Flow uses SBT as build system. To get started, include a dependency to flow-core in your project:
libraryDependencies += "com.github.jodersky" %% "flow" % "2.2.2"
NOTICE: flow uses native libraries to back serial communication, therefore before you can run any application depending on flow you must include flow’s native library! To do so, you have two options.
In case your OS/architecture combination is present in the table below, add a second dependency to your project:
libraryDependencies += "com.github.jodersky" % "flow-native" % "2.2.2"
OS | Architecture | Notes |
---|---|---|
Linux | x86 x86_64 ARM (v7) |
A user accessing a serial port will probably need to be in the dialout group. |
Mac OS X | x86_64 |
This will add a jar to your classpath containing native libraries for various platforms. At run time, the correct library for the current platform is selected, extracted and loaded. This solution enables running applications seamlessly, as if they were pure JVM applications. However, since the JVM does not enable full determination of the current platform (only OS and rough architecture are known), only a couple of platforms can be supported through this solution at the same time.
Do not include the second dependency above. Instead, for every end-user application that relies on flow, manually add the native library for the current platform to the JVM’s library path. This can be achieved through various ways, notably:
Per application: Run your program with the command-line option -Djava.library.path=".:<folder containing libflow3.so>"
. E.g. java -Djava.library.path=".:/home/<folder containing libflow3.so>" -jar your-app.jar
System- or user-wide:
Copy the native library to a place that is on the default Java library path and run your application normally. Such places usually include /usr/lib
and /usr/local/lib
.
Use a provided installer (currently Debian archive and Mac .pkg, available in releases)
The native library can either be obtained by building flow (see section Build) or by taking a pre-compiled one, found in releases in the GitHub project. Native libraries need to be of major version 3 to work with this version of flow.
It is recommended that you use the first option only for testing purposes or end-user applications. The second option is recomended for libraries, since it leaves more choice to the end-user.
The following is a general guide on the usage of flow. If you prefer a complete example, check out the code contained in the flow-samples
directory.
Flow’s API follows that of an actor based system, where each actor is assigned specific functions involved in serial communication. The two main actor types are:
Serial “manager”. The manager is a singleton actor that is instantiated once per actor system, a reference to it may be obtained with IO(Serial)
. It is typically used to open serial ports (see following section).
Serial “operators”. Operators are created once per open serial port and serve as an intermediate between client code and native code dealing with serial data transmission and reception. They isolate the user from threading issues and enable the reactive dispatch of incoming data. A serial operator is said to be “associated” to its underlying open serial port.
The messages understood by flow’s actors are all contained in the com.github.jodersky.flow.Serial
object. They are well documented and should serve as the entry point when searching the API documentation.
A serial port is opened by sending an Open
message to the serial manager. The response varies on the outcome of opening the underlying serial port.
In case of failure, the serial manager will respond with a CommandFailed
message to the original sender. The message contains details on the reason to why the opening failed.
In case of success, the sender is notified with an Opened
message. This message is sent from an operator actor, spawned by the serial manager. It is useful to capture the sender (i.e. the operator) of this message as all further communication with the newly opened port must pass through the operator.
import com.github.jodersky.flow.{ Serial, SerialSettings, AccessDeniedException }
val port = "/dev/ttyXXX"
val settings = SerialSettings(
baud = 115200,
characterSize = 8,
twoStopBits = false,
parity = Parity.None
)
IO(Serial) ! Serial.Open(port, settings)
def receive = {
case Serial.CommandFailed(cmd: Serial.Open, reason: AccessDeniedException) =>
println("You're not allowed to open that port!")
case Serial.CommandFailed(cmd: Serial.Open, reason) =>
println("Could not open port for some other reason: " + reason.getMessage)
case Serial.Opened(settings) => {
val operator = sender
//do stuff with the operator, e.g. context become opened(op)
}
}
Writing data is as simple as sending a Write
message to an operator. The data to send is an instance of akka.util.ByteString
:
operator ! Serial.Write(data)
Optionally, an acknowledgement for sent data can be requested by adding an ack
parameter to a Write
message. The ack
parameter is of type Int => Serial.Event
, i.e. a function that takes the number of actual bytes written and returns an event. Note that “bytes written” refers to bytes enqueued in a kernel buffer; no guarantees can be made on the actual transmission of the data.
case class MyPacketAck(wrote: Int) extends Serial.Event
operator ! Serial.Write(data, MyPacketAck(_))
operator ! Serial.Write(data, n => MyPacketAck(n))
def receive = {
case MyPacketAck(n) => println("Wrote " + n + " bytes of data")
}
The actor that opened a serial port (referred to as the client), exclusively receives incomming messages from the operator. These messages are in the form of akka.util.ByteString
s and wrapped in a Received
object.
def receive = {
case Serial.Received(data) => println("Received data: " + data.toString)
}
A port is closed by sending a Close
message to its operator:
operator ! Serial.Close
The operator will close the underlying serial port and respond with a final Closed
message before terminating.
