Merge pull request #4 from lpcvoid/add-http-functionality

Add http functionality

Author: lpcvoid

.github/workflows/test_matrix.yml

@@ -18,7 +18,7 @@ jobs:
             os: windows-latest,
             build_type: "Release", cc: "cl", cxx: "cl",
             #this path will change at some point when VC2022 is released I guess
-            environment_script: "C:/Program Files (x86)/Microsoft Visual Studio/2019/Enterprise/VC/Auxiliary/Build/vcvars64.bat"
+            environment_script: "C:/Program Files (x86)/Microsoft Visual Studio/2022/Enterprise/VC/Auxiliary/Build/vcvars64.bat"
           }
           - {
             name: "Ubuntu gcc",
@@ -31,9 +31,9 @@ jobs:
             build_type: "Release", cc: "clang", cxx: "clang++"
           }
           - {
-            name: "MacOS clang",
+            name: "MacOS gcc",
             os: macos-latest,
-            build_type: "Release", cc: "clang", cxx: "clang++"
+            build_type: "Release", cc: "gcc", cxx: "g++"
           }
     steps:
       - uses: actions/checkout@v2

.gitmodules

@@ -1,3 +1,6 @@
 [submodule "doctest"]
 	path = doctest
 	url = https://github.com/onqtam/doctest.git
+[submodule "extern/cpp-uri-parser"]
+	path = extern/cpp-uri-parser
+	url = https://github.com/lpcvoid/cpp-uri-parser

CMakeLists.txt

@@ -19,13 +19,29 @@ set(NETLIB_SRC
         src/service_resolver.hpp
         src/endpoint_accessor.hpp
         src/thread_pool.hpp
-        src/socket_operations.hpp)
+        src/socket_operations.hpp
+)
+
+set(NETLIB_HTTP
+        src/http/client.hpp
+        src/http/http.hpp
+)
 
 option(BUILD_TESTS "Build tests" ON)
 option(BUILD_EXAMPLES "Build example programs" ON)
+option(WITH_HTTP "Build with http support" ON)
 
 if(NOT EXISTS "${CMAKE_SOURCE_DIR}/doctest/CMakeLists.txt")
     set(BUILD_TESTS OFF)
+    message(WARNING "Cannot build tests without doctest! Deactivating tests.")
+endif()
+
+if (WITH_HTTP)
+    if(NOT EXISTS "${CMAKE_SOURCE_DIR}/extern/cpp-uri-parser/URI.hpp")
+        set(WITH_HTTP OFF)
+        message(WARNING "If you want HTTP support, you need to check out submodules (cpp-uri-parser)")
+    endif()
+    message(NOTICE "Building with HTTP support.")
 endif()
 
 if(BUILD_TESTS)

examples/CMakeLists.txt

@@ -1,5 +1,9 @@
 set(EXAMPLE_SOURCES echo_server.cpp daytime_client.cpp threadpool.cpp time_client.cpp)
 
+if (WITH_HTTP)
+    set(EXAMPLE_SOURCES ${EXAMPLE_SOURCES} http_client.cpp)
+endif()
+
 foreach (examplesource ${EXAMPLE_SOURCES})
     string(REPLACE ".cpp" "" examplename ${examplesource})
     add_executable(${examplename} ${examplesource})

examples/http_client.cpp

@@ -0,0 +1,35 @@
+#include "../src/netlib.hpp"
+#include "../src/http/client.hpp"
+#include "../src/http/http.hpp"
+#include <csignal>
+#include <iomanip>
+#include <iostream>
+
+
+void exit_handler(int s){
+    std::cout << "Goodbye!" << std::endl;
+    exit(EXIT_SUCCESS);
+}
+
+int main(int argc, char** argv)
+{
+    netlib::http::http_client client;
+
+    auto res = client.get("http://example.com");
+
+    if (res.second) {
+        std::cerr << "Error: " << res.second.message() << std::endl;
+        exit(1);
+    }
+
+    std::cout << "Got HTTP response: " << res.first->response_code <<
+        ", version " << res.first->version.first << "." << res.first->version.second << std::endl;
+    std::cout << "Header entries:" << std::endl;
+    std::for_each(res.first->headers.begin(), res.first->headers.end(), [](auto header_entry) {
+       std::cout << header_entry.first << " = " << header_entry.second << std::endl;
+    });
+    std::cout << "Body:" << std::endl;
+    std::cout << res.first.value().body << std::endl;
+
+    signal(SIGINT, exit_handler);
+}

