Queue *get_sasl_queue() = 0;
+
+ /**
+ * @returns the librdkafka background thread queue.
+ */
+ virtual Queue *get_background_queue() = 0;
+
+
+
+ /**
+ * @brief Allocate memory using the same allocator librdkafka uses.
+ *
+ * This is typically an abstraction for the malloc(3) call and makes sure
+ * the application can use the same memory allocator as librdkafka for
+ * allocating pointers that are used by librdkafka.
+ *
+ * @remark Memory allocated by mem_malloc() must be freed using
+ * mem_free().
+ */
+ virtual void *mem_malloc(size_t size) = 0;
+
+ /**
+ * @brief Free pointer returned by librdkafka
+ *
+ * This is typically an abstraction for the free(3) call and makes sure
+ * the application can use the same memory allocator as librdkafka for
+ * freeing pointers returned by librdkafka.
+ *
+ * In standard setups it is usually not necessary to use this interface
+ * rather than the free(3) function.
+ *
+ * @remark mem_free() must only be used for pointers returned by APIs
+ * that explicitly mention using this function for freeing.
+ */
+ virtual void mem_free(void *ptr) = 0;
+
+ /**
+ * @brief Sets SASL credentials used for SASL PLAIN and SCRAM mechanisms by
+ * this Kafka client.
+ *
+ * This function sets or resets the SASL username and password credentials
+ * used by this Kafka client. The new credentials will be used the next time
+ * this client needs to authenticate to a broker.
+ * will not disconnect existing connections that might have been made using
+ * the old credentials.
+ *
+ * @remark This function only applies to the SASL PLAIN and SCRAM mechanisms.
+ *
+ * @returns NULL on success or an error object on error.
+ */
+ virtual Error *sasl_set_credentials(const std::string &username,
+ const std::string &password) = 0;
+};
+
+
+/**@}*/
+
+
+/**
+ * @name Topic and partition objects
+ * @{
+ *
+ */
+
+/**
+ * @brief Topic+Partition
+ *
+ * This is a generic type to hold a single partition and various
+ * information about it.
+ *
+ * Is typically used with std::vector<RdKafka::TopicPartition*> to provide
+ * a list of partitions for different operations.
+ */
+class RD_EXPORT TopicPartition {
+ public:
+ /**
+ * @brief Create topic+partition object for \p topic and \p partition.
+ *
+ * Use \c delete to deconstruct.
+ */
+ static TopicPartition *create(const std::string &topic, int partition);
+
+ /**
+ * @brief Create topic+partition object for \p topic and \p partition
+ * with offset \p offset.
+ *
+ * Use \c delete to deconstruct.
+ */
+ static TopicPartition *create(const std::string &topic,
+ int partition,
+ int64_t offset);
+
+ virtual ~TopicPartition() = 0;
+
+ /**
+ * @brief Destroy/delete the TopicPartitions in \p partitions
+ * and clear the vector.
+ */
+ static void destroy(std::vector<TopicPartition *> &partitions);
+
+ /** @returns topic name */
+ virtual const std::string &topic() const = 0;
+
+ /** @returns partition id */
+ virtual int partition() const = 0;
+
+ /** @returns offset (if applicable) */
+ virtual int64_t offset() const = 0;
+
+ /** @brief Set offset */
+ virtual void set_offset(int64_t offset) = 0;
+
+ /** @returns error code (if applicable) */
+ virtual ErrorCode err() const = 0;
+
+ /** @brief Get partition leader epoch, or -1 if not known or relevant. */
+ virtual int32_t get_leader_epoch() = 0;
+
+ /** @brief Set partition leader epoch. */
+ virtual void set_leader_epoch(int32_t leader_epoch) = 0;
+};
+
+
+
+/**
+ * @brief Topic handle
+ *
+ */
+class RD_EXPORT Topic {
+ public:
+ /**
+ * @brief Unassigned partition.
+ *
+ * The unassigned partition is used by the producer API for messages
+ * that should be partitioned using the configured or default partitioner.
+ */
+ static const int32_t PARTITION_UA;
+
+ /** @brief Special offsets */
+ static const int64_t OFFSET_BEGINNING; /**< Consume from beginning */
+ static const int64_t OFFSET_END; /**< Consume from end */
+ static const int64_t OFFSET_STORED; /**< Use offset storage */
+ static const int64_t OFFSET_INVALID; /**< Invalid offset */
+
+
+ /**
+ * @brief Creates a new topic handle for topic named \p topic_str
+ *
+ * \p conf is an optional configuration for the topic that will be used
+ * instead of the default topic configuration.
+ * The \p conf object is reusable after this call.
+ *
+ * @returns the new topic handle or NULL on error (see \p errstr).
+ */
+ static Topic *create(Handle *base,
+ const std::string &topic_str,
+ const Conf *conf,
+ std::string &errstr);
+
+ virtual ~Topic() = 0;
+
+
+ /** @returns the topic name */
+ virtual std::string name() const = 0;
+
+ /**
+ * @returns true if \p partition is available for the topic (has leader).
+ * @warning \b MUST \b ONLY be called from within a
+ * RdKafka::PartitionerCb callback.
+ */
+ virtual bool partition_available(int32_t partition) const = 0;
+
+ /**
+ * @brief Store offset \p offset + 1 for topic partition \p partition.
+ * The offset will be committed (written) to the broker (or file) according
+ * to \p auto.commit.interval.ms or next manual offset-less commit call.
+ *
+ * @deprecated This API lacks support for partition leader epochs, which makes
+ * it at risk for unclean leader election log truncation issues.
+ * Use KafkaConsumer::offsets_store() or
+ * Message::offset_store() instead.
+ *
+ * @remark \c enable.auto.offset.store must be set to \c false when using
+ * this API.
+ *
+ * @returns RdKafka::ERR_NO_ERROR on success or an error code if none of the
+ * offsets could be stored.
+ */
+ virtual ErrorCode offset_store(int32_t partition, int64_t offset) = 0;
+
+ /**
+ * @brief Returns the underlying librdkafka C rd_kafka_topic_t handle.
+ *
+ * @warning Calling the C API on this handle is not recommended and there
+ * is no official support for it, but for cases where the C++ API
+ * does not provide the underlying functionality this C handle can be
+ * used to interact directly with the core librdkafka API.
+ *
+ * @remark The lifetime of the returned pointer is the same as the Topic
+ * object this method is called on.
+ *
+ * @remark Include <rdkafka/rdkafka.h> prior to including
+ * <rdkafka/rdkafkacpp.h>
+ *
+ * @returns \c rd_kafka_topic_t*
+ */
+ virtual struct rd_kafka_topic_s *c_ptr() = 0;
+};
+
+
+/**@}*/
+
+
+/**
+ * @name Message object
+ * @{
+ *
+ */
+
+
+/**
+ * @brief Message timestamp object
+ *
+ * Represents the number of milliseconds since the epoch (UTC).
+ *
+ * The MessageTimestampType dictates the timestamp type or origin.
+ *
+ * @remark Requires Apache Kafka broker version >= 0.10.0
+ *
+ */
+
+class RD_EXPORT MessageTimestamp {
+ public:
+ /*! Message timestamp type */
+ enum MessageTimestampType {
+ MSG_TIMESTAMP_NOT_AVAILABLE, /**< Timestamp not available */
+ MSG_TIMESTAMP_CREATE_TIME, /**< Message creation time (source) */
+ MSG_TIMESTAMP_LOG_APPEND_TIME /**< Message log append time (broker) */
+ };
+
+ MessageTimestampType type; /**< Timestamp type */
+ int64_t timestamp; /**< Milliseconds since epoch (UTC). */
+};
+
+
+/**
+ * @brief Headers object
+ *
+ * Represents message headers.
+ *
+ * https://cwiki.apache.org/confluence/display/KAFKA/KIP-82+-+Add+Record+Headers
+ *
+ * @remark Requires Apache Kafka >= 0.11.0 brokers
+ */
+class RD_EXPORT Headers {
+ public:
+ virtual ~Headers() = 0;
+
+ /**
+ * @brief Header object
+ *
+ * This object represents a single Header with a key value pair
+ * and an ErrorCode
+ *
+ * @remark dynamic allocation of this object is not supported.
+ */
+ class Header {
+ public:
+ /**
+ * @brief Header object to encapsulate a single Header
+ *
+ * @param key the string value for the header key
+ * @param value the bytes of the header value, or NULL
+ * @param value_size the length in bytes of the header value
+ *
+ * @remark key and value are copied.
+ *
+ */
+ Header(const std::string &key, const void *value, size_t value_size) :
+ key_(key), err_(ERR_NO_ERROR), value_size_(value_size) {
+ value_ = copy_value(value, value_size);
+ }
+
+ /**
+ * @brief Header object to encapsulate a single Header
+ *
+ * @param key the string value for the header key
+ * @param value the bytes of the header value
+ * @param value_size the length in bytes of the header value
+ * @param err the error code if one returned
+ *
+ * @remark The error code is used for when the Header is constructed
+ * internally by using RdKafka::Headers::get_last which constructs
+ * a Header encapsulating the ErrorCode in the process.
+ * If err is set, the value and value_size fields will be undefined.
+ */
+ Header(const std::string &key,
+ const void *value,
+ size_t value_size,
+ const RdKafka::ErrorCode err) :
+ key_(key), err_(err), value_(NULL), value_size_(value_size) {
+ if (err == ERR_NO_ERROR)
+ value_ = copy_value(value, value_size);
+ }
+
+ /**
+ * @brief Copy constructor
+ *
+ * @param other Header to make a copy of.
+ */
+ Header(const Header &other) :
+ key_(other.key_), err_(other.err_), value_size_(other.value_size_) {
+ value_ = copy_value(other.value_, value_size_);
+ }
+
+ /**
+ * @brief Assignment operator
+ *
+ * @param other Header to make a copy of.
+ */
+ Header &operator=(const Header &other) {
+ if (&other == this) {
+ return *this;
+ }
+
+ key_ = other.key_;
+ err_ = other.err_;
+ value_size_ = other.value_size_;
+
+ if (value_ != NULL)
+ mem_free(value_);
+
+ value_ = copy_value(other.value_, value_size_);
+
+ return *this;
+ }
+
+ ~Header() {
+ if (value_ != NULL)
+ mem_free(value_);
+ }
+
+ /** @returns the key/name associated with this Header */
+ std::string key() const {
+ return key_;
+ }
+
+ /** @returns returns the binary value, or NULL */
+ const void *value() const {
+ return value_;
+ }
+
+ /** @returns returns the value casted to a nul-terminated C string,
+ * or NULL. */
+ const char *value_string() const {
+ return static_cast<const char *>(value_);
+ }
+
+ /** @returns Value Size the length of the Value in bytes */
+ size_t value_size() const {
+ return value_size_;
+ }
+
+ /** @returns the error code of this Header (usually ERR_NO_ERROR) */
+ RdKafka::ErrorCode err() const {
+ return err_;
+ }
+
+ private:
+ char *copy_value(const void *value, size_t value_size) {
+ if (!value)
+ return NULL;
+
+ char *dest = (char *)mem_malloc(value_size + 1);
+ memcpy(dest, (const char *)value, value_size);
+ dest[value_size] = '\0';
+
+ return dest;
+ }
+
+ std::string key_;
+ RdKafka::ErrorCode err_;
+ char *value_;
+ size_t value_size_;
+ void *operator new(size_t); /* Prevent dynamic allocation */
+ };
+
+ /**
+ * @brief Create a new instance of the Headers object
+ *
+ * @returns an empty Headers list
+ */
+ static Headers *create();
+
+ /**
+ * @brief Create a new instance of the Headers object from a std::vector
+ *
+ * @param headers std::vector of RdKafka::Headers::Header objects.
+ * The headers are copied, not referenced.
+ *
+ * @returns a Headers list from std::vector set to the size of the std::vector
+ */
+ static Headers *create(const std::vector<Header> &headers);
+
+ /**
+ * @brief Adds a Header to the end of the list.
+ *
+ * @param key header key/name
+ * @param value binary value, or NULL
+ * @param value_size size of the value
+ *
+ * @returns an ErrorCode signalling success or failure to add the header.
+ */
+ virtual ErrorCode add(const std::string &key,
+ const void *value,
+ size_t value_size) = 0;
+
+ /**
+ * @brief Adds a Header to the end of the list.
+ *
+ * Convenience method for adding a std::string as a value for the header.
+ *
+ * @param key header key/name
+ * @param value value string
+ *
+ * @returns an ErrorCode signalling success or failure to add the header.
+ */
+ virtual ErrorCode add(const std::string &key, const std::string &value) = 0;
+
+ /**
+ * @brief Adds a Header to the end of the list.
+ *
+ * This method makes a copy of the passed header.
+ *
+ * @param header Existing header to copy
+ *
+ * @returns an ErrorCode signalling success or failure to add the header.
+ */
+ virtual ErrorCode add(const Header &header) = 0;
+
+ /**
+ * @brief Removes all the Headers of a given key
+ *
+ * @param key header key/name to remove
+ *
+ * @returns An ErrorCode signalling a success or failure to remove the Header.
+ */
+ virtual ErrorCode remove(const std::string &key) = 0;
+
+ /**
+ * @brief Gets all of the Headers of a given key
+ *
+ * @param key header key/name
+ *
+ * @remark If duplicate keys exist this will return them all as a std::vector
+ *
+ * @returns a std::vector containing all the Headers of the given key.
+ */
+ virtual std::vector<Header> get(const std::string &key) const = 0;
+
+ /**
+ * @brief Gets the last occurrence of a Header of a given key
+ *
+ * @param key header key/name
+ *
+ * @remark This will only return the most recently added header
+ *
+ * @returns the Header if found, otherwise a Header with an err set to
+ * ERR__NOENT.
+ */
+ virtual Header get_last(const std::string &key) const = 0;
+
+ /**
+ * @brief Returns all Headers
+ *
+ * @returns a std::vector containing all of the Headers
+ */
+ virtual std::vector<Header> get_all() const = 0;
+
+ /**
+ * @returns the number of headers.
+ */
+ virtual size_t size() const = 0;
+};
+
+
+/**
+ * @brief Message object
+ *
+ * This object represents either a single consumed or produced message,
+ * or an event (\p err() is set).
