This is a simple REST client for C++. It wraps libcurl for HTTP requests.
restclient-cpp provides two ways of interacting with REST endpoints. There is a simple one, which doesn't need you to configure an object to interact with an API. However the simple way doesn't provide a lot of configuration options either. So if you need more than just a simple HTTP call, you will probably want to check out the advanced usage.
The simple API is just some static methods modeled after the most common HTTP verbs:
#include "restclient-cpp/restclient.h"
RestClient::Response r = RestClient::get("http://url.com")
RestClient::Response r = RestClient::post("http://url.com/post", "application/json", "{\"foo\": \"bla\"}")
RestClient::Response r = RestClient::put("http://url.com/put", "application/json", "{\"foo\": \"bla\"}")
RestClient::Response r = RestClient::patch("http://url.com/patch", "application/json", "{\"foo\": \"bla\"}")
RestClient::Response r = RestClient::del("http://url.com/delete")
RestClient::Response r = RestClient::head("http://url.com")
RestClient::Response r = RestClient::options("http://url.com")
The response is of type RestClient::Response and has three attributes:
RestClient::Response.code // HTTP response code
RestClient::Response.body // HTTP response body
RestClient::Response.headers // HTTP response headers
However if you want more sophisticated features like connection reuse, timeouts or authentication, there is also a different, more configurable way.
#include "restclient-cpp/connection.h"
#include "restclient-cpp/restclient.h"
// initialize RestClient
RestClient::init();
// get a connection object
RestClient::Connection* conn = new RestClient::Connection("http://url.com");
// configure basic auth
conn->SetBasicAuth("WarMachine68", "WARMACHINEROX");
// set connection timeout to 5s
conn->SetTimeout(5);
// set custom user agent
// (this will result in the UA "foo/cool restclient-cpp/VERSION")
conn->SetUserAgent("foo/cool");
// enable following of redirects (default is off)
conn->FollowRedirects(true);
// and limit the number of redirects (default is -1, unlimited)
conn->FollowRedirects(true, 3);
// set headers
RestClient::HeaderFields headers;
headers["Accept"] = "application/json";
conn->SetHeaders(headers)
// append additional headers
conn->AppendHeader("X-MY-HEADER", "foo")
// if using a non-standard Certificate Authority (CA) trust file
conn->SetCAInfoFilePath("/etc/custom-ca.crt")
RestClient::Response r = conn->get("/get")
RestClient::Response r = conn->head("/get")
RestClient::Response r = conn->del("/delete")
RestClient::Response r = conn->options("/options")
// set different content header for POST, PUT and PATCH
conn->AppendHeader("Content-Type", "application/json")
RestClient::Response r = conn->post("/post", "{\"foo\": \"bla\"}")
RestClient::Response r = conn->put("/put", "application/json", "{\"foo\": \"bla\"}")
RestClient::Response r = conn->patch("/patch", "text/plain", "foobar")
// deinit RestClient. After calling this you have to call RestClient::init()
// again before you can use it
RestClient::disable();
The responses are again of type RestClient::Response and have three attributes:
RestClient::Response.code // HTTP response code
RestClient::Response.body // HTTP response body
RestClient::Response.headers // HTTP response headers
The connection object also provides a simple way to get some diagnostics and
metrics information via conn->GetInfo()
. The result is a
RestClient::Connection::Info
struct and looks like this:
typedef struct {
std::string base_url;
RestClients::HeaderFields headers;
int timeout;
struct {
std::string username;
std::string password;
} basicAuth;
std::string certPath;
std::string certType;
std::string keyPath;
std::string keyPassword;
std::string customUserAgent;
std::string uriProxy;
struct {
// total time of the last request in seconds Total time of previous
// transfer. See CURLINFO_TOTAL_TIME
int totalTime;
// time spent in DNS lookup in seconds Time from start until name
// resolving completed. See CURLINFO_NAMELOOKUP_TIME
int nameLookupTime;
// time it took until Time from start until remote host or proxy
// completed. See CURLINFO_CONNECT_TIME
int connectTime;
// Time from start until SSL/SSH handshake completed. See
// CURLINFO_APPCONNECT_TIME
int appConnectTime;
// Time from start until just before the transfer begins. See
// CURLINFO_PRETRANSFER_TIME
int preTransferTime;
// Time from start until just when the first byte is received. See
// CURLINFO_STARTTRANSFER_TIME
int startTransferTime;
// Time taken for all redirect steps before the final transfer. See
// CURLINFO_REDIRECT_TIME
int redirectTime;
// number of redirects followed. See CURLINFO_REDIRECT_COUNT
int redirectCount;
} lastRequest;
} Info;
The connection object stores the curl easy handle in an instance variable and uses that for the lifetime of the object. This means curl will automatically reuse connections made with that handle.
