LCM
|
Publish and receive messages. More...
Classes | |
struct | lcm_recv_buf_t |
Typedefs | |
typedef struct _lcm_t | lcm_t |
typedef struct _lcm_subscription_t | lcm_subscription_t |
typedef void(* | lcm_msg_handler_t) (const lcm_recv_buf_t *rbuf, const char *channel, void *user_data) |
Callback function prototype. More... | |
Functions | |
lcm_t * | lcm_create (const char *provider) |
Constructor. More... | |
void | lcm_destroy (lcm_t *lcm) |
Destructor. | |
int | lcm_get_fileno (lcm_t *lcm) |
Returns a file descriptor or socket that can be used with select() , poll() , or other event loops for asynchronous notification of incoming messages. More... | |
lcm_subscription_t * | lcm_subscribe (lcm_t *lcm, const char *channel, lcm_msg_handler_t handler, void *userdata) |
Subscribe a callback function to a channel, without automatic message decoding. More... | |
int | lcm_unsubscribe (lcm_t *lcm, lcm_subscription_t *handler) |
Unsubscribe a message handler. More... | |
int | lcm_publish (lcm_t *lcm, const char *channel, const void *data, unsigned int datalen) |
Publish a message, specified as a raw byte buffer. More... | |
int | lcm_handle (lcm_t *lcm) |
Wait for and dispatch the next incoming message. More... | |
int | lcm_handle_timeout (lcm_t *lcm, int timeout_millis) |
Wait for and dispatch the next incoming message, up to a time limit. More... | |
int | lcm_subscription_set_queue_capacity (lcm_subscription_t *handler, int num_messages) |
Adjusts the maximum number of received messages that can be queued up for a subscription. More... | |
int | lcm_subscription_get_queue_size (lcm_subscription_t *handler) |
Query the current number of unhandled messages queued up for a subscription. | |
Publish and receive messages.
All LCM functions are internally synchronized and thread-safe.
Linking:
pkg-config --libs lcm
typedef void(* lcm_msg_handler_t) (const lcm_recv_buf_t *rbuf, const char *channel, void *user_data) |
Callback function prototype.
Pass instances of this to lcm_subscribe()
rbuf | the message timestamp and payload |
channel | the channel the message was received on |
user_data | the user-specified parameter passed to lcm_subscribe() |
typedef struct _lcm_subscription_t lcm_subscription_t |
An opaque data structure that identifies an LCM subscription.
typedef struct _lcm_t lcm_t |
Opaque data structure containing the LCM context.
lcm_t* lcm_create | ( | const char * | provider | ) |
Constructor.
Allocates and initializes a lcm_t. provider must be either NULL, or a string of the form
"provider://network?option1=value1&option2=value2&...&optionN=valueN"
provider | Initialization string specifying the LCM network provider. If this is NULL, and the environment variable "LCM_DEFAULT_URL" is defined, then the environment variable is used instead. If this is NULL and the environment variable is not defined, then default settings are used. |
The currently supported providers are:
udpm:// UDP Multicast provider network can be of the form "multicast_address:port". Either the multicast address or the port may be ommitted for the default. options: recv_buf_size = N size of the kernel UDP receive buffer to request. Defaults to operating system defaults ttl = N time to live of transmitted packets. Default 0 examples: "udpm://239.255.76.67:7667" Default initialization string "udpm://239.255.76.67:7667?ttl=1" Sets the multicast TTL to 1 so that packets published will enter the local network.
file:// LCM Log file-based provider network should be the path to the log file Events are read from or written to the log file. In read mode, events are generated from the log file in real-time, or at the rate specified by the speed option. In write mode, events published to the LCM instance will be written to the log file in real-time. options: speed = N Scale factor controlling the playback speed of the log file. Defaults to 1. If less than or equal to zero, then the events in the log file are played back as fast as possible. Events are never skipped in read mode, so actual playback speed may be slower than requested, depending on the handlers. mode = r | w Specifies the log file mode. Defaults to 'r' start_timestamp = USEC Seeks to USEC microseconds in the logfile, where USEC is given in microseconds since 00:00:00 UTC on 1 January 1970. If USEC is before the first event, then playback begins at the start of the log file. If it is after the last event, calls to lcm_handle will return -1. examples: "file:///home/albert/path/to/logfile" Loads the file "/home/albert/path/to/logfile" as an LCM event source. "file:///home/albert/path/to/logfile?speed=4" Loads the file "/home/albert/path/to/logfile" as an LCM event source. Events are played back at 4x speed.
memq:// Memory queue test provider This provider is primarily useful for testing, especially unit testing code that uses LCM. It provides a pub/sub interface that is private to the LCM instance, and does _not_ provide any interprocess communication, or even inter-instance communication. Use this provider to implement unit tests that require deterministic and predictable behavior that is independent of a system's network configuration. example: "memq://" This is the only valid way to instantiate this provider.
int lcm_get_fileno | ( | lcm_t * | lcm | ) |
Returns a file descriptor or socket that can be used with select()
, poll()
, or other event loops for asynchronous notification of incoming messages.