+ *
+ * An application must check RdKafka::Message::err() to see if the
+ * object is a proper message (error is RdKafka::ERR_NO_ERROR) or a
+ * an error event.
+ *
+ */
+class RD_EXPORT Message {
+ public:
+ /** @brief Message persistence status can be used by the application to
+ * find out if a produced message was persisted in the topic log. */
+ enum Status {
+ /** Message was never transmitted to the broker, or failed with
+ * an error indicating it was not written to the log.
+ * Application retry risks ordering, but not duplication. */
+ MSG_STATUS_NOT_PERSISTED = 0,
+
+ /** Message was transmitted to broker, but no acknowledgement was
+ * received.
+ * Application retry risks ordering and duplication. */
+ MSG_STATUS_POSSIBLY_PERSISTED = 1,
+
+ /** Message was written to the log and fully acknowledged.
+ * No reason for application to retry.
+ * Note: this value should only be trusted with \c acks=all. */
+ MSG_STATUS_PERSISTED = 2,
+ };
+
+ /**
+ * @brief Accessor functions*
+ * @remark Not all fields are present in all types of callbacks.
+ */
+
+ /** @returns The error string if object represent an error event,
+ * else an empty string. */
+ virtual std::string errstr() const = 0;
+
+ /** @returns The error code if object represents an error event, else 0. */
+ virtual ErrorCode err() const = 0;
+
+ /** @returns the RdKafka::Topic object for a message (if applicable),
+ * or NULL if a corresponding RdKafka::Topic object has not been
+ * explicitly created with RdKafka::Topic::create().
+ * In this case use topic_name() instead. */
+ virtual Topic *topic() const = 0;
+
+ /** @returns Topic name (if applicable, else empty string) */
+ virtual std::string topic_name() const = 0;
+
+ /** @returns Partition (if applicable) */
+ virtual int32_t partition() const = 0;
+
+ /** @returns Message payload (if applicable) */
+ virtual void *payload() const = 0;
+
+ /** @returns Message payload length (if applicable) */
+ virtual size_t len() const = 0;
+
+ /** @returns Message key as string (if applicable) */
+ virtual const std::string *key() const = 0;
+
+ /** @returns Message key as void pointer (if applicable) */
+ virtual const void *key_pointer() const = 0;
+
+ /** @returns Message key's binary length (if applicable) */
+ virtual size_t key_len() const = 0;
+
+ /** @returns Message or error offset (if applicable) */
+ virtual int64_t offset() const = 0;
+
+ /** @returns Message timestamp (if applicable) */
+ virtual MessageTimestamp timestamp() const = 0;
+
+ /** @returns The \p msg_opaque as provided to RdKafka::Producer::produce() */
+ virtual void *msg_opaque() const = 0;
+
+ virtual ~Message() = 0;
+
+ /** @returns the latency in microseconds for a produced message measured
+ * from the produce() call, or -1 if latency is not available. */
+ virtual int64_t latency() const = 0;
+
+ /**
+ * @brief Returns the underlying librdkafka C rd_kafka_message_t handle.
+ *
+ * @warning Calling the C API on this handle is not recommended and there
+ * is no official support for it, but for cases where the C++ API
+ * does not provide the underlying functionality this C handle can be
+ * used to interact directly with the core librdkafka API.
+ *
+ * @remark The lifetime of the returned pointer is the same as the Message
+ * object this method is called on.
+ *
+ * @remark Include <rdkafka/rdkafka.h> prior to including
+ * <rdkafka/rdkafkacpp.h>
+ *
+ * @returns \c rd_kafka_message_t*
+ */
+ virtual struct rd_kafka_message_s *c_ptr() = 0;
+
+ /**
+ * @brief Returns the message's persistence status in the topic log.
+ */
+ virtual Status status() const = 0;
+
+ /** @returns the Headers instance for this Message, or NULL if there
+ * are no headers.
+ *
+ * @remark The lifetime of the Headers are the same as the Message. */
+ virtual RdKafka::Headers *headers() = 0;
+
+ /** @returns the Headers instance for this Message (if applicable).
+ * If NULL is returned the reason is given in \p err, which
+ * is either ERR__NOENT if there were no headers, or another
+ * error code if header parsing failed.
+ *
+ * @remark The lifetime of the Headers are the same as the Message. */
+ virtual RdKafka::Headers *headers(RdKafka::ErrorCode *err) = 0;
+
+ /** @returns the broker id of the broker the message was produced to or
+ * fetched from, or -1 if not known/applicable. */
+ virtual int32_t broker_id() const = 0;
+
+ /** @returns the message's partition leader epoch at the time the message was
+ * fetched and if known, else -1. */
+ virtual int32_t leader_epoch() const = 0;
+
+ /**
+ * @brief Store offset +1 for the consumed message.
+ *
+ * The message offset + 1 will be committed to broker according
+ * to \c `auto.commit.interval.ms` or manual offset-less commit()
+ *
+ * @warning This method may only be called for partitions that are currently
+ * assigned.
+ * Non-assigned partitions will fail with ERR__STATE.
+ *
+ * @warning Avoid storing offsets after calling seek() (et.al) as
+ * this may later interfere with resuming a paused partition, instead
+ * store offsets prior to calling seek.
+ *
+ * @remark \c `enable.auto.offset.store` must be set to "false" when using
+ * this API.
+ *
+ * @returns NULL on success or an error object on failure.
+ */
+ virtual Error *offset_store() = 0;
+};
+
+/**@}*/
+
+
+/**
+ * @name Queue interface
+ * @{
+ *
+ */
+
+
+/**
+ * @brief Queue interface
+ *
+ * Create a new message queue. Message queues allows the application
+ * to re-route consumed messages from multiple topic+partitions into
+ * one single queue point. This queue point, containing messages from
+ * a number of topic+partitions, may then be served by a single
+ * consume() method, rather than one per topic+partition combination.
+ *
+ * See the RdKafka::Consumer::start(), RdKafka::Consumer::consume(), and
+ * RdKafka::Consumer::consume_callback() methods that take a queue as the first
+ * parameter for more information.
+ */
+class RD_EXPORT Queue {
+ public:
+ /**
+ * @brief Create Queue object
+ */
+ static Queue *create(Handle *handle);
+
+ /**
+ * @brief Forward/re-route queue to \p dst.
+ * If \p dst is \c NULL, the forwarding is removed.
+ *
+ * The internal refcounts for both queues are increased.
+ *
+ * @remark Regardless of whether \p dst is NULL or not, after calling this
+ * function, \p src will not forward it's fetch queue to the consumer
+ * queue.
+ */
+ virtual ErrorCode forward(Queue *dst) = 0;
+
+
+ /**
+ * @brief Consume message or get error event from the queue.
+ *
+ * @remark Use \c delete to free the message.
+ *
+ * @returns One of:
+ * - proper message (RdKafka::Message::err() is ERR_NO_ERROR)
+ * - error event (RdKafka::Message::err() is != ERR_NO_ERROR)
+ * - timeout due to no message or event in \p timeout_ms
+ * (RdKafka::Message::err() is ERR__TIMED_OUT)
+ */
+ virtual Message *consume(int timeout_ms) = 0;
+
+ /**
+ * @brief Poll queue, serving any enqueued callbacks.
+ *
+ * @remark Must NOT be used for queues containing messages.
+ *
+ * @returns the number of events served or 0 on timeout.
+ */
+ virtual int poll(int timeout_ms) = 0;
+
+ virtual ~Queue() = 0;
+
+ /**
+ * @brief Enable IO event triggering for queue.
+ *
+ * To ease integration with IO based polling loops this API
+ * allows an application to create a separate file-descriptor
+ * that librdkafka will write \p payload (of size \p size) to
+ * whenever a new element is enqueued on a previously empty queue.
+ *
+ * To remove event triggering call with \p fd = -1.
+ *
+ * librdkafka will maintain a copy of the \p payload.
+ *
+ * @remark When using forwarded queues the IO event must only be enabled
+ * on the final forwarded-to (destination) queue.
+ */
+ virtual void io_event_enable(int fd, const void *payload, size_t size) = 0;
+};
+
+/**@}*/
+
+/**
+ * @name ConsumerGroupMetadata
+ * @{
+ *
+ */
+/**
+ * @brief ConsumerGroupMetadata holds a consumer instance's group
+ * metadata state.
+ *
+ * This class currently does not have any public methods.
+ */
+class RD_EXPORT ConsumerGroupMetadata {
+ public:
+ virtual ~ConsumerGroupMetadata() = 0;
+};
+
+/**@}*/
+
+/**
+ * @name KafkaConsumer
+ * @{
+ *
+ */
+
+
+/**
+ * @brief High-level KafkaConsumer (for brokers 0.9 and later)
+ *
+ * @remark Requires Apache Kafka >= 0.9.0 brokers
+ *
+ * Currently supports the \c range and \c roundrobin partition assignment
+ * strategies (see \c partition.assignment.strategy)
+ */
+class RD_EXPORT KafkaConsumer : public virtual Handle {
+ public:
+ /**
+ * @brief Creates a KafkaConsumer.
+ *
+ * The \p conf object must have \c group.id set to the consumer group to join.
+ *
+ * Use RdKafka::KafkaConsumer::close() to shut down the consumer.
+ *
+ * @sa RdKafka::RebalanceCb
+ * @sa CONFIGURATION.md for \c group.id, \c session.timeout.ms,
+ * \c partition.assignment.strategy, etc.
+ */
+ static KafkaConsumer *create(const Conf *conf, std::string &errstr);
+
+ virtual ~KafkaConsumer() = 0;
+
+
+ /** @brief Returns the current partition assignment as set by
+ * RdKafka::KafkaConsumer::assign() */
+ virtual ErrorCode assignment(
+ std::vector<RdKafka::TopicPartition *> &partitions) = 0;
+
+ /** @brief Returns the current subscription as set by
+ * RdKafka::KafkaConsumer::subscribe() */
+ virtual ErrorCode subscription(std::vector<std::string> &topics) = 0;
+
+ /**
+ * @brief Update the subscription set to \p topics.
+ *
+ * Any previous subscription will be unassigned and unsubscribed first.
+ *
+ * The subscription set denotes the desired topics to consume and this
+ * set is provided to the partition assignor (one of the elected group
+ * members) for all clients which then uses the configured
+ * \c partition.assignment.strategy to assign the subscription sets's
+ * topics's partitions to the consumers, depending on their subscription.
+ *
+ * The result of such an assignment is a rebalancing which is either
+ * handled automatically in librdkafka or can be overridden by the application
+ * by providing a RdKafka::RebalanceCb.
+ *
+ * The rebalancing passes the assigned partition set to
+ * RdKafka::KafkaConsumer::assign() to update what partitions are actually
+ * being fetched by the KafkaConsumer.
+ *
+ * Regex pattern matching automatically performed for topics prefixed
+ * with \c \"^\" (e.g. \c \"^myPfx[0-9]_.*\"
+ *
+ * @remark A consumer error will be raised for each unavailable topic in the
+ * \p topics. The error will be ERR_UNKNOWN_TOPIC_OR_PART
+ * for non-existent topics, and
+ * ERR_TOPIC_AUTHORIZATION_FAILED for unauthorized topics.
+ * The consumer error will be raised through consume() (et.al.)
+ * with the \c RdKafka::Message::err() returning one of the
+ * error codes mentioned above.
+ * The subscribe function itself is asynchronous and will not return
+ * an error on unavailable topics.
+ *
+ * @returns an error if the provided list of topics is invalid.
+ */
+ virtual ErrorCode subscribe(const std::vector<std::string> &topics) = 0;
+
+ /** @brief Unsubscribe from the current subscription set. */
+ virtual ErrorCode unsubscribe() = 0;
+
+ /**
+ * @brief Update the assignment set to \p partitions.
+ *
+ * The assignment set is the set of partitions actually being consumed
+ * by the KafkaConsumer.
+ */
+ virtual ErrorCode assign(const std::vector<TopicPartition *> &partitions) = 0;
+
+ /**
+ * @brief Stop consumption and remove the current assignment.
+ */
+ virtual ErrorCode unassign() = 0;
+
+ /**
+ * @brief Consume message or get error event, triggers callbacks.
+ *
+ * Will automatically call registered callbacks for any such queued events,
+ * including RdKafka::RebalanceCb, RdKafka::EventCb, RdKafka::OffsetCommitCb,
+ * etc.
+ *
+ * @remark Use \c delete to free the message.
+ *
+ * @remark An application should make sure to call consume() at regular
+ * intervals, even if no messages are expected, to serve any
+ * queued callbacks waiting to be called. This is especially
+ * important when a RebalanceCb has been registered as it needs
+ * to be called and handled properly to synchronize internal
+ * consumer state.
+ *
+ * @remark Application MUST NOT call \p poll() on KafkaConsumer objects.
+ *
+ * @returns One of:
+ * - proper message (RdKafka::Message::err() is ERR_NO_ERROR)
+ * - error event (RdKafka::Message::err() is != ERR_NO_ERROR)
+ * - timeout due to no message or event in \p timeout_ms
+ * (RdKafka::Message::err() is ERR__TIMED_OUT)
+ */
+ virtual Message *consume(int timeout_ms) = 0;
+
+ /**
+ * @brief Commit offsets for the current assignment.
+ *
+ * @remark This is the synchronous variant that blocks until offsets
+ * are committed or the commit fails (see return value).
+ *
+ * @remark If a RdKafka::OffsetCommitCb callback is registered it will
+ * be called with commit details on a future call to
+ * RdKafka::KafkaConsumer::consume()
+
+ *
+ * @returns ERR_NO_ERROR or error code.
+ */
+ virtual ErrorCode commitSync() = 0;
+
+ /**
+ * @brief Asynchronous version of RdKafka::KafkaConsumer::CommitSync()
+ *
+ * @sa RdKafka::KafkaConsumer::commitSync()
+ */
+ virtual ErrorCode commitAsync() = 0;
+
+ /**
+ * @brief Commit offset for a single topic+partition based on \p message
+ *
+ * @remark The offset committed will be the message's offset + 1.
+ *
+ * @remark This is the synchronous variant.
+ *
+ * @sa RdKafka::KafkaConsumer::commitSync()
+ */
+ virtual ErrorCode commitSync(Message *message) = 0;
+
+ /**
+ * @brief Commit offset for a single topic+partition based on \p message
+ *
+ * @remark The offset committed will be the message's offset + 1.