The operator has a deathwatch on the client actor that opened the port, this means that if the latter crashes, the operator closes the port and equally terminates, freeing any allocated resources.
The opposite is not true by default, i.e. if the operator crashes (this can happen for example on IO errors) it dies silently and the client is not informed. Therefore, it is recommended that the client keep a deathwatch on the operator.
As of version 2.2.0, flow can watch directories for new files. On most unix systems this can be used for watching for new serial ports in /dev/
. Watching happens through a message-based, publish-subscribe protocol as explained in the sections below.
A client actor may watch – i.e subscribe to notifications on – a directory by sending a Watch
command to the serial manager.
Should an error be encountered whilst trying to obtain the watch, the manager will respond with a CommandFailed
message. Otherwise, the client may be considered “subscribed” to the directory and the serial manager will thenceforth notify the client on new files.
IO(Serial) ! Serial.Watch("/dev/")
def receive = {
case Serial.CommandFailed(w: Watch, reason) =>
println(s"Cannot obtain a watch on ${w.directory}: ${reason.getMessage}")
}
Whilst subscribed to a directory, a client actor is informed of any new files in said directory by receiving Connected
messages from the manager.
def receive = {
case Serial.Connected(port) if port matches "/dev/ttyUSB\\d+" =>
// do something with the available port, e.g.
// IO(Serial) ! Open(port, settings)
}
Unsubscribing from events on a directory is done by sending an Unsubscribe
message to the serial manager.
IO(Serial) ! Unwatch("/dev/")
Note that the manager has a deathwatch on every subscribed client. Hence, should a client die, any underlying resources will be freed.
Flow uses Java’s WatchService
s under the hood, therefore a Java runtime of a version of at least 1.7 is required.
A complete build of flow involves two parts
Building Scala sources (the front-end), resulting in a platform independent artifact (i.e. a jar file).
Building C sources (the back-end), yielding a native library that may only be used on systems resembling the platform for which it was compiled
Both steps are independent, their only interaction being a header file generated by the JDK utility javah
(see sbt javah
for details), and may therefore be built in any order.
Run sbt flow/packageBin
in the base directory. This simply compiles Scala sources as with any standard SBT project and packages the resulting class-files in a jar.
The back-end is managed by GNU Autotools and all relevant files are contained in flow-native
.
Several steps are involved in producing the native library:
When compiling for the first time, initialize Autotools by running the script ./bootstrap
.
Once initialized, ./configure && make
will build the back-end.
The native library is now ready and can be:
copied to a local directory: DESTDIR=$(pwd)/<directory> make install
installed system-wide: make install
put into a “fat” jar, useful for dependency management with SBT (see next section)
The native library produced in the previous step may be bundled into a “fat” jar so that it can be included in SBT projects through its regular dependency mechanisms. In this process, SBT basically acts as a wrapper script around Autotools, calling the native build process and packaging generated libraries. Running sbt flow-native/packageBin
in the base directory produces the fat jar in flow-native-sbt/target
.
Note: an important feature of fat jars is to include native libraries for several platforms. To copy binaries compiled on other platforms to the fat jar, place them in a subfolder of flow-native-sbt/lib_native
. The subfolder should have the name $(os.name)-$(os.arch)
, where os.name
and os.arch
are the Java system properties of the respective platforms.
The project and package versions follow a sematic pattern: M.m.p
, where
M
is the major version, representing backwards incompatible changes
m
is the minor version, indicating backwards compatible changes such as new feature additions
p
is the patch number, representing internal modifications such as bug-fixes
Usually (following most Linux distribution’s conventions), shared libraries produced by a project name
of version M.m.p
are named libname.so.M.m.p
. However, since when accessing shared libraries through the JVM, only the name
can be specified and no particular version, the convention adopted by flow is to append M
to the library name and always keep the major version at zero. E.g. libflow.so.3.1.2
becomes libflow3.so.0.1.2
.