The virtual interface can be used as a way to write OS and driver independent tests. Any VirtualBus instances connecting to the same channel (from within the same Python process) will receive each others messages.
If messages shall be sent across process or host borders, consider using the Multicast IP Interface and refer to (the next section) for a comparison and general discussion of different virtual interfaces.
Other Virtual Interfaces¶
There are quite a few implementations for CAN networks that do not require physical CAN hardware. This section also describes common limitations of current virtual interfaces.
The following table compares some known virtual interfaces:
|Within Process||Between Processes||Via (IP) Networks||Without Central Server||Transport Technology||Serialization Format|
||included||✓||✗||✗||✓||Singleton & Mutex (reliable)||none|
||included||✓||✓||✓||✓||UDP via IP multicast (unreliable)||custom using msgpack|
|christiansandberg/ python-can-remote||external||✓||✓||✓||✗||Websockets via TCP/IP (reliable)||custom binary|
|windelbouwman/ virtualcan||external||✓||✓||✓||✗||ZeroMQ via TCP/IP (reliable)||custom binary |
|||The only option in this list that implements interoperability with other languages out of the box. For the others (except the first intra-process one), other programs written in potentially different languages could effortlessly interface with the bus once they mimic the serialization format. The last one, however, has already implemented the entire bus functionality in C++ and Rust, besides the Python variant.|
Guaranteed delivery and message ordering is one major point of difference:
While in a physical CAN network, a message is either sent or in queue (or an explicit error occurred),
this may not be the case for virtual networks.
udp_multicast bus for example, drops this property for the benefit of lower
latencies by using unreliable UDP/IP instead of reliable TCP/IP (and because normal IP multicast
is inherently unreliable, as the recipients are unknown by design). The other three buses faithfully
model a physical CAN network in this regard: They ensure that all recipients actually receive
(and acknowledge each message), much like in a physical CAN network. They also ensure that
messages are relayed in the order they have arrived at the central server and that messages
arrive at the recipients exactly once. Both is not guaranteed to hold for the best-effort
udp_multicast bus as it uses UDP/IP as a transport layer.
Central servers are, however, required by interfaces 3 and 4 (the external tools) to provide
these guarantees of message delivery and message ordering. The central servers receive and distribute
the CAN messages to all other bus participants, unlike in a real physical CAN network.
The first intra-process
virtual interface only runs within one Python process, effectively the
Python instance of
VirtualBus acts as a central server. Notably the
does not require a central server.
Arbitration and throughput are two interrelated functions/properties of CAN networks which are typically abstracted in virtual interfaces. In all four interfaces, an unlimited amount of messages can be sent per unit of time (given the computational power of the machines and networks that are involved). In a real CAN/CAN FD networks, however, throughput is usually much more restricted and prioritization of arbitration IDs is thus an important feature once the bus is starting to get saturated. None of the interfaces presented above support any sort of throttling or ID arbitration under high loads.
import can bus1 = can.interface.Bus('test', bustype='virtual') bus2 = can.interface.Bus('test', bustype='virtual') msg1 = can.Message(arbitration_id=0xabcde, data=[1,2,3]) bus1.send(msg1) msg2 = bus2.recv() assert msg1 == msg2
Bus Class Documentation¶
VirtualBus(channel=None, receive_own_messages=False, rx_queue_size=0, **kwargs)¶
A virtual CAN bus using an internal message queue. It can be used for example for testing.
In this interface, a channel is an arbitrary object used as an identifier for connected buses.
can.VirtualBus._detect_available_configs()for how it behaves here.
The timeout when sending a message applies to each receiver individually. This means that sending can block up to 5 seconds if a message is sent to 5 receivers with the timeout set to 1.0.
This interface guarantees reliable delivery and message ordering, but does not implement rate limiting or ID arbitration/prioritization under high loads. Please refer to the section Other Virtual Interfaces for more information on this and a comparison to alternatives.
Transmit a message to the CAN bus.
Override this method to enable the transmit path.
- msg (Message) – A message object.
- timeout (
float]) – If > 0, wait up to this many seconds for message to be ACK’ed or for transmit queue to be ready depending on driver implementation. If timeout is exceeded, an exception will be raised. Might not be supported by all interfaces. None blocks indefinitely.
can.CanError – if the message could not be sent
Called to carry out any interface specific cleanup required in shutting down a bus.