+ *
+ * @remark This is the asynchronous variant.
+ *
+ * @sa RdKafka::KafkaConsumer::commitSync()
+ */
+ virtual ErrorCode commitAsync(Message *message) = 0;
+
+ /**
+ * @brief Commit offsets for the provided list of partitions.
+ *
+ * @remark The \c .offset of the partitions in \p offsets should be the
+ * offset where consumption will resume, i.e., the last
+ * processed offset + 1.
+ *
+ * @remark This is the synchronous variant.
+ */
+ virtual ErrorCode commitSync(std::vector<TopicPartition *> &offsets) = 0;
+
+ /**
+ * @brief Commit offset for the provided list of partitions.
+ *
+ * @remark The \c .offset of the partitions in \p offsets should be the
+ * offset where consumption will resume, i.e., the last
+ * processed offset + 1.
+ *
+ * @remark This is the asynchronous variant.
+ */
+ virtual ErrorCode commitAsync(
+ const std::vector<TopicPartition *> &offsets) = 0;
+
+ /**
+ * @brief Commit offsets for the current assignment.
+ *
+ * @remark This is the synchronous variant that blocks until offsets
+ * are committed or the commit fails (see return value).
+ *
+ * @remark The provided callback will be called from this function.
+ *
+ * @returns ERR_NO_ERROR or error code.
+ */
+ virtual ErrorCode commitSync(OffsetCommitCb *offset_commit_cb) = 0;
+
+ /**
+ * @brief Commit offsets for the provided list of partitions.
+ *
+ * @remark This is the synchronous variant that blocks until offsets
+ * are committed or the commit fails (see return value).
+ *
+ * @remark The provided callback will be called from this function.
+ *
+ * @returns ERR_NO_ERROR or error code.
+ */
+ virtual ErrorCode commitSync(std::vector<TopicPartition *> &offsets,
+ OffsetCommitCb *offset_commit_cb) = 0;
+
+
+
+ /**
+ * @brief Retrieve committed offsets for topics+partitions.
+ *
+ * @returns ERR_NO_ERROR on success in which case the
+ * \p offset or \p err field of each \p partitions' element is filled
+ * in with the stored offset, or a partition specific error.
+ * Else returns an error code.
+ */
+ virtual ErrorCode committed(std::vector<TopicPartition *> &partitions,
+ int timeout_ms) = 0;
+
+ /**
+ * @brief Retrieve current positions (offsets) for topics+partitions.
+ *
+ * @returns ERR_NO_ERROR on success in which case the
+ * \p offset or \p err field of each \p partitions' element is filled
+ * in with the stored offset, or a partition specific error.
+ * Else returns an error code.
+ */
+ virtual ErrorCode position(std::vector<TopicPartition *> &partitions) = 0;
+
+
+ /**
+ * For pausing and resuming consumption, see
+ * @sa RdKafka::Handle::pause() and RdKafka::Handle::resume()
+ */
+
+
+ /**
+ * @brief Close and shut down the consumer.
+ *
+ * This call will block until the following operations are finished:
+ * - Trigger a local rebalance to void the current assignment (if any).
+ * - Stop consumption for current assignment (if any).
+ * - Commit offsets (if any).
+ * - Leave group (if applicable).
+ *
+ * The maximum blocking time is roughly limited to session.timeout.ms.
+ *
+ * @remark Callbacks, such as RdKafka::RebalanceCb and
+ * RdKafka::OffsetCommitCb, etc, may be called.
+ *
+ * @remark The consumer object must later be freed with \c delete
+ */
+ virtual ErrorCode close() = 0;
+
+
+ /**
+ * @brief Seek consumer for topic+partition to offset which is either an
+ * absolute or logical offset.
+ *
+ * If \p timeout_ms is not 0 the call will wait this long for the
+ * seek to be performed. If the timeout is reached the internal state
+ * will be unknown and this function returns `ERR__TIMED_OUT`.
+ * If \p timeout_ms is 0 it will initiate the seek but return
+ * immediately without any error reporting (e.g., async).
+ *
+ * This call triggers a fetch queue barrier flush.
+ *
+ * @remark Consumption for the given partition must have started for the
+ * seek to work. Use assign() to set the starting offset.
+ *
+ * @returns an ErrorCode to indicate success or failure.
+ */
+ virtual ErrorCode seek(const TopicPartition &partition, int timeout_ms) = 0;
+
+
+ /**
+ * @brief Store offset \p offset for topic partition \p partition.
+ * The offset will be committed (written) to the offset store according
+ * to \p auto.commit.interval.ms or the next manual offset-less commit*()
+ *
+ * Per-partition success/error status propagated through TopicPartition.err()
+ *
+ * @remark The \c .offset field is stored as is, it will NOT be + 1.
+ *
+ * @remark \c enable.auto.offset.store must be set to \c false when using
+ * this API.
+ *
+ * @remark The leader epoch, if set, will be used to fence outdated partition
+ * leaders. See TopicPartition::set_leader_epoch().
+ *
+ * @returns RdKafka::ERR_NO_ERROR on success, or
+ * RdKafka::ERR___UNKNOWN_PARTITION if none of the offsets could
+ * be stored, or
+ * RdKafka::ERR___INVALID_ARG if \c enable.auto.offset.store is true.
+ */
+ virtual ErrorCode offsets_store(std::vector<TopicPartition *> &offsets) = 0;
+
+
+ /**
+ * @returns the current consumer group metadata associated with this consumer,
+ * or NULL if the consumer is configured with a \c group.id.
+ * This metadata object should be passed to the transactional
+ * producer's RdKafka::Producer::send_offsets_to_transaction() API.
+ *
+ * @remark The returned object must be deleted by the application.
+ *
+ * @sa RdKafka::Producer::send_offsets_to_transaction()
+ */
+ virtual ConsumerGroupMetadata *groupMetadata() = 0;
+
+
+ /** @brief Check whether the consumer considers the current assignment to
+ * have been lost involuntarily. This method is only applicable for
+ * use with a subscribing consumer. Assignments are revoked
+ * immediately when determined to have been lost, so this method is
+ * only useful within a rebalance callback. Partitions that have
+ * been lost may already be owned by other members in the group and
+ * therefore commiting offsets, for example, may fail.
+ *
+ * @remark Calling assign(), incremental_assign() or incremental_unassign()
+ * resets this flag.
+ *
+ * @returns Returns true if the current partition assignment is considered
+ * lost, false otherwise.
+ */
+ virtual bool assignment_lost() = 0;
+
+ /**
+ * @brief The rebalance protocol currently in use. This will be
+ * "NONE" if the consumer has not (yet) joined a group, else it will
+ * match the rebalance protocol ("EAGER", "COOPERATIVE") of the
+ * configured and selected assignor(s). All configured
+ * assignors must have the same protocol type, meaning
+ * online migration of a consumer group from using one
+ * protocol to another (in particular upgading from EAGER
+ * to COOPERATIVE) without a restart is not currently
+ * supported.
+ *
+ * @returns an empty string on error, or one of
+ * "NONE", "EAGER", "COOPERATIVE" on success.
+ */
+
+ virtual std::string rebalance_protocol() = 0;
+
+
+ /**
+ * @brief Incrementally add \p partitions to the current assignment.
+ *
+ * If a COOPERATIVE assignor (i.e. incremental rebalancing) is being used,
+ * this method should be used in a rebalance callback to adjust the current
+ * assignment appropriately in the case where the rebalance type is
+ * ERR__ASSIGN_PARTITIONS. The application must pass the partition list
+ * passed to the callback (or a copy of it), even if the list is empty.
+ * This method may also be used outside the context of a rebalance callback.
+ *
+ * @returns NULL on success, or an error object if the operation was
+ * unsuccessful.
+ *
+ * @remark The returned object must be deleted by the application.
+ */
+ virtual Error *incremental_assign(
+ const std::vector<TopicPartition *> &partitions) = 0;
+
+
+ /**
+ * @brief Incrementally remove \p partitions from the current assignment.
+ *
+ * If a COOPERATIVE assignor (i.e. incremental rebalancing) is being used,
+ * this method should be used in a rebalance callback to adjust the current
+ * assignment appropriately in the case where the rebalance type is
+ * ERR__REVOKE_PARTITIONS. The application must pass the partition list
+ * passed to the callback (or a copy of it), even if the list is empty.
+ * This method may also be used outside the context of a rebalance callback.
+ *
+ * @returns NULL on success, or an error object if the operation was
+ * unsuccessful.
+ *
+ * @remark The returned object must be deleted by the application.
+ */
+ virtual Error *incremental_unassign(
+ const std::vector<TopicPartition *> &partitions) = 0;
+
+ /**
+ * @brief Close and shut down the consumer.
+ *
+ * Performs the same actions as RdKafka::KafkaConsumer::close() but in a
+ * background thread.
+ *
+ * Rebalance events/callbacks (etc) will be forwarded to the
+ * application-provided \p queue. The application must poll this queue until
+ * RdKafka::KafkaConsumer::closed() returns true.
+ *
+ * @remark Depending on consumer group join state there may or may not be
+ * rebalance events emitted on \p rkqu.
+ *
+ * @returns an error object if the consumer close failed, else NULL.
+ *
+ * @sa RdKafka::KafkaConsumer::closed()
+ */
+ virtual Error *close(Queue *queue) = 0;
+
+
+ /** @returns true if the consumer is closed, else 0.
+ *
+ * @sa RdKafka::KafkaConsumer::close()
+ */
+ virtual bool closed() = 0;
+};
+
+
+/**@}*/
+
+
+/**
+ * @name Simple Consumer (legacy)
+ * @{
+ *
+ */
+
+/**
+ * @brief Simple Consumer (legacy)
+ *
+ * A simple non-balanced, non-group-aware, consumer.
+ */
+class RD_EXPORT Consumer : public virtual Handle {
+ public:
+ /**
+ * @brief Creates a new Kafka consumer handle.
+ *
+ * \p conf is an optional object that will be used instead of the default
+ * configuration.
+ * The \p conf object is reusable after this call.
+ *
+ * @returns the new handle on success or NULL on error in which case
+ * \p errstr is set to a human readable error message.
+ */
+ static Consumer *create(const Conf *conf, std::string &errstr);
+
+ virtual ~Consumer() = 0;
+
+
+ /**
+ * @brief Start consuming messages for topic and \p partition
+ * at offset \p offset which may either be a proper offset (0..N)
+ * or one of the the special offsets: \p OFFSET_BEGINNING or \p OFFSET_END.
+ *
+ * rdkafka will attempt to keep \p queued.min.messages (config property)
+ * messages in the local queue by repeatedly fetching batches of messages
+ * from the broker until the threshold is reached.
+ *
+ * The application shall use one of the \p ..->consume*() functions
+ * to consume messages from the local queue, each kafka message being
+ * represented as a `RdKafka::Message *` object.
+ *
+ * \p ..->start() must not be called multiple times for the same
+ * topic and partition without stopping consumption first with
+ * \p ..->stop().
+ *
+ * @returns an ErrorCode to indicate success or failure.
+ */
+ virtual ErrorCode start(Topic *topic, int32_t partition, int64_t offset) = 0;
+
+ /**
+ * @brief Start consuming messages for topic and \p partition on
+ * queue \p queue.
+ *
+ * @sa RdKafka::Consumer::start()
+ */
+ virtual ErrorCode start(Topic *topic,
+ int32_t partition,
+ int64_t offset,
+ Queue *queue) = 0;
+
+ /**
+ * @brief Stop consuming messages for topic and \p partition, purging
+ * all messages currently in the local queue.
+ *
+ * The application needs to be stop all consumers before destroying
+ * the Consumer handle.
+ *
+ * @returns an ErrorCode to indicate success or failure.
+ */
+ virtual ErrorCode stop(Topic *topic, int32_t partition) = 0;
+
+ /**
+ * @brief Seek consumer for topic+partition to \p offset which is either an
+ * absolute or logical offset.
+ *
+ * If \p timeout_ms is not 0 the call will wait this long for the
+ * seek to be performed. If the timeout is reached the internal state
+ * will be unknown and this function returns `ERR__TIMED_OUT`.
+ * If \p timeout_ms is 0 it will initiate the seek but return
+ * immediately without any error reporting (e.g., async).
+ *
+ * This call triggers a fetch queue barrier flush.
+ *
+ * @returns an ErrorCode to indicate success or failure.
+ */
+ virtual ErrorCode seek(Topic *topic,
+ int32_t partition,
+ int64_t offset,
+ int timeout_ms) = 0;
+
+ /**
+ * @brief Consume a single message from \p topic and \p partition.
+ *
+ * \p timeout_ms is maximum amount of time to wait for a message to be
+ * received.
+ * Consumer must have been previously started with \p ..->start().
+ *
+ * @returns a Message object, the application needs to check if message
+ * is an error or a proper message RdKafka::Message::err() and checking for
+ * \p ERR_NO_ERROR.
+ *
+ * The message object must be destroyed when the application is done with it.
+ *
+ * Errors (in RdKafka::Message::err()):
+ * - ERR__TIMED_OUT - \p timeout_ms was reached with no new messages fetched.
+ * - ERR__PARTITION_EOF - End of partition reached, not an error.
+ */
+ virtual Message *consume(Topic *topic, int32_t partition, int timeout_ms) = 0;
+
+ /**
+ * @brief Consume a single message from the specified queue.
+ *
+ * \p timeout_ms is maximum amount of time to wait for a message to be
+ * received.
+ * Consumer must have been previously started on the queue with
+ * \p ..->start().
+ *
+ * @returns a Message object, the application needs to check if message
+ * is an error or a proper message \p Message->err() and checking for
+ * \p ERR_NO_ERROR.
+ *
+ * The message object must be destroyed when the application is done with it.
+ *
+ * Errors (in RdKafka::Message::err()):
+ * - ERR__TIMED_OUT - \p timeout_ms was reached with no new messages fetched
+ *
+ * Note that Message->topic() may be nullptr after certain kinds of
+ * errors, so applications should check that it isn't null before
+ * dereferencing it.
+ */
+ virtual Message *consume(Queue *queue, int timeout_ms) = 0;
+
+ /**
+ * @brief Consumes messages from \p topic and \p partition, calling
+ * the provided callback for each consumed messsage.
+ *
+ * \p consume_callback() provides higher throughput performance
+ * than \p consume().