extern/cpp-uri-parser

@@ -13,10 +13,9 @@
 namespace netlib {
 
 using namespace std::chrono_literals;
+static constexpr std::chrono::milliseconds DEFAULT_TIMEOUT = 1000ms;
 
 class client {
-private:
-    static constexpr std::chrono::milliseconds DEFAULT_TIMEOUT = 1000ms;
 protected:
     std::optional<netlib::socket> _socket;
     addrinfo *_endpoint_addr = nullptr;

src/client.hpp

@@ -0,0 +1,83 @@
+#include "../extern/cpp-uri-parser/URI.hpp"
+#include "http.hpp"
+#pragma once
+
+namespace netlib::http {
+
+using namespace std::chrono_literals;
+
+class http_client {
+private:
+    netlib::thread_pool _thread_pool;
+    netlib::client _client;
+public:
+
+    inline http_client() {}
+
+    inline std::pair<std::optional<netlib::http::http_response>, std::error_condition> get(const std::string& url) {
+        auto uri = URI(url);
+        if (uri.get_result() != URI::URIParsingResult::success) {
+            return std::make_pair(std::nullopt, std::errc::bad_address);
+        }
+        if (!_client.is_connected()) {
+            uint16_t port = 80;
+            if (uri.get_protocol().has_value()) {
+                std::string_view protocol_str = uri.get_protocol().value();
+                if (protocol_str != "http"){
+                    // no support for TLS so far...
+                    // also, only http and https shall ever be expected
+                    return std::make_pair(std::nullopt, std::errc::protocol_not_supported);
+                }
+            }
+            auto res = _client.connect(std::string(uri.get_host().value()), port, AddressFamily::IPv4, AddressProtocol::TCP);
+            if (res) {
+                return std::make_pair(std::nullopt, res);
+            }
+        }
+
+        std::string query = (uri.get_query().has_value() ? std::string(uri.get_query().value()) : "/");
+        std::string http_get = "GET " + query + " HTTP/1.1\r\nHost:" + std::string(uri.get_host().value()) + "\r\n\r\n";
+        const std::vector<uint8_t> get_data(http_get.begin(), http_get.end());
+        auto send_res = _client.send(get_data);
+        if (send_res.first != get_data.size()) {
+            return std::make_pair(std::nullopt, send_res.second);
+        }
+        std::vector<uint8_t> data_buffer;
+        std::chrono::milliseconds time_spent = std::chrono::milliseconds(0);
+        const std::chrono::milliseconds TICK_TIME = std::chrono::milliseconds(50);
+        while (true) {
+            auto recv_res = _client.recv(0, TICK_TIME);
+            time_spent += TICK_TIME;
+            if (!recv_res.first.empty()) {
+                data_buffer.insert(data_buffer.begin(), recv_res.first.begin(), recv_res.first.end());
+            }
+            if ((recv_res.second == std::errc::timed_out) && (!data_buffer.empty())) {
+                netlib::http::http_response response;
+                std::string raw_response (data_buffer.begin(), data_buffer.end());
+                std::error_condition parse_resp = response.from_raw_response(raw_response);
+                if (!parse_resp) {
+                    return {response, {}};
+                } else {
+                    return {std::nullopt, parse_resp};
+                }
+            }
+            if (time_spent > DEFAULT_TIMEOUT) {
+                return {std::nullopt, std::errc::timed_out};
+            }
+        }
+    }
+
+    inline std::future<std::pair<std::optional<netlib::http::http_response>, std::error_condition>> get_async(const std::string& url) {
+        return _thread_pool.add_task(
+            [&](std::string url) {
+                return this->get(url);
+            },
+            url);
+    }
+
+
+
+
+};
+
+}

