Sockets

Sockets in Scalability Protocols provide the handle for communication between peers. Sockets also encapsulate protocol specific semantics, such as filtering subscriptions, or automatically retrying requests.

Socket Structure

#define NNG_SOCKET_INITIALIZER // opaque value

typedef struct nng_socket_s nng_socket;

The nng_socket structure represents socket. This is a handle, and the members of it are opaque. However, unlike a pointer, it is usually passed by value.

A socket may be initialized statically with the NNG_SOCKET_INITIALIZER macro, to ensure that it cannot be confused with a valid open socket.

Socket Identity

int nng_socket_id(nng_socket s);
int nng_socket_raw(nng_socket s, bool *raw);
int nng_socket_proto_id(nng_socket s, uint16_t *proto);
int nng_socket_peer_id(nng_socket s, uint16_t *proto);
int nng_socket_proto_name(nng_socket s, const char **name);
int nng_socket_peer_name(nng_socket s, const char **name);

These functions are used to provide fundamental information about the socket s. Most applications will not need to use these functions.

The nng_socket_id function returns the numeric id, which will be a non-negative value, associated with the socket. If the socket is uninitialized (has never been opened), then the return value may be -1.

The nng_socket_proto_id and nng_socket_peer_id functions provide the 16-bit protocol identifier for the socket’s protocol, and of the protocol peers will use when communicating with the socket.

The nng_socket_proto_name and nng_socket_peer_name functions provide the ASCII names of the socket’s protocol, and of the protocol peers of the socket use. The value stored in name is a fixed string located in program text, and must not be freed or altered. It is guaranteed to remain valid while this library is present.

The nng_socket_raw function determines whether the socket is in raw mode or not, storing true in raw if it is, or false if it is not.

Opening a Socket

int nng_bus0_open(nng_socket *s);
int nng_pub0_open(nng_socket *s);
int nng_pull0_open(nng_socket *s);
int nng_push0_open(nng_socket *s);
int nng_rep0_open(nng_socket *s);
int nng_req0_open(nng_socket *s);
int nng_respondent0_open(nng_socket *s);
int nng_sub0_open(nng_socket *s);
int nng_surveyor0_open(nng_socket *s);

These functions open a socket, returning it in s. The constructors for sockets are protocol specific so please refer to protocol documentation for more specific information.

The following functions open a socket in normal mode:

• nng_bus0_open - BUS version 0
• nng_pair0_open - PAIR version 0
• nng_pair1_open - PAIR version 1
• nng_pair1_open_poly - PAIR version 1, polyamorous mode
• nng_pub0_open - PUB version 0
• nng_pull0_open - PULL version 0
• nng_push0_open - PUSH version 0
• nng_rep0_open - REP version 0
• nng_req0_open - REQ version 0
• nng_respondent0_open - RESPONDENT version 0
• nng_sub0_open - SUB version 0
• nng_surveyor0_open - SURVEYOR version 0

Raw Mode Sockets

int nng_bus0_open_raw(nng_socket *s);
int nng_pub0_open_raw(nng_socket *s);
int nng_pull0_open_raw(nng_socket *s);
int nng_push0_open_raw(nng_socket *s);
int nng_rep0_open_raw(nng_socket *s);
int nng_req0_open_raw(nng_socket *s);
int nng_respondent0_open_raw(nng_socket *s);
int nng_sub0_open_raw(nng_socket *s);
int nng_surveyor0_open_raw(nng_socket *s);

Raw mode sockets are used in circumstances when the application needs direct access to the message headers to control the protocol details.

Such sockets require greater sophistication on the part of the application to use, as the application must process the protocol headers specifically. The details of the protocol headers, and requirements, are described in the protocol documentation for each protocol.

Raw mode sockets do not have any kind of state machine associated with them, as all of the protocol specific processing must be performed by the application.

The following functions open a socket in raw mode:

• nng_bus0_open_raw - BUS version 0, raw mode
• nng_pair0_open_raw - PAIR version 0, raw mode
• nng_pair1_open_raw - PAIR version 1, raw mode
• nng_pub0_open_raw - PUB version 0, raw mode
• nng_pull0_open_raw - PULL version 0, raw mode
• nng_push0_open_raw - PUSH version 0, raw mode
• nng_rep0_open_raw - REP version 0, raw mode
• nng_req0_open_raw - REP version 0, raw mode
• nng_respondent0_open_raw - RESPONDENT version 0, raw mode
• nng_sub0_open_raw - SUB version 0, raw mode
• nng_surveyor0_open_raw - SURVEYOR version 0, raw mode

Closing a Socket

int nng_socket_close(nng_socket s);

The nng_socket_close function closes a socket, releasing all resources associated with it. Any operations that are in progress will be terminated with a result of NNG_ECLOSED.