+ *
+ * \p timeout_ms is the maximum amount of time to wait for one or
+ * more messages to arrive.
+ *
+ * The provided \p consume_cb instance has its \p consume_cb function
+ * called for every message received.
+ *
+ * The \p opaque argument is passed to the \p consume_cb as \p opaque.
+ *
+ * @returns the number of messages processed or -1 on error.
+ *
+ * @sa RdKafka::Consumer::consume()
+ */
+ virtual int consume_callback(Topic *topic,
+ int32_t partition,
+ int timeout_ms,
+ ConsumeCb *consume_cb,
+ void *opaque) = 0;
+
+ /**
+ * @brief Consumes messages from \p queue, calling the provided callback for
+ * each consumed messsage.
+ *
+ * @sa RdKafka::Consumer::consume_callback()
+ */
+ virtual int consume_callback(Queue *queue,
+ int timeout_ms,
+ RdKafka::ConsumeCb *consume_cb,
+ void *opaque) = 0;
+
+ /**
+ * @brief Converts an offset into the logical offset from the tail of a topic.
+ *
+ * \p offset is the (positive) number of items from the end.
+ *
+ * @returns the logical offset for message \p offset from the tail, this value
+ * may be passed to Consumer::start, et.al.
+ * @remark The returned logical offset is specific to librdkafka.
+ */
+ static int64_t OffsetTail(int64_t offset);
+};
+
+/**@}*/
+
+
+/**
+ * @name Producer
+ * @{
+ *
+ */
+
+
+/**
+ * @brief Producer
+ */
+class RD_EXPORT Producer : public virtual Handle {
+ public:
+ /**
+ * @brief Creates a new Kafka producer handle.
+ *
+ * \p conf is an optional object that will be used instead of the default
+ * configuration.
+ * The \p conf object is reusable after this call.
+ *
+ * @returns the new handle on success or NULL on error in which case
+ * \p errstr is set to a human readable error message.
+ */
+ static Producer *create(const Conf *conf, std::string &errstr);
+
+
+ virtual ~Producer() = 0;
+
+ /**
+ * @brief RdKafka::Producer::produce() \p msgflags
+ *
+ * These flags are optional.
+ */
+ enum {
+ RK_MSG_FREE = 0x1, /**< rdkafka will free(3) \p payload
+ * when it is done with it.
+ * Mutually exclusive with RK_MSG_COPY. */
+ RK_MSG_COPY = 0x2, /**< the \p payload data will be copied
+ * and the \p payload pointer will not
+ * be used by rdkafka after the
+ * call returns.
+ * Mutually exclusive with RK_MSG_FREE. */
+ RK_MSG_BLOCK = 0x4 /**< Block produce*() on message queue
+ * full.
+ * WARNING:
+ * If a delivery report callback
+ * is used the application MUST
+ * call rd_kafka_poll() (or equiv.)
+ * to make sure delivered messages
+ * are drained from the internal
+ * delivery report queue.
+ * Failure to do so will result
+ * in indefinately blocking on
+ * the produce() call when the
+ * message queue is full.
+ */
+
+
+ /**@cond NO_DOC*/
+ /* For backwards compatibility: */
+#ifndef MSG_COPY /* defined in sys/msg.h */
+ , /** this comma must exist betwen
+ * RK_MSG_BLOCK and MSG_FREE
+ */
+ MSG_FREE = RK_MSG_FREE,
+ MSG_COPY = RK_MSG_COPY
+#endif
+ /**@endcond*/
+ };
+
+ /**
+ * @brief Produce and send a single message to broker.
+ *
+ * This is an asynch non-blocking API.
+ *
+ * \p partition is the target partition, either:
+ * - RdKafka::Topic::PARTITION_UA (unassigned) for
+ * automatic partitioning using the topic's partitioner function, or
+ * - a fixed partition (0..N)
+ *
+ * \p msgflags is zero or more of the following flags OR:ed together:
+ * RK_MSG_BLOCK - block \p produce*() call if
+ * \p queue.buffering.max.messages or
+ * \p queue.buffering.max.kbytes are exceeded.
+ * Messages are considered in-queue from the point they
+ * are accepted by produce() until their corresponding
+ * delivery report callback/event returns.
+ * It is thus a requirement to call
+ * poll() (or equiv.) from a separate
+ * thread when RK_MSG_BLOCK is used.
+ * See WARNING on \c RK_MSG_BLOCK above.
+ * RK_MSG_FREE - rdkafka will free(3) \p payload when it is done with it.
+ * RK_MSG_COPY - the \p payload data will be copied and the \p payload
+ * pointer will not be used by rdkafka after the
+ * call returns.
+ *
+ * NOTE: RK_MSG_FREE and RK_MSG_COPY are mutually exclusive.
+ *
+ * If the function returns an error code and RK_MSG_FREE was specified, then
+ * the memory associated with the payload is still the caller's
+ * responsibility.
+ *
+ * \p payload is the message payload of size \p len bytes.
+ *
+ * \p key is an optional message key, if non-NULL it
+ * will be passed to the topic partitioner as well as be sent with the
+ * message to the broker and passed on to the consumer.
+ *
+ * \p msg_opaque is an optional application-provided per-message opaque
+ * pointer that will provided in the delivery report callback (\p dr_cb) for
+ * referencing this message.
+ *
+ * @returns an ErrorCode to indicate success or failure:
+ * - ERR_NO_ERROR - message successfully enqueued for transmission.
+ *
+ * - ERR__QUEUE_FULL - maximum number of outstanding messages has been
+ * reached: \c queue.buffering.max.message
+ *
+ * - ERR_MSG_SIZE_TOO_LARGE - message is larger than configured max size:
+ * \c messages.max.bytes
+ *
+ * - ERR__UNKNOWN_PARTITION - requested \p partition is unknown in the
+ * Kafka cluster.
+ *
+ * - ERR__UNKNOWN_TOPIC - topic is unknown in the Kafka cluster.
+ */
+ virtual ErrorCode produce(Topic *topic,
+ int32_t partition,
+ int msgflags,
+ void *payload,
+ size_t len,
+ const std::string *key,
+ void *msg_opaque) = 0;
+
+ /**
+ * @brief Variant produce() that passes the key as a pointer and length
+ * instead of as a const std::string *.
+ */
+ virtual ErrorCode produce(Topic *topic,
+ int32_t partition,
+ int msgflags,
+ void *payload,
+ size_t len,
+ const void *key,
+ size_t key_len,
+ void *msg_opaque) = 0;
+
+ /**
+ * @brief produce() variant that takes topic as a string (no need for
+ * creating a Topic object), and also allows providing the
+ * message timestamp (milliseconds since beginning of epoch, UTC).
+ * Otherwise identical to produce() above.
+ */
+ virtual ErrorCode produce(const std::string topic_name,
+ int32_t partition,
+ int msgflags,
+ void *payload,
+ size_t len,
+ const void *key,
+ size_t key_len,
+ int64_t timestamp,
+ void *msg_opaque) = 0;
+
+ /**
+ * @brief produce() variant that that allows for Header support on produce
+ * Otherwise identical to produce() above.
+ *
+ * @warning The \p headers will be freed/deleted if the produce() call
+ * succeeds, or left untouched if produce() fails.
+ */
+ virtual ErrorCode produce(const std::string topic_name,
+ int32_t partition,
+ int msgflags,
+ void *payload,
+ size_t len,
+ const void *key,
+ size_t key_len,
+ int64_t timestamp,
+ RdKafka::Headers *headers,
+ void *msg_opaque) = 0;
+
+
+ /**
+ * @brief Variant produce() that accepts vectors for key and payload.
+ * The vector data will be copied.
+ */
+ virtual ErrorCode produce(Topic *topic,
+ int32_t partition,
+ const std::vector<char> *payload,
+ const std::vector<char> *key,
+ void *msg_opaque) = 0;
+
+
+ /**
+ * @brief Wait until all outstanding produce requests, et.al, are completed.
+ * This should typically be done prior to destroying a producer
+ * instance to make sure all queued and in-flight produce requests are
+ * completed before terminating.
+ *
+ * @remark The \c linger.ms time will be ignored for the duration of the call,
+ * queued messages will be sent to the broker as soon as possible.
+ *
+ * @remark This function will call Producer::poll() and thus
+ * trigger callbacks.
+ *
+ * @returns ERR__TIMED_OUT if \p timeout_ms was reached before all
+ * outstanding requests were completed, else ERR_NO_ERROR
+ */
+ virtual ErrorCode flush(int timeout_ms) = 0;
+
+
+ /**
+ * @brief Purge messages currently handled by the producer instance.
+ *
+ * @param purge_flags tells which messages should be purged and how.
+ *
+ * The application will need to call Handle::poll() or Producer::flush()
+ * afterwards to serve the delivery report callbacks of the purged messages.
+ *
+ * Messages purged from internal queues fail with the delivery report
+ * error code set to ERR__PURGE_QUEUE, while purged messages that
+ * are in-flight to or from the broker will fail with the error code set to
+ * ERR__PURGE_INFLIGHT.
+ *
+ * @warning Purging messages that are in-flight to or from the broker
+ * will ignore any sub-sequent acknowledgement for these messages
+ * received from the broker, effectively making it impossible
+ * for the application to know if the messages were successfully
+ * produced or not. This may result in duplicate messages if the
+ * application retries these messages at a later time.
+ *
+ * @remark This call may block for a short time while background thread
+ * queues are purged.
+ *
+ * @returns ERR_NO_ERROR on success,
+ * ERR__INVALID_ARG if the \p purge flags are invalid or unknown,
+ * ERR__NOT_IMPLEMENTED if called on a non-producer client instance.
+ */
+ virtual ErrorCode purge(int purge_flags) = 0;
+
+ /**
+ * @brief RdKafka::Handle::purge() \p purge_flags
+ */
+ enum {
+ PURGE_QUEUE = 0x1, /**< Purge messages in internal queues */
+
+ PURGE_INFLIGHT = 0x2, /*! Purge messages in-flight to or from the broker.
+ * Purging these messages will void any future
+ * acknowledgements from the broker, making it
+ * impossible for the application to know if these
+ * messages were successfully delivered or not.
+ * Retrying these messages may lead to duplicates. */
+
+ PURGE_NON_BLOCKING = 0x4 /* Don't wait for background queue
+ * purging to finish. */
+ };
+
+ /**
+ * @name Transactional API
+ * @{
+ *
+ * Requires Kafka broker version v0.11.0 or later
+ *
+ * See the Transactional API documentation in rdkafka.h for more information.
+ */
+
+ /**
+ * @brief Initialize transactions for the producer instance.
+ *
+ * @param timeout_ms The maximum time to block. On timeout the operation
+ * may continue in the background, depending on state,
+ * and it is okay to call init_transactions() again.
+ *
+ * @returns an RdKafka::Error object on error, or NULL on success.
+ * Check whether the returned error object permits retrying
+ * by calling RdKafka::Error::is_retriable(), or whether a fatal
+ * error has been raised by calling RdKafka::Error::is_fatal().
+ *
+ * @remark The returned error object (if not NULL) must be deleted.
+ *
+ * See rd_kafka_init_transactions() in rdkafka.h for more information.
+ *
+ */
+ virtual Error *init_transactions(int timeout_ms) = 0;
+
+
+ /**
+ * @brief init_transactions() must have been called successfully
+ * (once) before this function is called.
+ *
+ * @returns an RdKafka::Error object on error, or NULL on success.
+ * Check whether a fatal error has been raised by calling
+ * RdKafka::Error::is_fatal_error().
+ *
+ * @remark The returned error object (if not NULL) must be deleted.
+ *
+ * See rd_kafka_begin_transaction() in rdkafka.h for more information.
+ */
+ virtual Error *begin_transaction() = 0;
+
+ /**
+ * @brief Sends a list of topic partition offsets to the consumer group
+ * coordinator for \p group_metadata, and marks the offsets as part
+ * part of the current transaction.
+ * These offsets will be considered committed only if the transaction
+ * is committed successfully.
+ *
+ * The offsets should be the next message your application will
+ * consume,
+ * i.e., the last processed message's offset + 1 for each partition.
+ * Either track the offsets manually during processing or use
+ * RdKafka::KafkaConsumer::position() (on the consumer) to get the
+ * current offsets for
+ * the partitions assigned to the consumer.
+ *
+ * Use this method at the end of a consume-transform-produce loop prior
+ * to committing the transaction with commit_transaction().
+ *
+ * @param offsets List of offsets to commit to the consumer group upon
+ * successful commit of the transaction. Offsets should be
+ * the next message to consume,
+ * e.g., last processed message + 1.
+ * @param group_metadata The current consumer group metadata as returned by
+ * RdKafka::KafkaConsumer::groupMetadata() on the consumer
+ * instance the provided offsets were consumed from.
+ * @param timeout_ms Maximum time allowed to register the
+ * offsets on the broker.
+ *
+ * @remark This function must be called on the transactional producer
+ * instance, not the consumer.
+ *
+ * @remark The consumer must disable auto commits
+ * (set \c enable.auto.commit to false on the consumer).
+ *
+ * @returns an RdKafka::Error object on error, or NULL on success.
+ * Check whether the returned error object permits retrying
+ * by calling RdKafka::Error::is_retriable(), or whether an abortable
+ * or fatal error has been raised by calling
+ * RdKafka::Error::txn_requires_abort() or RdKafka::Error::is_fatal()
+ * respectively.
+ *
+ * @remark The returned error object (if not NULL) must be deleted.
+ *
+ * See rd_kafka_send_offsets_to_transaction() in rdkafka.h for
+ * more information.
+ */
+ virtual Error *send_offsets_to_transaction(
+ const std::vector<TopicPartition *> &offsets,
+ const ConsumerGroupMetadata *group_metadata,
+ int timeout_ms) = 0;
+
+ /**
+ * @brief Commit the current transaction as started with begin_transaction().
+ *
+ * Any outstanding messages will be flushed (delivered) before actually
+ * committing the transaction.
+ *
+ * @param timeout_ms The maximum time to block. On timeout the operation
+ * may continue in the background, depending on state,
+ * and it is okay to call this function again.
+ * Pass -1 to use the remaining transaction timeout,
+ * this is the recommended use.
+ *
+ * @remark It is strongly recommended to always pass -1 (remaining transaction
+ * time) as the \p timeout_ms. Using other values risk internal
+ * state desynchronization in case any of the underlying protocol
+ * requests fail.