Two wrapper functions are provided to setup the progress callback for uploads/downloads.
Calling conn->SetFileProgressCallback(callback)
with a callback parameter matching the prototype int progress_callback(void *clientp, double dltotal, double dlnow, double ultotal, double ulnow)
will setup the progress callback.
Calling conn->SetFileProgressCallbackData(data)
is optional. This will set the data pointer which is the first parameter fed back to the progress callback - clientp
. If this isn't set then clientp
will default to the connection object conn
.
// set CURLOPT_NOPROGRESS
// set CURLOPT_PROGRESSFUNCTION
conn->SetFileProgressCallback(progressFunc);
// set CURLOPT_PROGRESSDATA
conn->SetFileProgressCallbackData(data);
A write callback function can be provided for processing data as it's received from a GET call (for instance the Kubernetes Watch API).
Calling conn->SetWriteFunction(callback)
with a function parameter matching the prototype size_t write_function(void *data, size_t size, size_t nmemb, void *userdata)int progress_callback(void *clientp, double dltotal, double dlnow, double ultotal, double ulnow)
will setup the write function.
Here is an example of a write callback function, processing result data line by line.
auto writeCallback = [](void *data, size_t size, size_t nmemb, void *userdata) -> size_t
{
size_t bytes = size * nmemb;
try
{
// Add to the buffer
auto res = reinterpret_cast<RestClient::Response *>(userdata);
res->body.append(static_cast<char*>(data), bytes);
// If the last character is not a new line, wait for the rest.
if ('\n' != *(res->body.end() - 1))
{
return bytes;
}
// Process data one line at a time.
std::stringstream stream(res->body);
std::string line;
while (std::getline(stream, line))
{
// Do something with the line here...
}
// Done processing the line
res->body.clear();
}
catch(std::exception e)
{
// Log caught exception here
return 0;
}
return bytes;
};
When restclient-cpp encounters an error, generally the error (or "status") code is returned in the Response
(see
Response struct in restclient.h). This error code can be either
an HTTP error code, or if a lower-level cURL error was encountered, it may be
a CURLCode. Currently, libcurl only defines 92 error codes, which means
there is no overlap between cURL error codes and HTTP response codes (which start at 1xx). However, if in the future, libcurl defines more than 99
error codes, meaning that cURL errors overlap with the HTTP 1xx class of responses, restclient-cpp will return a -1 if the CURLCode is 100 or higher.
In this case, callers can use GetInfo().lastRequest.curlCode
to inspect the actual cURL error.
restclient-cpp leans heavily on libcurl as it aims to provide a thin wrapper
around it. This means it adheres to the basic level of thread safety provided
by libcurl. The RestClient::init()
and
RestClient::disable()
methods basically correspond to curl_global_init
and
curl_global_cleanup
and thus need to be called right at the beginning of
your program and before shutdown respectively. These set up the environment
and are not thread-safe. After that you can create connection objects in
your threads. Do not share connection objects across threads as this would
mean accessing curl handles from multiple threads at the same time which is
not allowed.