Sending Messages

int nng_send(nng_socket s, void *data, size_t size, int flags);
int nng_sendmsg(nng_socket s, nng_msg *msg, int flags);
void nng_socket_send(nng_socket s, nng_aio *aio);

These functions (nng_send, nng_sendmsg, and nng_socket_send) send messages over the socket s. The differences in their behaviors are as follows.

nng_send

The nng_send function is the simplest to use, but is the least efficient. It sends the content in data, as a message of size bytes. The flags is a bit mask made up of zero or more of the following values:

• NNG_FLAG_NONBLOCK: If the socket cannot accept more data at this time, it does not block, but returns immediately with a status of NNG_EAGAIN. If this flag is absent, the function will wait until data can be sent.

nng_sendmsg

The nng_sendmsg function sends the msg over the socket s.

If this function returns zero, then the socket will dispose of msg when the transmission is complete. If the function returns a non-zero status, then the call retains the responsibility for disposing of msg.

The flags can contain the value NNG_FLAG_NONBLOCK, indicating that the function should not wait if the socket cannot accept more data for sending. In such a case, it will return NNG_EAGAIN.

nng_socket_send

The nng_socket_send function sends a message asynchronously, using the nng_aio aio, over the socket s. The message to send must have been set on aio using the nng_aio_set_msg function.

If the operation completes successfully, then the socket will have disposed of the message. However, if it fails, then callback of aio should arrange for a final disposition of the message. (The message can be retrieved from aio with nng_aio_get_msg.)

Note that callback associated with aio may be called before the message is finally delivered to the recipient. For example, the message may be sitting in queue, or located in TCP buffers, or even in flight.

Receiving Messages

int nng_recv(nng_socket s, void *data, size_t *sizep, int flags);
int nng_recvmsg(nng_socket s, nng_msg **msgp, int flags);
void nng_socket_recv(nng_socket s, nng_aio *aio);

These functions (nng_recv, nng_recvmsg, and nng_socket_recv) receive messages over the socket s. The differences in their behaviors are as follows.

nng_recv

The nng_recv function is the simplest to use, but is the least efficient. It receives the content in data, as a message size (in bytes) of up to the value stored in sizep.

Upon success, the size of the message received will be stored in sizep.

The flags is a bit mask made up of zero or more of the following values:

• NNG_FLAG_NONBLOCK: If the socket has no messages pending for reception at this time, it does not block, but returns immediately with a status of NNG_EAGAIN. If this flag is absent, the function will wait until data can be received.

nng_recvmsg

The nng_recvmsg function receives a message and stores a pointer to the nng_msg for that message in msgp.

The flags can contain the value NNG_FLAG_NONBLOCK, indicating that the function should not wait if the socket has no messages available to receive. In such a case, it will return NNG_EAGAIN.

nng_socket_recv

The nng_socket_send function receives a message asynchronously, using the nng_aio aio, over the socket s. On success, the received message can be retrieved from the aio using the nng_aio_get_msg function.

Socket Options

int nng_socket_get_bool(nng_socket s, const char *opt, bool *valp);
int nng_socket_get_int(nng_socket s, const char *opt, int *valp);
int nng_socket_get_ms(nng_socket s, const char *opt, nng_duration *valp);
int nng_socket_get_size(nng_socket s, const char *opt, size_t *valp);

int nng_socket_set_bool(nng_socket s, const char *opt, int val);
int nng_socket_set_int(nng_socket s, const char *opt, int val);
int nng_socket_set_ms(nng_socket s, const char *opt, nng_duration val);
int nng_socket_set_size(nng_socket s, const char *opt, size_t val);

Protocols usually have protocol specific behaviors that can be adjusted via options.

These functions are used to retrieve or change the value of an option named opt from the context ctx. The nng_socket_get_ functions retrieve the value from the socket s, and store it in the location valp references. The nng_socket_set_ functions change the value for the socket s, taking it from val.

These functions access an option as a specific type. The protocol documentation will have details about which options are available, whether they can be read or written, and the appropriate type to use.

Common Options

The following options are available for many protocols, and always use the same types and semantics described below.