+ *
+ * @returns an RdKafka::Error object on error, or NULL on success.
+ * Check whether the returned error object permits retrying
+ * by calling RdKafka::Error::is_retriable(), or whether an abortable
+ * or fatal error has been raised by calling
+ * RdKafka::Error::txn_requires_abort() or RdKafka::Error::is_fatal()
+ * respectively.
+ *
+ * @remark The returned error object (if not NULL) must be deleted.
+ *
+ * See rd_kafka_commit_transaction() in rdkafka.h for more information.
+ */
+ virtual Error *commit_transaction(int timeout_ms) = 0;
+
+ /**
+ * @brief Aborts the ongoing transaction.
+ *
+ * This function should also be used to recover from non-fatal
+ * abortable transaction errors.
+ *
+ * Any outstanding messages will be purged and fail with
+ * RdKafka::ERR__PURGE_INFLIGHT or RdKafka::ERR__PURGE_QUEUE.
+ * See RdKafka::Producer::purge() for details.
+ *
+ * @param timeout_ms The maximum time to block. On timeout the operation
+ * may continue in the background, depending on state,
+ * and it is okay to call this function again.
+ * Pass -1 to use the remaining transaction timeout,
+ * this is the recommended use.
+ *
+ * @remark It is strongly recommended to always pass -1 (remaining transaction
+ * time) as the \p timeout_ms. Using other values risk internal
+ * state desynchronization in case any of the underlying protocol
+ * requests fail.
+ *
+ * @returns an RdKafka::Error object on error, or NULL on success.
+ * Check whether the returned error object permits retrying
+ * by calling RdKafka::Error::is_retriable(), or whether a
+ * fatal error has been raised by calling RdKafka::Error::is_fatal().
+ *
+ * @remark The returned error object (if not NULL) must be deleted.
+ *
+ * See rd_kafka_abort_transaction() in rdkafka.h for more information.
+ */
+ virtual Error *abort_transaction(int timeout_ms) = 0;
+
+ /**@}*/
+};
+
+/**@}*/
+
+
+/**
+ * @name Metadata interface
+ * @{
+ *
+ */
+
+
+/**
+ * @brief Metadata: Broker information
+ */
+class BrokerMetadata {
+ public:
+ /** @returns Broker id */
+ virtual int32_t id() const = 0;
+
+ /** @returns Broker hostname */
+ virtual std::string host() const = 0;
+
+ /** @returns Broker listening port */
+ virtual int port() const = 0;
+
+ virtual ~BrokerMetadata() = 0;
+};
+
+
+
+/**
+ * @brief Metadata: Partition information
+ */
+class PartitionMetadata {
+ public:
+ /** @brief Replicas */
+ typedef std::vector<int32_t> ReplicasVector;
+ /** @brief ISRs (In-Sync-Replicas) */
+ typedef std::vector<int32_t> ISRSVector;
+
+ /** @brief Replicas iterator */
+ typedef ReplicasVector::const_iterator ReplicasIterator;
+ /** @brief ISRs iterator */
+ typedef ISRSVector::const_iterator ISRSIterator;
+
+
+ /** @returns Partition id */
+ virtual int32_t id() const = 0;
+
+ /** @returns Partition error reported by broker */
+ virtual ErrorCode err() const = 0;
+
+ /** @returns Leader broker (id) for partition */
+ virtual int32_t leader() const = 0;
+
+ /** @returns Replica brokers */
+ virtual const std::vector<int32_t> *replicas() const = 0;
+
+ /** @returns In-Sync-Replica brokers
+ * @warning The broker may return a cached/outdated list of ISRs.
+ */
+ virtual const std::vector<int32_t> *isrs() const = 0;
+
+ virtual ~PartitionMetadata() = 0;
+};
+
+
+
+/**
+ * @brief Metadata: Topic information
+ */
+class TopicMetadata {
+ public:
+ /** @brief Partitions */
+ typedef std::vector<const PartitionMetadata *> PartitionMetadataVector;
+ /** @brief Partitions iterator */
+ typedef PartitionMetadataVector::const_iterator PartitionMetadataIterator;
+
+ /** @returns Topic name */
+ virtual std::string topic() const = 0;
+
+ /** @returns Partition list */
+ virtual const PartitionMetadataVector *partitions() const = 0;
+
+ /** @returns Topic error reported by broker */
+ virtual ErrorCode err() const = 0;
+
+ virtual ~TopicMetadata() = 0;
+};
+
+
+/**
+ * @brief Metadata container
+ */
+class Metadata {
+ public:
+ /** @brief Brokers */
+ typedef std::vector<const BrokerMetadata *> BrokerMetadataVector;
+ /** @brief Topics */
+ typedef std::vector<const TopicMetadata *> TopicMetadataVector;
+
+ /** @brief Brokers iterator */
+ typedef BrokerMetadataVector::const_iterator BrokerMetadataIterator;
+ /** @brief Topics iterator */
+ typedef TopicMetadataVector::const_iterator TopicMetadataIterator;
+
+
+ /**
+ * @brief Broker list
+ * @remark Ownership of the returned pointer is retained by the instance of
+ * Metadata that is called.
+ */
+ virtual const BrokerMetadataVector *brokers() const = 0;
+
+ /**
+ * @brief Topic list
+ * @remark Ownership of the returned pointer is retained by the instance of
+ * Metadata that is called.
+ */
+ virtual const TopicMetadataVector *topics() const = 0;
+
+ /** @brief Broker (id) originating this metadata */
+ virtual int32_t orig_broker_id() const = 0;
+
+ /** @brief Broker (name) originating this metadata */
+ virtual std::string orig_broker_name() const = 0;
+
+ virtual ~Metadata() = 0;
+};
+
+/**@}*/
+
+} // namespace RdKafka
+
+
+#endif /* _RDKAFKACPP_H_ */
diff --git a/fluent-bit/lib/librdkafka-2.1.0/src-cpp/rdkafkacpp_int.h b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/rdkafkacpp_int.h
new file mode 100644
index 00000000..bc024ebe
--- /dev/null
+++ b/fluent-bit/lib/librdkafka-2.1.0/src-cpp/rdkafkacpp_int.h
@@ -0,0 +1,1628 @@
+/*
+ * librdkafka - Apache Kafka C/C++ library
+ *
+ * Copyright (c) 2014 Magnus Edenhill
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ * this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _RDKAFKACPP_INT_H_
+#define _RDKAFKACPP_INT_H_
+
+#include <string>
+#include <iostream>
+#include <cstring>
+#include <stdlib.h>
+
+#include "rdkafkacpp.h"
+
+extern "C" {
+#include "../src/rdkafka.h"
+}
+
+#ifdef _WIN32
+/* Visual Studio */
+#include "../src/win32_config.h"
+#else
+/* POSIX / UNIX based systems */
+#include "../config.h" /* mklove output */
+#endif
+
+#ifdef _MSC_VER
+typedef int mode_t;
+#pragma warning(disable : 4250)
+#endif
+
+
+namespace RdKafka {
+
+void consume_cb_trampoline(rd_kafka_message_t *msg, void *opaque);
+void log_cb_trampoline(const rd_kafka_t *rk,
+ int level,
+ const char *fac,
+ const char *buf);
+void error_cb_trampoline(rd_kafka_t *rk,
+ int err,
+ const char *reason,
+ void *opaque);
+void throttle_cb_trampoline(rd_kafka_t *rk,
+ const char *broker_name,
+ int32_t broker_id,
+ int throttle_time_ms,
+ void *opaque);
+int stats_cb_trampoline(rd_kafka_t *rk,
+ char *json,
+ size_t json_len,
+ void *opaque);
+int socket_cb_trampoline(int domain, int type, int protocol, void *opaque);
+int open_cb_trampoline(const char *pathname,
+ int flags,
+ mode_t mode,
+ void *opaque);
+void rebalance_cb_trampoline(rd_kafka_t *rk,
+ rd_kafka_resp_err_t err,
+ rd_kafka_topic_partition_list_t *c_partitions,
+ void *opaque);
+void offset_commit_cb_trampoline0(rd_kafka_t *rk,
+ rd_kafka_resp_err_t err,
+ rd_kafka_topic_partition_list_t *c_offsets,
+ void *opaque);
+void oauthbearer_token_refresh_cb_trampoline(rd_kafka_t *rk,
+ const char *oauthbearer_config,
+ void *opaque);
+
+int ssl_cert_verify_cb_trampoline(rd_kafka_t *rk,
+ const char *broker_name,
+ int32_t broker_id,
+ int *x509_error,
+ int depth,
+ const char *buf,
+ size_t size,
+ char *errstr,
+ size_t errstr_size,
+ void *opaque);
+
+rd_kafka_topic_partition_list_t *partitions_to_c_parts(
+ const std::vector<TopicPartition *> &partitions);
+
+/**
+ * @brief Update the application provided 'partitions' with info from 'c_parts'
+ */
+void update_partitions_from_c_parts(
+ std::vector<TopicPartition *> &partitions,
+ const rd_kafka_topic_partition_list_t *c_parts);
+
+
+class ErrorImpl : public Error {
+ public:
+ ~ErrorImpl() {
+ rd_kafka_error_destroy(c_error_);
+ }
+
+ ErrorImpl(ErrorCode code, const std::string *errstr) {
+ c_error_ = rd_kafka_error_new(static_cast<rd_kafka_resp_err_t>(code),
+ errstr ? "%s" : NULL,
+ errstr ? errstr->c_str() : NULL);
+ }
+
+ ErrorImpl(rd_kafka_error_t *c_error) : c_error_(c_error) {
+ }
+
+ static Error *create(ErrorCode code, const std::string *errstr) {
+ return new ErrorImpl(code, errstr);
+ }
+
+ ErrorCode code() const {
+ return static_cast<ErrorCode>(rd_kafka_error_code(c_error_));
+ }
+
+ std::string name() const {
+ return std::string(rd_kafka_error_name(c_error_));
+ }
+
+ std::string str() const {
+ return std::string(rd_kafka_error_string(c_error_));
+ }
+
+ bool is_fatal() const {
+ return !!rd_kafka_error_is_fatal(c_error_);
+ }
+
+ bool is_retriable() const {
+ return !!rd_kafka_error_is_retriable(c_error_);
+ }
+
+ bool txn_requires_abort() const {
+ return !!rd_kafka_error_txn_requires_abort(c_error_);
+ }
+
+ rd_kafka_error_t *c_error_;
+};
+
+
+class EventImpl : public Event {
+ public:
+ ~EventImpl() {
+ }
+
+ EventImpl(Type type,
+ ErrorCode err,
+ Severity severity,
+ const char *fac,
+ const char *str) :
+ type_(type),
+ err_(err),
+ severity_(severity),
+ fac_(fac ? fac : ""),
+ str_(str),
+ id_(0),
+ throttle_time_(0),
+ fatal_(false) {
+ }
+
+ EventImpl(Type type) :
+ type_(type),
+ err_(ERR_NO_ERROR),
+ severity_(EVENT_SEVERITY_EMERG),
+ fac_(""),
+ str_(""),
+ id_(0),
+ throttle_time_(0),
+ fatal_(false) {
+ }
+
+ Type type() const {
+ return type_;
+ }
+ ErrorCode err() const {
+ return err_;
+ }
+ Severity severity() const {
+ return severity_;
+ }
+ std::string fac() const {
+ return fac_;
+ }
+ std::string str() const {
+ return str_;
+ }
+ std::string broker_name() const {
+ if (type_ == EVENT_THROTTLE)
+ return str_;
+ else
+ return std::string("");
+ }
+ int broker_id() const {
+ return id_;
+ }
+ int throttle_time() const {
+ return throttle_time_;
+ }
+
+ bool fatal() const {
+ return fatal_;
+ }
+
+ Type type_;
+ ErrorCode err_;
+ Severity severity_;
+ std::string fac_;
+ std::string str_; /* reused for THROTTLE broker_name */
+ int id_;
+ int throttle_time_;
+ bool fatal_;
+};
+
+class QueueImpl : virtual public Queue {
+ public:
+ QueueImpl(rd_kafka_queue_t *c_rkqu) : queue_(c_rkqu) {
+ }
+ ~QueueImpl() {
+ rd_kafka_queue_destroy(queue_);
+ }
+ static Queue *create(Handle *base);
+ ErrorCode forward(Queue *queue);
+ Message *consume(int timeout_ms);
+ int poll(int timeout_ms);
+ void io_event_enable(int fd, const void *payload, size_t size);
+
+ rd_kafka_queue_t *queue_;
+};
+
+
+
+class HeadersImpl : public Headers {
+ public:
+ HeadersImpl() : headers_(rd_kafka_headers_new(8)) {
+ }
+
+ HeadersImpl(rd_kafka_headers_t *headers) : headers_(headers) {
+ }
+
+ HeadersImpl(const std::vector<Header> &headers) {
+ if (headers.size() > 0) {
+ headers_ = rd_kafka_headers_new(headers.size());
+ from_vector(headers);
+ } else {
+ headers_ = rd_kafka_headers_new(8);
+ }
+ }
+
+ ~HeadersImpl() {
+ if (headers_) {
+ rd_kafka_headers_destroy(headers_);
+ }
+ }
+
+ ErrorCode add(const std::string &key, const char *value) {
+ rd_kafka_resp_err_t err;
+ err = rd_kafka_header_add(headers_, key.c_str(), key.size(), value, -1);
+ return static_cast<RdKafka::ErrorCode>(err);
+ }
+
+ ErrorCode add(const std::string &key, const void *value, size_t value_size) {
+ rd_kafka_resp_err_t err;
+ err = rd_kafka_header_add(headers_, key.c_str(), key.size(), value,
+ value_size);
+ return static_cast<RdKafka::ErrorCode>(err);
+ }
+
+ ErrorCode add(const std::string &key, const std::string &value) {
+ rd_kafka_resp_err_t err;
+ err = rd_kafka_header_add(headers_, key.c_str(), key.size(), value.c_str(),
+ value.size());
+ return static_cast<RdKafka::ErrorCode>(err);
+ }
+
+ ErrorCode add(const Header &header) {
+ rd_kafka_resp_err_t err;
+ err =
+ rd_kafka_header_add(headers_, header.key().c_str(), header.key().size(),
+ header.value(), header.value_size());
+ return static_cast<RdKafka::ErrorCode>(err);
+ }
+
+ ErrorCode remove(const std::string &key) {
+ rd_kafka_resp_err_t err;
+ err = rd_kafka_header_remove(headers_, key.c_str());
+ return static_cast<RdKafka::ErrorCode>(err);
+ }
+
+ std::vector<Headers::Header> get(const std::string &key) const {
+ std::vector<Headers::Header> headers;
+ const void *value;
+ size_t size;
+ rd_kafka_resp_err_t err;
+ for (size_t idx = 0; !(err = rd_kafka_header_get(headers_, idx, key.c_str(),