src/http/client.hpp

@@ -0,0 +1,90 @@
+#pragma once
+
+#include <string>
+#include <system_error>
+#include <utility>
+#include <vector>
+#include <algorithm>
+
+namespace netlib::http {
+
+using http_header_entry = std::pair<std::string, std::string>;
+using http_headers = std::vector<http_header_entry>;
+
+struct http_response {
+    http_headers headers;
+    uint32_t response_code;
+    std::pair<uint32_t, uint32_t> version;
+    std::string body;
+    inline std::error_condition from_raw_response(const std::string& raw_response) {
+        // a very rudimentary http response parser
+        if (raw_response.empty()) {
+            return std::errc::no_message;
+        }
+
+        /* strategy:
+         * split into multiple (at least two) parts, delimited by \r\n\r\n"
+         * within the first part:
+         *      first, parse the status line
+         *      second, parse the header fields, until we arrive at an empty line (only CR LF)
+         *      last, an optional body
+         * then, for the rest of the parts, concat into body
+         */
+
+        auto split = [](const std::string& str, const std::string& delimiter) -> std::vector<std::string> {
+            std::vector<std::string> split_tokens;
+            std::size_t start;
+            std::size_t end = 0;
+            while ((start = str.find_first_not_of(delimiter, end)) != std::string::npos)
+            {
+                end = str.find(delimiter, start);
+                split_tokens.push_back(str.substr(start, end - start));
+            }
+            return split_tokens;
+        };
+
+        std::vector<std::string> header_body_split = split(raw_response, "\r\n\r\n");
+        //split header part of response into response_header_lines
+        std::vector<std::string> response_header_lines = split(header_body_split.front(), "\r\n");
+        //first line should start with "HTTP"
+        if (!response_header_lines.front().starts_with("HTTP")) {
+            return std::errc::result_out_of_range;
+        }
+        //attempt to parse status line
+        //split into parts by space
+        auto status_parts = split(response_header_lines.front(), " ");
+        if (status_parts.size() < 3) {
+            return std::errc::bad_message;
+        }
+        //parse "HTTP/x.x"
+        auto version_parts = split(status_parts.front(), "/");
+        if (version_parts.size() != 2) {
+            return std::errc::bad_message;
+        }
+        //parse "x.x"
+        auto version_components = split(version_parts.back(), ".");
+        version.first = std::stoi(version_components.front());
+        version.second = std::stoi(version_components.back());
+        //parse response code
+        response_code = std::stoi(status_parts[1]);
+        //there can be an optional code description in the first line, but we ignore that here
+        //parse the response header lines until the end
+        //start at second line, first is status
+        std::for_each(response_header_lines.begin(), response_header_lines.end(), [&](const std::string& header_component){
+            auto component_parts = split(header_component, ":");
+            if (component_parts.size() == 2) {
+                headers.emplace_back(component_parts.front(), component_parts.back());
+            }
+        });
+
+        //now, take the body part(s) and concat them
+        std::for_each(header_body_split.begin() + 1, header_body_split.end(), [&](const std::string& body_line){
+            body += body_line;
+        });
+
+        return {};
+
+    };
+};
+
+}

src/http/http.hpp

@@ -25,6 +25,14 @@ using ssize_t = signed long long int;
 #define MSG_NOSIGNAL 0
 //poll implementation
 #define poll_syscall ::WSAPoll
+//we ignore unused parameter warning on windows, missing impls
+#pragma warning(disable: 4100)
+//use unsafe functions under windows
+#define _CRT_SECURE_NO_WARNINGS
+#pragma warning(disable:4996)
+#pragma warning(disable:4267)
+#pragma warning(disable:4244)
+
 #else
 // headers
 #include <arpa/inet.h>
@@ -42,7 +50,9 @@ using socket_t = int32_t;
 // this is actually a nice one to have
 #define INVALID_SOCKET (-1)
 #ifdef __APPLE__
-#define MSG_NOSIGNAL SO_NOSIGPIPE
+    #ifndef MSG_NOSIGNAL
+        #define MSG_NOSIGNAL SO_NOSIGPIPE
+    #endif
 #endif
 #endif