Polling Socket Events

int nng_socket_get_recv_poll_fd(nng_socket s, int *fdp);
int nng_socket_get_send_poll_fd(nng_socket s, int *fdp);

Sometimes it is necessary to integrate a socket into a poll or select driven event loop. (Or, on Linux, epoll, or on BSD derived systems like macOS kqueue).

For these occasions, a suitable file descriptor for polling is provided by these two functions.

The nng_socket_get_recv_poll_fd function obtains a file descriptor that will poll as readable when a message is ready for receiving for the socket.

The nng_socket_get_send_poll_fd function obtains a file descriptor that will poll as readable when the socket can accept a message for sending.

These file descriptors should only be polled for readability, and no other operation performed on them. The socket will read from, or write to, these file descriptors to provide a level-signaled behavior automatically.

Additionally the socket will close these file descriptors when the socket itself is closed.

These functions replace the NNG_OPT_SENDFD and NNG_OPT_RECVFD socket options that were available in previous versions of NNG.

Examples

Example 1: Initializing a Socket

nng_socket s = NNG_SOCKET_INITIALIZER;

Example 2: Publishing a Timestamp

This example demonstrates the use of nng_aio, nng_socket_send, and nng_sleep_aio to build a service that publishes a timestamp at one second intervals. Error handling is elided for the sake of clarity.

#include <stdlib.h>
#include <stdio.h>
#include <nng/nng.h>
#include <nng/protocol/pubsub0/pub.h>

struct state {
    nng_socket s;
    bool sleeping;
    nng_aio *aio;
};

static struct state state;

void callback(void *arg) {
    nng_msg *msg;
    nng_time now;
    struct state *state = arg;
    if (nng_aio_result(state->aio) != 0) {
        fprintf(stderr, "Error %s occurred", nng_strerror(nng_aio_result(state->aio)));
        return; // terminate the callback loop
    }
    if (state->sleeping) {
        state->sleeping = false;
        nng_msg_alloc(&msg, sizeof (nng_time));
        now = nng_clock();
        nng_msg_append(msg, &now, sizeof (now)); // note: native endian
        nng_aio_set_msg(state->aio, msg);
        nng_socket_send(state->s, state->aio);
    } else {
        state->sleeping = true;
        nng_sleep_aio(1000, state->aio); // 1000 ms == 1 second
    }
}

int main(int argc, char **argv) {
    const char *url = argv[1]; // should check this

    nng_aio_alloc(&state.aio, NULL, NULL);
    nng_pub0_open(&state.s);
    nng_listen(state.s, url, NULL, 0);
    state.sleeping = 0;
    nng_sleep_aio(1, state.aio); // kick it off right away
    for(;;) {
        nng_msleep(0x7FFFFFFF); // infinite, could use pause or sigsuspend
    }
}

Example 3: Watching a Periodic Timestamp

This example demonstrates the use of nng_aio, nng_socket_recv, to build a client to watch for messages received from the service created in Example 2. Error handling is elided for the sake of clarity.

#include <stdlib.h>
#include <stdio.h>
#include <string.h>>
#include <nng/nng.h>
#include <nng/protocol/pubsub0/sub.h>

struct state {
    nng_socket s;
    nng_aio *aio;
};

static struct state state;

void callback(void *arg) {
    nng_msg *msg;
    nng_time now;
    struct state *state = arg;
    if (nng_aio_result(state->aio) != 0) {
        fprintf(stderr, "Error %s occurred", nng_strerror(nng_aio_result(state->aio)));
        return; // terminate the callback loop
    }
    msg = nng_aio_get_msg(state->aio);
    memcpy(&now, nng_msg_body(msg), sizeof (now)); // should check the length!
    printf("Timestamp is %lu\n", (unsigned long)now);
    nng_msg_free(msg);
    nng_aio_set_msg(state->aio, NULL);
    nng_socket_recv(state->s, state->aio);
}

int main(int argc, char **argv) {
    const char *url = argv[1]; // should check this

    nng_aio_alloc(&state.aio, NULL, NULL);
    nng_sub0_open(&state.s);
    nng_sub0_socket_subscribe(state.s, NULL, 0); // subscribe to everything
    nng_dial(state.s, url, NULL, 0);
    nng_socket_recv(state.s, state.aio); // kick it off right away
    for(;;) {
        nng_msleep(0x7FFFFFFF); // infinite, could use pause or sigsuspend
    }
}