+ &value, &size));
+ idx++) {
+ headers.push_back(Headers::Header(key, value, size));
+ }
+ return headers;
+ }
+
+ Headers::Header get_last(const std::string &key) const {
+ const void *value;
+ size_t size;
+ rd_kafka_resp_err_t err;
+ err = rd_kafka_header_get_last(headers_, key.c_str(), &value, &size);
+ return Headers::Header(key, value, size,
+ static_cast<RdKafka::ErrorCode>(err));
+ }
+
+ std::vector<Headers::Header> get_all() const {
+ std::vector<Headers::Header> headers;
+ size_t idx = 0;
+ const char *name;
+ const void *valuep;
+ size_t size;
+ while (!rd_kafka_header_get_all(headers_, idx++, &name, &valuep, &size)) {
+ headers.push_back(Headers::Header(name, valuep, size));
+ }
+ return headers;
+ }
+
+ size_t size() const {
+ return rd_kafka_header_cnt(headers_);
+ }
+
+ /** @brief Reset the C headers pointer to NULL. */
+ void c_headers_destroyed() {
+ headers_ = NULL;
+ }
+
+ /** @returns the underlying C headers, or NULL. */
+ rd_kafka_headers_t *c_ptr() {
+ return headers_;
+ }
+
+
+ private:
+ void from_vector(const std::vector<Header> &headers) {
+ if (headers.size() == 0)
+ return;
+ for (std::vector<Header>::const_iterator it = headers.begin();
+ it != headers.end(); it++)
+ this->add(*it);
+ }
+
+ HeadersImpl(HeadersImpl const &) /*= delete*/;
+ HeadersImpl &operator=(HeadersImpl const &) /*= delete*/;
+
+ rd_kafka_headers_t *headers_;
+};
+
+
+
+class MessageImpl : public Message {
+ public:
+ ~MessageImpl() {
+ if (free_rkmessage_)
+ rd_kafka_message_destroy(const_cast<rd_kafka_message_t *>(rkmessage_));
+ if (key_)
+ delete key_;
+ if (headers_)
+ delete headers_;
+ }
+
+ MessageImpl(rd_kafka_type_t rk_type,
+ RdKafka::Topic *topic,
+ rd_kafka_message_t *rkmessage) :
+ topic_(topic),
+ rkmessage_(rkmessage),
+ free_rkmessage_(true),
+ key_(NULL),
+ headers_(NULL),
+ rk_type_(rk_type) {
+ }
+
+ MessageImpl(rd_kafka_type_t rk_type,
+ RdKafka::Topic *topic,
+ rd_kafka_message_t *rkmessage,
+ bool dofree) :
+ topic_(topic),
+ rkmessage_(rkmessage),
+ free_rkmessage_(dofree),
+ key_(NULL),
+ headers_(NULL),
+ rk_type_(rk_type) {
+ }
+
+ MessageImpl(rd_kafka_type_t rk_type, rd_kafka_message_t *rkmessage) :
+ topic_(NULL),
+ rkmessage_(rkmessage),
+ free_rkmessage_(true),
+ key_(NULL),
+ headers_(NULL),
+ rk_type_(rk_type) {
+ if (rkmessage->rkt) {
+ /* Possibly NULL */
+ topic_ = static_cast<Topic *>(rd_kafka_topic_opaque(rkmessage->rkt));
+ }
+ }
+
+ /* Create errored message */
+ MessageImpl(rd_kafka_type_t rk_type,
+ RdKafka::Topic *topic,
+ RdKafka::ErrorCode err) :
+ topic_(topic),
+ free_rkmessage_(false),
+ key_(NULL),
+ headers_(NULL),
+ rk_type_(rk_type) {
+ rkmessage_ = &rkmessage_err_;
+ memset(&rkmessage_err_, 0, sizeof(rkmessage_err_));
+ rkmessage_err_.err = static_cast<rd_kafka_resp_err_t>(err);
+ }
+
+ std::string errstr() const {
+ const char *es;
+ /* message_errstr() is only available for the consumer. */
+ if (rk_type_ == RD_KAFKA_CONSUMER)
+ es = rd_kafka_message_errstr(rkmessage_);
+ else
+ es = rd_kafka_err2str(rkmessage_->err);
+
+ return std::string(es ? es : "");
+ }
+
+ ErrorCode err() const {
+ return static_cast<RdKafka::ErrorCode>(rkmessage_->err);
+ }
+
+ Topic *topic() const {
+ return topic_;
+ }
+ std::string topic_name() const {
+ if (rkmessage_->rkt)
+ return rd_kafka_topic_name(rkmessage_->rkt);
+ else
+ return "";
+ }
+ int32_t partition() const {
+ return rkmessage_->partition;
+ }
+ void *payload() const {
+ return rkmessage_->payload;
+ }
+ size_t len() const {
+ return rkmessage_->len;
+ }
+ const std::string *key() const {
+ if (key_) {
+ return key_;
+ } else if (rkmessage_->key) {
+ key_ = new std::string(static_cast<char const *>(rkmessage_->key),
+ rkmessage_->key_len);
+ return key_;
+ }
+ return NULL;
+ }
+ const void *key_pointer() const {
+ return rkmessage_->key;
+ }
+ size_t key_len() const {
+ return rkmessage_->key_len;
+ }
+
+ int64_t offset() const {
+ return rkmessage_->offset;
+ }
+
+ MessageTimestamp timestamp() const {
+ MessageTimestamp ts;
+ rd_kafka_timestamp_type_t tstype;
+ ts.timestamp = rd_kafka_message_timestamp(rkmessage_, &tstype);
+ ts.type = static_cast<MessageTimestamp::MessageTimestampType>(tstype);
+ return ts;
+ }
+
+ void *msg_opaque() const {
+ return rkmessage_->_private;
+ }
+
+ int64_t latency() const {
+ return rd_kafka_message_latency(rkmessage_);
+ }
+
+ struct rd_kafka_message_s *c_ptr() {
+ return rkmessage_;
+ }
+
+ Status status() const {
+ return static_cast<Status>(rd_kafka_message_status(rkmessage_));
+ }
+
+ Headers *headers() {
+ ErrorCode err;
+ return headers(&err);
+ }
+
+ Headers *headers(ErrorCode *err) {
+ *err = ERR_NO_ERROR;
+
+ if (!headers_) {
+ rd_kafka_headers_t *c_hdrs;
+ rd_kafka_resp_err_t c_err;
+
+ if ((c_err = rd_kafka_message_detach_headers(rkmessage_, &c_hdrs))) {
+ *err = static_cast<RdKafka::ErrorCode>(c_err);
+ return NULL;
+ }
+
+ headers_ = new HeadersImpl(c_hdrs);
+ }
+
+ return headers_;
+ }
+
+ int32_t broker_id() const {
+ return rd_kafka_message_broker_id(rkmessage_);
+ }
+
+ int32_t leader_epoch() const {
+ return rd_kafka_message_leader_epoch(rkmessage_);
+ }
+
+
+ Error *offset_store() {
+ rd_kafka_error_t *c_error;
+
+ c_error = rd_kafka_offset_store_message(rkmessage_);
+
+ if (c_error)
+ return new ErrorImpl(c_error);
+ else
+ return NULL;
+ }
+
+ RdKafka::Topic *topic_;
+ rd_kafka_message_t *rkmessage_;
+ bool free_rkmessage_;
+ /* For error signalling by the C++ layer the .._err_ message is
+ * used as a place holder and rkmessage_ is set to point to it. */
+ rd_kafka_message_t rkmessage_err_;
+ mutable std::string *key_; /* mutable because it's a cached value */
+
+ private:
+ /* "delete" copy ctor + copy assignment, for safety of key_ */
+ MessageImpl(MessageImpl const &) /*= delete*/;
+ MessageImpl &operator=(MessageImpl const &) /*= delete*/;
+
+ RdKafka::Headers *headers_;
+ const rd_kafka_type_t rk_type_; /**< Client type */
+};
+
+
+class ConfImpl : public Conf {
+ public:
+ ConfImpl(ConfType conf_type) :
+ consume_cb_(NULL),
+ dr_cb_(NULL),
+ event_cb_(NULL),
+ socket_cb_(NULL),
+ open_cb_(NULL),
+ partitioner_cb_(NULL),
+ partitioner_kp_cb_(NULL),
+ rebalance_cb_(NULL),
+ offset_commit_cb_(NULL),
+ oauthbearer_token_refresh_cb_(NULL),
+ ssl_cert_verify_cb_(NULL),
+ conf_type_(conf_type),
+ rk_conf_(NULL),
+ rkt_conf_(NULL) {
+ }
+ ~ConfImpl() {
+ if (rk_conf_)
+ rd_kafka_conf_destroy(rk_conf_);
+ else if (rkt_conf_)
+ rd_kafka_topic_conf_destroy(rkt_conf_);
+ }
+
+ Conf::ConfResult set(const std::string &name,
+ const std::string &value,
+ std::string &errstr);
+
+ Conf::ConfResult set(const std::string &name,
+ DeliveryReportCb *dr_cb,
+ std::string &errstr) {
+ if (name != "dr_cb") {
+ errstr = "Invalid value type, expected RdKafka::DeliveryReportCb";
+ return Conf::CONF_INVALID;
+ }
+
+ if (!rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ return Conf::CONF_INVALID;
+ }
+
+ dr_cb_ = dr_cb;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult set(const std::string &name,
+ OAuthBearerTokenRefreshCb *oauthbearer_token_refresh_cb,
+ std::string &errstr) {
+ if (name != "oauthbearer_token_refresh_cb") {
+ errstr =
+ "Invalid value type, expected RdKafka::OAuthBearerTokenRefreshCb";
+ return Conf::CONF_INVALID;
+ }
+
+ if (!rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ return Conf::CONF_INVALID;
+ }
+
+ oauthbearer_token_refresh_cb_ = oauthbearer_token_refresh_cb;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult set(const std::string &name,
+ EventCb *event_cb,
+ std::string &errstr) {
+ if (name != "event_cb") {
+ errstr = "Invalid value type, expected RdKafka::EventCb";
+ return Conf::CONF_INVALID;
+ }
+
+ if (!rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ return Conf::CONF_INVALID;
+ }
+
+ event_cb_ = event_cb;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult set(const std::string &name,
+ const Conf *topic_conf,
+ std::string &errstr) {
+ const ConfImpl *tconf_impl =
+ dynamic_cast<const RdKafka::ConfImpl *>(topic_conf);
+ if (name != "default_topic_conf" || !tconf_impl->rkt_conf_) {
+ errstr = "Invalid value type, expected RdKafka::Conf";
+ return Conf::CONF_INVALID;
+ }
+
+ if (!rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ return Conf::CONF_INVALID;
+ }
+
+ rd_kafka_conf_set_default_topic_conf(
+ rk_conf_, rd_kafka_topic_conf_dup(tconf_impl->rkt_conf_));
+
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult set(const std::string &name,
+ PartitionerCb *partitioner_cb,
+ std::string &errstr) {
+ if (name != "partitioner_cb") {
+ errstr = "Invalid value type, expected RdKafka::PartitionerCb";
+ return Conf::CONF_INVALID;
+ }
+
+ if (!rkt_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_TOPIC object";
+ return Conf::CONF_INVALID;
+ }
+
+ partitioner_cb_ = partitioner_cb;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult set(const std::string &name,
+ PartitionerKeyPointerCb *partitioner_kp_cb,
+ std::string &errstr) {
+ if (name != "partitioner_key_pointer_cb") {
+ errstr = "Invalid value type, expected RdKafka::PartitionerKeyPointerCb";
+ return Conf::CONF_INVALID;
+ }
+
+ if (!rkt_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_TOPIC object";
+ return Conf::CONF_INVALID;
+ }
+
+ partitioner_kp_cb_ = partitioner_kp_cb;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult set(const std::string &name,
+ SocketCb *socket_cb,
+ std::string &errstr) {
+ if (name != "socket_cb") {
+ errstr = "Invalid value type, expected RdKafka::SocketCb";
+ return Conf::CONF_INVALID;
+ }
+
+ if (!rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ return Conf::CONF_INVALID;
+ }
+
+ socket_cb_ = socket_cb;
+ return Conf::CONF_OK;
+ }
+
+
+ Conf::ConfResult set(const std::string &name,
+ OpenCb *open_cb,
+ std::string &errstr) {
+ if (name != "open_cb") {
+ errstr = "Invalid value type, expected RdKafka::OpenCb";
+ return Conf::CONF_INVALID;
+ }
+
+ if (!rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ return Conf::CONF_INVALID;
+ }
+
+ open_cb_ = open_cb;
+ return Conf::CONF_OK;
+ }
+
+
+
+ Conf::ConfResult set(const std::string &name,
+ RebalanceCb *rebalance_cb,
+ std::string &errstr) {
+ if (name != "rebalance_cb") {
+ errstr = "Invalid value type, expected RdKafka::RebalanceCb";
+ return Conf::CONF_INVALID;
+ }
+
+ if (!rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ return Conf::CONF_INVALID;
+ }
+
+ rebalance_cb_ = rebalance_cb;
+ return Conf::CONF_OK;
+ }
+
+
+ Conf::ConfResult set(const std::string &name,
+ OffsetCommitCb *offset_commit_cb,
+ std::string &errstr) {
+ if (name != "offset_commit_cb") {
+ errstr = "Invalid value type, expected RdKafka::OffsetCommitCb";
+ return Conf::CONF_INVALID;
+ }
+
+ if (!rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ return Conf::CONF_INVALID;
+ }
+
+ offset_commit_cb_ = offset_commit_cb;
+ return Conf::CONF_OK;
+ }
+
+
+ Conf::ConfResult set(const std::string &name,
+ SslCertificateVerifyCb *ssl_cert_verify_cb,
+ std::string &errstr) {
+ if (name != "ssl_cert_verify_cb") {
+ errstr = "Invalid value type, expected RdKafka::SslCertificateVerifyCb";
+ return Conf::CONF_INVALID;
+ }
+
+ if (!rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ return Conf::CONF_INVALID;
+ }
+
+ ssl_cert_verify_cb_ = ssl_cert_verify_cb;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult set_engine_callback_data(void *value, std::string &errstr) {
+ if (!rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ return Conf::CONF_INVALID;
+ }
+
+ rd_kafka_conf_set_engine_callback_data(rk_conf_, value);
+ return Conf::CONF_OK;
+ }
+
+
+ Conf::ConfResult set_ssl_cert(RdKafka::CertificateType cert_type,
+ RdKafka::CertificateEncoding cert_enc,
+ const void *buffer,
+ size_t size,
+ std::string &errstr) {
+ rd_kafka_conf_res_t res;
+ char errbuf[512];
+
+ if (!rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ return Conf::CONF_INVALID;
+ }
+
+ res = rd_kafka_conf_set_ssl_cert(
+ rk_conf_, static_cast<rd_kafka_cert_type_t>(cert_type),
+ static_cast<rd_kafka_cert_enc_t>(cert_enc), buffer, size, errbuf,
+ sizeof(errbuf));
+
+ if (res != RD_KAFKA_CONF_OK)
+ errstr = errbuf;
+