Each LCM instance has a file descriptor that can be used to asynchronously receive notification when incoming messages have been received. This file descriptor can typically be incorporated into an application event loop (e.g., GTK+, QT, etc.) For an example using select(), see examples/c/listener-async.c
int lcm_handle | ( | lcm_t * | lcm | ) |
Wait for and dispatch the next incoming message.
Message handlers are invoked one at a time from the thread that calls this function, and in the order that they were subscribed.
This function waits indefinitely. If you want timeout behavior, (e.g., wait 100ms for a message) then consider using lcm_get_fileno() together with select() or poll()
Recursive calls to lcm_handle are not allowed – do not call lcm_handle from within a message handler. All other functions are okay (e.g., it is okay to call lcm_publish from within a message handler).
lcm | the LCM object |
int lcm_handle_timeout | ( | lcm_t * | lcm, |
int | timeout_millis | ||
) |
Wait for and dispatch the next incoming message, up to a time limit.
This function is equivalent to lcm_handle(), but if no messages are received and handled by the time timeout_millis
milliseconds elapses, then the function returns.
This function largely exists for convenience, and its behavior can be replicated by using lcm_fileno() and lcm_handle() in conjunction with select() or poll().
New in LCM 1.1.0.
lcm | the LCM object |
timeout_millis | the maximum amount of time to wait for a message, in milliseconds. If 0, then dispatches any available messages and then returns immediately. Values less than 0 are not allowed. |
int lcm_publish | ( | lcm_t * | lcm, |
const char * | channel, | ||
const void * | data, | ||
unsigned int | datalen | ||
) |
Publish a message, specified as a raw byte buffer.
In general, you probably don't want to use this function, as it does not automatically encode messages. Instead, use the message-specific publish function generated by lcm-gen
. Use this function only when you want to publish raw byte buffers.
lcm | The LCM object |
channel | The channel to publish on |
data | The raw byte buffer |
datalen | Size of the byte buffer |
lcm_subscription_t* lcm_subscribe | ( | lcm_t * | lcm, |
const char * | channel, | ||
lcm_msg_handler_t | handler, | ||
void * | userdata | ||
) |
Subscribe a callback function to a channel, without automatic message decoding.
In general, you probably don't want to use this function, as it does not automatically decode messages. Instead, use the message-specific subscribe function generated by lcm-gen
. Use this function only when you want to work with the raw message itself. TODO link to example or more details.
The callback function will be invoked during calls to lcm_handle() any time a message on the specified channel is received. Multiple callbacks can be subscribed for the same channel.
lcm | the LCM object |
channel | the channel to listen on. This can also be a GLib regular expression, and is treated as a regex implicitly surrounded by '^' and '$' |
handler | the callback function to be invoked when a message is received on the specified channel |
userdata | this will be passed to the callback function |
int lcm_subscription_set_queue_capacity | ( | lcm_subscription_t * | handler, |
int | num_messages | ||
) |
Adjusts the maximum number of received messages that can be queued up for a subscription.
In general, you probably don't want to use this function. Instead, use the message-specific set_queue_capacity function generated by lcm-gen
. Use this function only when you want to work with untyped subscriptions. TODO link to example or more details.
Setting this to a low number may reduce overall latency at the expense of dropping more messages. Conversely, setting this to a high number may drop fewer messages at the expense of increased latency. A value of 0 indicates no limit, and should be used carefully.
handler | the subscription object |
num_messages | the maximum queue size, in messages. The default is 30. |
int lcm_unsubscribe | ( | lcm_t * | lcm, |
lcm_subscription_t * | handler | ||
) |
Unsubscribe a message handler.
In general, you probably don't want to use this function. Instead, use the message-specific unsubscribe function generated by lcm-gen
. Use this function only when you want to work with the raw message itself. TODO link to example or more details.
The callback function for the subscription will no longer be invoked when messages on the corresponding channel are received. After this function returns, handler
is no longer valid and should not be used anymore.
handler
is not a valid subscription.