The connection level method SetNoSignal can be set to skip all signal handling. This is important in multi-threaded applications as DNS resolution timeouts use signals. The signal handlers quite readily get executed on other threads. Note that with this option DNS resolution timeouts do not work. If you have crashes in your multi-threaded executable that appear to be in DNS resolution, this is probably why.
In order to provide an easy to use API, the simple usage via the static methods implicitly calls the curl global functions and is therefore also not thread-safe.
Simple wrapper functions are provided to allow clients to authenticate using certificates.
Under the hood these wrappers set cURL options, e.g. CURLOPT_SSLCERT
, using curl_easy_setopt
.
Note: currently libcurl
compiled with gnutls
(e.g. libcurl4-gnutls-dev
on
ubuntu) is buggy in that it returns a wrong error code when these options are set to invalid values.
// set CURLOPT_SSLCERT
conn->SetCertPath(certPath);
// set CURLOPT_SSLCERTTYPE
conn->SetCertType(type);
// set CURLOPT_SSLKEY
conn->SetKeyPath(keyPath);
// set CURLOPT_KEYPASSWD
conn->SetKeyPassword(keyPassword);
An HTTP Proxy can be set to use for the upcoming request.
To specify a port number, append :[port] to the end of the host name. If not specified, libcurl
will default to using port 1080 for proxies. The proxy string may be prefixed with http://
or https://
. If no HTTP(S) scheme is specified, the address provided to libcurl
will be prefixed with http://
to specify an HTTP proxy. A proxy host string can embedded user + password.
The operation will be tunneled through the proxy as curl option CURLOPT_HTTPPROXYTUNNEL
is enabled by default.
A numerical IPv6 address must be written within [brackets].
// set CURLOPT_PROXY
conn->SetProxy("https://37.187.100.23:3128");
/* or you can set it without the protocol scheme and
http:// will be prefixed by default */
conn->SetProxy("37.187.100.23:3128");
/* the following request will be tunneled through the proxy */
RestClient::Response res = conn->get("/get");
- https://docs.docker.com/develop/sdk/examples/
- $ curl --unix-socket /var/run/docker.sock http:/v1.24/containers/json
Note that the URL used with a unix socket has only ONE leading forward slash.
RestClient::Connection* conn = new RestClient::Connection("http:/v1.30");
conn->SetUnixSocketPath("/var/run/docker.sock");
RestClient::HeaderFields headers;
headers["Accept"] = "application/json; charset=UTF-8";
headers["Expect"] = "";
conn->SetHeaders(headers);
auto resp = conn->get("/images/json");
There are some packages available for Linux on packagecloud. And for OSX you can get it from the mrtazz/oss homebrew tap:
brew tap mrtazz/oss
brew install restclient-cpp
Otherwise you can do the regular autotools dance:
./autogen.sh
./configure
make install
Alternatively, you can build and install restclient-cpp using vcpkg dependency manager:
git clone https://github.com/Microsoft/vcpkg.git
cd vcpkg
./bootstrap-vcpkg.sh
./vcpkg integrate install
./vcpkg install restclient-cpp
The restclient-cpp port in vcpkg is kept up to date by Microsoft team members and community contributors. If the version is out of date, please create an issue or pull request on the vcpkg repository.
Another option is to use Mingw64 and CMake to compile on Windows. This requires you to have Mingw64 installed with CMake and libCurl (MSYS2 would also work fine).
# Make sure cmake, mingw32-make and g++ are in PATH
# Clone the repo
git clone https://github.com/mrtazz/restclient-cpp && cd restclient-cpp
# Build library
mkdir build && cd build
# Run cmake
cmake -DCMAKE_CXX_COMPILER=g++ -DCMAKE_CC_COMPILER=gcc -DCMAKE_MAKE_PROGRAM=mingw32-make -G "MinGW Makefiles" ..
# Builds the dll.a lib and dll file
mingw32-make
All contributions are highly appreciated. This includes filing issues, updating documentation and writing code. Please take a look at the contributing guidelines before so your contribution can be merged as fast as possible.