+ return static_cast<Conf::ConfResult>(res);
+ }
+
+ Conf::ConfResult enable_sasl_queue(bool enable, std::string &errstr) {
+ if (!rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ return Conf::CONF_INVALID;
+ }
+
+ rd_kafka_conf_enable_sasl_queue(rk_conf_, enable ? 1 : 0);
+
+ return Conf::CONF_OK;
+ }
+
+
+ Conf::ConfResult get(const std::string &name, std::string &value) const {
+ if (name.compare("dr_cb") == 0 || name.compare("event_cb") == 0 ||
+ name.compare("partitioner_cb") == 0 ||
+ name.compare("partitioner_key_pointer_cb") == 0 ||
+ name.compare("socket_cb") == 0 || name.compare("open_cb") == 0 ||
+ name.compare("rebalance_cb") == 0 ||
+ name.compare("offset_commit_cb") == 0 ||
+ name.compare("oauthbearer_token_refresh_cb") == 0 ||
+ name.compare("ssl_cert_verify_cb") == 0 ||
+ name.compare("set_engine_callback_data") == 0 ||
+ name.compare("enable_sasl_queue") == 0) {
+ return Conf::CONF_INVALID;
+ }
+ rd_kafka_conf_res_t res = RD_KAFKA_CONF_INVALID;
+
+ /* Get size of property */
+ size_t size;
+ if (rk_conf_)
+ res = rd_kafka_conf_get(rk_conf_, name.c_str(), NULL, &size);
+ else if (rkt_conf_)
+ res = rd_kafka_topic_conf_get(rkt_conf_, name.c_str(), NULL, &size);
+ if (res != RD_KAFKA_CONF_OK)
+ return static_cast<Conf::ConfResult>(res);
+
+ char *tmpValue = new char[size];
+
+ if (rk_conf_)
+ res = rd_kafka_conf_get(rk_conf_, name.c_str(), tmpValue, &size);
+ else if (rkt_conf_)
+ res = rd_kafka_topic_conf_get(rkt_conf_, name.c_str(), tmpValue, &size);
+
+ if (res == RD_KAFKA_CONF_OK)
+ value.assign(tmpValue);
+ delete[] tmpValue;
+
+ return static_cast<Conf::ConfResult>(res);
+ }
+
+ Conf::ConfResult get(DeliveryReportCb *&dr_cb) const {
+ if (!rk_conf_)
+ return Conf::CONF_INVALID;
+ dr_cb = this->dr_cb_;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult get(
+ OAuthBearerTokenRefreshCb *&oauthbearer_token_refresh_cb) const {
+ if (!rk_conf_)
+ return Conf::CONF_INVALID;
+ oauthbearer_token_refresh_cb = this->oauthbearer_token_refresh_cb_;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult get(EventCb *&event_cb) const {
+ if (!rk_conf_)
+ return Conf::CONF_INVALID;
+ event_cb = this->event_cb_;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult get(PartitionerCb *&partitioner_cb) const {
+ if (!rkt_conf_)
+ return Conf::CONF_INVALID;
+ partitioner_cb = this->partitioner_cb_;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult get(PartitionerKeyPointerCb *&partitioner_kp_cb) const {
+ if (!rkt_conf_)
+ return Conf::CONF_INVALID;
+ partitioner_kp_cb = this->partitioner_kp_cb_;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult get(SocketCb *&socket_cb) const {
+ if (!rk_conf_)
+ return Conf::CONF_INVALID;
+ socket_cb = this->socket_cb_;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult get(OpenCb *&open_cb) const {
+ if (!rk_conf_)
+ return Conf::CONF_INVALID;
+ open_cb = this->open_cb_;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult get(RebalanceCb *&rebalance_cb) const {
+ if (!rk_conf_)
+ return Conf::CONF_INVALID;
+ rebalance_cb = this->rebalance_cb_;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult get(OffsetCommitCb *&offset_commit_cb) const {
+ if (!rk_conf_)
+ return Conf::CONF_INVALID;
+ offset_commit_cb = this->offset_commit_cb_;
+ return Conf::CONF_OK;
+ }
+
+ Conf::ConfResult get(SslCertificateVerifyCb *&ssl_cert_verify_cb) const {
+ if (!rk_conf_)
+ return Conf::CONF_INVALID;
+ ssl_cert_verify_cb = this->ssl_cert_verify_cb_;
+ return Conf::CONF_OK;
+ }
+
+ std::list<std::string> *dump();
+
+
+ Conf::ConfResult set(const std::string &name,
+ ConsumeCb *consume_cb,
+ std::string &errstr) {
+ if (name != "consume_cb") {
+ errstr = "Invalid value type, expected RdKafka::ConsumeCb";
+ return Conf::CONF_INVALID;
+ }
+
+ if (!rk_conf_) {
+ errstr = "Requires RdKafka::Conf::CONF_GLOBAL object";
+ return Conf::CONF_INVALID;
+ }
+
+ consume_cb_ = consume_cb;
+ return Conf::CONF_OK;
+ }
+
+ struct rd_kafka_conf_s *c_ptr_global() {
+ if (conf_type_ == CONF_GLOBAL)
+ return rk_conf_;
+ else
+ return NULL;
+ }
+
+ struct rd_kafka_topic_conf_s *c_ptr_topic() {
+ if (conf_type_ == CONF_TOPIC)
+ return rkt_conf_;
+ else
+ return NULL;
+ }
+
+ ConsumeCb *consume_cb_;
+ DeliveryReportCb *dr_cb_;
+ EventCb *event_cb_;
+ SocketCb *socket_cb_;
+ OpenCb *open_cb_;
+ PartitionerCb *partitioner_cb_;
+ PartitionerKeyPointerCb *partitioner_kp_cb_;
+ RebalanceCb *rebalance_cb_;
+ OffsetCommitCb *offset_commit_cb_;
+ OAuthBearerTokenRefreshCb *oauthbearer_token_refresh_cb_;
+ SslCertificateVerifyCb *ssl_cert_verify_cb_;
+ ConfType conf_type_;
+ rd_kafka_conf_t *rk_conf_;
+ rd_kafka_topic_conf_t *rkt_conf_;
+};
+
+
+class HandleImpl : virtual public Handle {
+ public:
+ ~HandleImpl() {
+ }
+ HandleImpl() {
+ }
+ std::string name() const {
+ return std::string(rd_kafka_name(rk_));
+ }
+ std::string memberid() const {
+ char *str = rd_kafka_memberid(rk_);
+ std::string memberid = str ? str : "";
+ if (str)
+ rd_kafka_mem_free(rk_, str);
+ return memberid;
+ }
+ int poll(int timeout_ms) {
+ return rd_kafka_poll(rk_, timeout_ms);
+ }
+ int outq_len() {
+ return rd_kafka_outq_len(rk_);
+ }
+
+ void set_common_config(const RdKafka::ConfImpl *confimpl);
+
+ RdKafka::ErrorCode metadata(bool all_topics,
+ const Topic *only_rkt,
+ Metadata **metadatap,
+ int timeout_ms);
+
+ ErrorCode pause(std::vector<TopicPartition *> &partitions);
+ ErrorCode resume(std::vector<TopicPartition *> &partitions);
+
+ ErrorCode query_watermark_offsets(const std::string &topic,
+ int32_t partition,
+ int64_t *low,
+ int64_t *high,
+ int timeout_ms) {
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_query_watermark_offsets(
+ rk_, topic.c_str(), partition, low, high, timeout_ms));
+ }
+
+ ErrorCode get_watermark_offsets(const std::string &topic,
+ int32_t partition,
+ int64_t *low,
+ int64_t *high) {
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_get_watermark_offsets(
+ rk_, topic.c_str(), partition, low, high));
+ }
+
+ Queue *get_partition_queue(const TopicPartition *partition);
+
+ Queue *get_sasl_queue() {
+ rd_kafka_queue_t *rkqu;
+ rkqu = rd_kafka_queue_get_sasl(rk_);
+
+ if (rkqu == NULL)
+ return NULL;
+
+ return new QueueImpl(rkqu);
+ }
+
+ Queue *get_background_queue() {
+ rd_kafka_queue_t *rkqu;
+ rkqu = rd_kafka_queue_get_background(rk_);
+
+ if (rkqu == NULL)
+ return NULL;
+
+ return new QueueImpl(rkqu);
+ }
+
+
+ ErrorCode offsetsForTimes(std::vector<TopicPartition *> &offsets,
+ int timeout_ms) {
+ rd_kafka_topic_partition_list_t *c_offsets = partitions_to_c_parts(offsets);
+ ErrorCode err = static_cast<ErrorCode>(
+ rd_kafka_offsets_for_times(rk_, c_offsets, timeout_ms));
+ update_partitions_from_c_parts(offsets, c_offsets);
+ rd_kafka_topic_partition_list_destroy(c_offsets);
+ return err;
+ }
+
+ ErrorCode set_log_queue(Queue *queue);
+
+ void yield() {
+ rd_kafka_yield(rk_);
+ }
+
+ std::string clusterid(int timeout_ms) {
+ char *str = rd_kafka_clusterid(rk_, timeout_ms);
+ std::string clusterid = str ? str : "";
+ if (str)
+ rd_kafka_mem_free(rk_, str);
+ return clusterid;
+ }
+
+ struct rd_kafka_s *c_ptr() {
+ return rk_;
+ }
+
+ int32_t controllerid(int timeout_ms) {
+ return rd_kafka_controllerid(rk_, timeout_ms);
+ }
+
+ ErrorCode fatal_error(std::string &errstr) const {
+ char errbuf[512];
+ RdKafka::ErrorCode err = static_cast<RdKafka::ErrorCode>(
+ rd_kafka_fatal_error(rk_, errbuf, sizeof(errbuf)));
+ if (err)
+ errstr = errbuf;
+ return err;
+ }
+
+ ErrorCode oauthbearer_set_token(const std::string &token_value,
+ int64_t md_lifetime_ms,
+ const std::string &md_principal_name,
+ const std::list<std::string> &extensions,
+ std::string &errstr) {
+ char errbuf[512];
+ ErrorCode err;
+ const char **extensions_copy = new const char *[extensions.size()];
+ int elem = 0;
+
+ for (std::list<std::string>::const_iterator it = extensions.begin();
+ it != extensions.end(); it++)
+ extensions_copy[elem++] = it->c_str();
+ err = static_cast<ErrorCode>(rd_kafka_oauthbearer_set_token(
+ rk_, token_value.c_str(), md_lifetime_ms, md_principal_name.c_str(),
+ extensions_copy, extensions.size(), errbuf, sizeof(errbuf)));
+ delete[] extensions_copy;
+
+ if (err != ERR_NO_ERROR)
+ errstr = errbuf;
+
+ return err;
+ }
+
+ ErrorCode oauthbearer_set_token_failure(const std::string &errstr) {
+ return static_cast<ErrorCode>(
+ rd_kafka_oauthbearer_set_token_failure(rk_, errstr.c_str()));
+ }
+
+ Error *sasl_background_callbacks_enable() {
+ rd_kafka_error_t *c_error = rd_kafka_sasl_background_callbacks_enable(rk_);
+
+ if (c_error)
+ return new ErrorImpl(c_error);
+
+ return NULL;
+ }
+
+ Error *sasl_set_credentials(const std::string &username,
+ const std::string &password) {
+ rd_kafka_error_t *c_error =
+ rd_kafka_sasl_set_credentials(rk_, username.c_str(), password.c_str());
+
+ if (c_error)
+ return new ErrorImpl(c_error);
+
+ return NULL;
+ };
+
+ void *mem_malloc(size_t size) {
+ return rd_kafka_mem_malloc(rk_, size);
+ }
+
+ void mem_free(void *ptr) {
+ rd_kafka_mem_free(rk_, ptr);
+ }
+
+ rd_kafka_t *rk_;
+ /* All Producer and Consumer callbacks must reside in HandleImpl and
+ * the opaque provided to rdkafka must be a pointer to HandleImpl, since
+ * ProducerImpl and ConsumerImpl classes cannot be safely directly cast to
+ * HandleImpl due to the skewed diamond inheritance. */
+ ConsumeCb *consume_cb_;
+ EventCb *event_cb_;
+ SocketCb *socket_cb_;
+ OpenCb *open_cb_;
+ DeliveryReportCb *dr_cb_;
+ PartitionerCb *partitioner_cb_;
+ PartitionerKeyPointerCb *partitioner_kp_cb_;
+ RebalanceCb *rebalance_cb_;
+ OffsetCommitCb *offset_commit_cb_;
+ OAuthBearerTokenRefreshCb *oauthbearer_token_refresh_cb_;
+ SslCertificateVerifyCb *ssl_cert_verify_cb_;
+};
+
+
+class TopicImpl : public Topic {
+ public:
+ ~TopicImpl() {
+ rd_kafka_topic_destroy(rkt_);
+ }
+
+ std::string name() const {
+ return rd_kafka_topic_name(rkt_);
+ }
+
+ bool partition_available(int32_t partition) const {
+ return !!rd_kafka_topic_partition_available(rkt_, partition);
+ }
+
+ ErrorCode offset_store(int32_t partition, int64_t offset) {
+ return static_cast<RdKafka::ErrorCode>(
+ rd_kafka_offset_store(rkt_, partition, offset));
+ }
+
+ static Topic *create(Handle &base, const std::string &topic, Conf *conf);
+
+ struct rd_kafka_topic_s *c_ptr() {
+ return rkt_;
+ }
+
+ rd_kafka_topic_t *rkt_;
+ PartitionerCb *partitioner_cb_;
+ PartitionerKeyPointerCb *partitioner_kp_cb_;
+};
+
+
+/**
+ * Topic and Partition
+ */
+class TopicPartitionImpl : public TopicPartition {
+ public:
+ ~TopicPartitionImpl() {
+ }
+
+ static TopicPartition *create(const std::string &topic, int partition);
+
+ TopicPartitionImpl(const std::string &topic, int partition) :
+ topic_(topic),
+ partition_(partition),
+ offset_(RdKafka::Topic::OFFSET_INVALID),
+ err_(ERR_NO_ERROR),
+ leader_epoch_(-1) {
+ }
+
+ TopicPartitionImpl(const std::string &topic, int partition, int64_t offset) :
+ topic_(topic),
+ partition_(partition),
+ offset_(offset),
+ err_(ERR_NO_ERROR),
+ leader_epoch_(-1) {
+ }
+
+ TopicPartitionImpl(const rd_kafka_topic_partition_t *c_part) {
+ topic_ = std::string(c_part->topic);
+ partition_ = c_part->partition;
+ offset_ = c_part->offset;
+ err_ = static_cast<ErrorCode>(c_part->err);
+ leader_epoch_ = rd_kafka_topic_partition_get_leader_epoch(c_part);
+ // FIXME: metadata
+ }
+
+ static void destroy(std::vector<TopicPartition *> &partitions);
+
+ int partition() const {
+ return partition_;
+ }
+ const std::string &topic() const {
+ return topic_;
+ }
+
+ int64_t offset() const {
+ return offset_;
+ }
+
+ ErrorCode err() const {
+ return err_;
+ }
+
+ void set_offset(int64_t offset) {
+ offset_ = offset;
+ }
+
+ int32_t get_leader_epoch() {
+ return leader_epoch_;
+ }
+
+ void set_leader_epoch(int32_t leader_epoch) {
+ leader_epoch_ = leader_epoch_;
+ }
+
+ std::ostream &operator<<(std::ostream &ostrm) const {
+ return ostrm << topic_ << " [" << partition_ << "]";
+ }
+
+ std::string topic_;
+ int partition_;
+ int64_t offset_;
+ ErrorCode err_;
+ int32_t leader_epoch_;
+};
+
+
+/**
+ * @class ConsumerGroupMetadata wraps the
+ * C rd_kafka_consumer_group_metadata_t object.
+ */
+class ConsumerGroupMetadataImpl : public ConsumerGroupMetadata {
+ public:
+ ~ConsumerGroupMetadataImpl() {
+ rd_kafka_consumer_group_metadata_destroy(cgmetadata_);
+ }
+
+ ConsumerGroupMetadataImpl(rd_kafka_consumer_group_metadata_t *cgmetadata) :
+ cgmetadata_(cgmetadata) {
+ }
+
+ rd_kafka_consumer_group_metadata_t *cgmetadata_;
+};
+
+
+class KafkaConsumerImpl : virtual public KafkaConsumer,
+ virtual public HandleImpl {
+ public:
+ ~KafkaConsumerImpl() {
+ if (rk_)
+ rd_kafka_destroy_flags(rk_, RD_KAFKA_DESTROY_F_NO_CONSUMER_CLOSE);
+ }
+
+ static KafkaConsumer *create(Conf *conf, std::string &errstr);
+
+ ErrorCode assignment(std::vector<TopicPartition *> &partitions);
+ bool assignment_lost();
+ std::string rebalance_protocol() {
+ const char *str = rd_kafka_rebalance_protocol(rk_);
+ return std::string(str ? str : "");
+ }
+ ErrorCode subscription(std::vector<std::string> &topics);
+ ErrorCode subscribe(const std::vector<std::string> &topics);
+ ErrorCode unsubscribe();
+ ErrorCode assign(const std::vector<TopicPartition *> &partitions);
+ ErrorCode unassign();
+ Error *incremental_assign(const std::vector<TopicPartition *> &partitions);
+ Error *incremental_unassign(const std::vector<TopicPartition *> &partitions);
+
+ Message *consume(int timeout_ms);
+ ErrorCode commitSync() {
+ return static_cast<ErrorCode>(rd_kafka_commit(rk_, NULL, 0 /*sync*/));
+ }
+ ErrorCode commitAsync() {
+ return static_cast<ErrorCode>(rd_kafka_commit(rk_, NULL, 1 /*async*/));
+ }
+ ErrorCode commitSync(Message *message) {
+ MessageImpl *msgimpl = dynamic_cast<MessageImpl *>(message);
+ return static_cast<ErrorCode>(
+ rd_kafka_commit_message(rk_, msgimpl->rkmessage_, 0 /*sync*/));
+ }
+ ErrorCode commitAsync(Message *message) {
+ MessageImpl *msgimpl = dynamic_cast<MessageImpl *>(message);
+ return static_cast<ErrorCode>(
+ rd_kafka_commit_message(rk_, msgimpl->rkmessage_, 1 /*async*/));
+ }
+
+ ErrorCode commitSync(std::vector<TopicPartition *> &offsets) {
+ rd_kafka_topic_partition_list_t *c_parts = partitions_to_c_parts(offsets);
+ rd_kafka_resp_err_t err = rd_kafka_commit(rk_, c_parts, 0);
+ if (!err)
+ update_partitions_from_c_parts(offsets, c_parts);
+ rd_kafka_topic_partition_list_destroy(c_parts);
+ return static_cast<ErrorCode>(err);
+ }
+
+ ErrorCode commitAsync(const std::vector<TopicPartition *> &offsets) {
+ rd_kafka_topic_partition_list_t *c_parts = partitions_to_c_parts(offsets);
+ rd_kafka_resp_err_t err = rd_kafka_commit(rk_, c_parts, 1);
+ rd_kafka_topic_partition_list_destroy(c_parts);
+ return static_cast<ErrorCode>(err);
+ }
+
+ ErrorCode commitSync(OffsetCommitCb *offset_commit_cb) {
+ return static_cast<ErrorCode>(rd_kafka_commit_queue(
+ rk_, NULL, NULL, RdKafka::offset_commit_cb_trampoline0,
+ offset_commit_cb));
+ }
+
+ ErrorCode commitSync(std::vector<TopicPartition *> &offsets,
+ OffsetCommitCb *offset_commit_cb) {
+ rd_kafka_topic_partition_list_t *c_parts = partitions_to_c_parts(offsets);
+ rd_kafka_resp_err_t err = rd_kafka_commit_queue(
+ rk_, c_parts, NULL, RdKafka::offset_commit_cb_trampoline0,
+ offset_commit_cb);
+ rd_kafka_topic_partition_list_destroy(c_parts);
+ return static_cast<ErrorCode>(err);
+ }
+
+ ErrorCode committed(std::vector<TopicPartition *> &partitions,
+ int timeout_ms);
+ ErrorCode position(std::vector<TopicPartition *> &partitions);
+
+ ConsumerGroupMetadata *groupMetadata() {
+ rd_kafka_consumer_group_metadata_t *cgmetadata;
+
+ cgmetadata = rd_kafka_consumer_group_metadata(rk_);
+ if (!cgmetadata)
+ return NULL;
+
+ return new ConsumerGroupMetadataImpl(cgmetadata);
+ }
+
+ ErrorCode close();
+
+ Error *close(Queue *queue);
+
+ bool closed() {
+ return rd_kafka_consumer_closed(rk_) ? true : false;
+ }
+
+ ErrorCode seek(const TopicPartition &partition, int timeout_ms);
+
+ ErrorCode offsets_store(std::vector<TopicPartition *> &offsets) {
+ rd_kafka_topic_partition_list_t *c_parts = partitions_to_c_parts(offsets);
+ rd_kafka_resp_err_t err = rd_kafka_offsets_store(rk_, c_parts);
+ update_partitions_from_c_parts(offsets, c_parts);
+ rd_kafka_topic_partition_list_destroy(c_parts);
+ return static_cast<ErrorCode>(err);
+ }
+};
+
+
+class MetadataImpl : public Metadata {
+ public:
+ MetadataImpl(const rd_kafka_metadata_t *metadata);
+ ~MetadataImpl();
+
+ const std::vector<const BrokerMetadata *> *brokers() const {
+ return &brokers_;
+ }
+
+ const std::vector<const TopicMetadata *> *topics() const {
+ return &topics_;
+ }
+
+ std::string orig_broker_name() const {
+ return std::string(metadata_->orig_broker_name);
+ }
+
+ int32_t orig_broker_id() const {
+ return metadata_->orig_broker_id;
+ }
+
+ private:
+ const rd_kafka_metadata_t *metadata_;
+ std::vector<const BrokerMetadata *> brokers_;
+ std::vector<const TopicMetadata *> topics_;
+ std::string orig_broker_name_;
+};
+
+
+
+class ConsumerImpl : virtual public Consumer, virtual public HandleImpl {
+ public:
+ ~ConsumerImpl() {
+ if (rk_)
+ rd_kafka_destroy(rk_);
+ }
+ static Consumer *create(Conf *conf, std::string &errstr);
+
+ ErrorCode start(Topic *topic, int32_t partition, int64_t offset);
+ ErrorCode start(Topic *topic,
+ int32_t partition,
+ int64_t offset,
+ Queue *queue);
+ ErrorCode stop(Topic *topic, int32_t partition);
+ ErrorCode seek(Topic *topic,
+ int32_t partition,
+ int64_t offset,
+ int timeout_ms);
+ Message *consume(Topic *topic, int32_t partition, int timeout_ms);
+ Message *consume(Queue *queue, int timeout_ms);
+ int consume_callback(Topic *topic,
+ int32_t partition,
+ int timeout_ms,
+ ConsumeCb *cb,
+ void *opaque);
+ int consume_callback(Queue *queue,
+ int timeout_ms,
+ RdKafka::ConsumeCb *consume_cb,
+ void *opaque);
+};
+
+
+
+class ProducerImpl : virtual public Producer, virtual public HandleImpl {
+ public:
+ ~ProducerImpl() {
+ if (rk_)
+ rd_kafka_destroy(rk_);
+ }
+
+ ErrorCode produce(Topic *topic,
+ int32_t partition,
+ int msgflags,
+ void *payload,
+ size_t len,
+ const std::string *key,
+ void *msg_opaque);
+
+ ErrorCode produce(Topic *topic,
+ int32_t partition,
+ int msgflags,
+ void *payload,
+ size_t len,
+ const void *key,
+ size_t key_len,
+ void *msg_opaque);
+
+ ErrorCode produce(Topic *topic,
+ int32_t partition,
+ const std::vector<char> *payload,
+ const std::vector<char> *key,
+ void *msg_opaque);
+
+ ErrorCode produce(const std::string topic_name,
+ int32_t partition,
+ int msgflags,
+ void *payload,
+ size_t len,
+ const void *key,
+ size_t key_len,
+ int64_t timestamp,
+ void *msg_opaque);
+
+ ErrorCode produce(const std::string topic_name,
+ int32_t partition,
+ int msgflags,
+ void *payload,
+ size_t len,
+ const void *key,
+ size_t key_len,
+ int64_t timestamp,
+ RdKafka::Headers *headers,
+ void *msg_opaque);
+
+ ErrorCode flush(int timeout_ms) {
+ return static_cast<RdKafka::ErrorCode>(rd_kafka_flush(rk_, timeout_ms));
+ }
+
+ ErrorCode purge(int purge_flags) {
+ return static_cast<RdKafka::ErrorCode>(
+ rd_kafka_purge(rk_, (int)purge_flags));
+ }
+
+ Error *init_transactions(int timeout_ms) {
+ rd_kafka_error_t *c_error;
+
+ c_error = rd_kafka_init_transactions(rk_, timeout_ms);
+
+ if (c_error)
+ return new ErrorImpl(c_error);
+ else
+ return NULL;
+ }
+
+ Error *begin_transaction() {
+ rd_kafka_error_t *c_error;
+
+ c_error = rd_kafka_begin_transaction(rk_);
+
+ if (c_error)
+ return new ErrorImpl(c_error);
+ else
+ return NULL;
+ }
+
+ Error *send_offsets_to_transaction(
+ const std::vector<TopicPartition *> &offsets,
+ const ConsumerGroupMetadata *group_metadata,
+ int timeout_ms) {
+ rd_kafka_error_t *c_error;
+ const RdKafka::ConsumerGroupMetadataImpl *cgmdimpl =
+ dynamic_cast<const RdKafka::ConsumerGroupMetadataImpl *>(
+ group_metadata);
+ rd_kafka_topic_partition_list_t *c_offsets = partitions_to_c_parts(offsets);
+
+ c_error = rd_kafka_send_offsets_to_transaction(
+ rk_, c_offsets, cgmdimpl->cgmetadata_, timeout_ms);
+
+ rd_kafka_topic_partition_list_destroy(c_offsets);
+
+ if (c_error)
+ return new ErrorImpl(c_error);
+ else
+ return NULL;
+ }
+
+ Error *commit_transaction(int timeout_ms) {
+ rd_kafka_error_t *c_error;
+
+ c_error = rd_kafka_commit_transaction(rk_, timeout_ms);
+
+ if (c_error)
+ return new ErrorImpl(c_error);
+ else
+ return NULL;
+ }
+
+ Error *abort_transaction(int timeout_ms) {
+ rd_kafka_error_t *c_error;
+
+ c_error = rd_kafka_abort_transaction(rk_, timeout_ms);
+
+ if (c_error)
+ return new ErrorImpl(c_error);
+ else
+ return NULL;
+ }
+
+ static Producer *create(Conf *conf, std::string &errstr);
+};
+
+
+
+} // namespace RdKafka
+
+#endif /* _RDKAFKACPP_INT_H_ */