PhpFlo: Flow-based programming for PHP
PhpFlo is a simple flow-based programming implementation for PHP. It is a PHP port of NoFlo, a similar tool for Node.js. From WikiPedia:
In computer science, flow-based programming (FBP) is a programming paradigm that defines applications as networks of "black box" processes, which exchange data across predefined connections by message passing, where the connections are specified externally to the processes. These black box processes can be reconnected endlessly to form different applications without having to be changed internally. FBP is thus naturally component-oriented.
Developers used to the Unix philosophy should be immediately familiar with FBP:
This is the Unix philosophy: Write programs that do one thing and do it well. Write programs to work together. Write programs to handle text streams, because that is a universal interface.
It also fits well in Alan Kay's original idea of object-oriented programming:
I thought of objects being like biological cells and/or individual computers on a network, only able to communicate with messages (so messaging came at the very beginning -- it took a while to see how to do messaging in a programming language efficiently enough to be useful).
The system has been heavily inspired by J. Paul Morrison's book Flow-Based Programming.
PhpFlo is still quite experimental, but may be useful for implementing flow control in PHP applications.
Installing
PhpFlo can be installed from Packagist.org with the composer package manager. Just ensure your composer.json
has the following:
{
"require": {
"phpflo/phpflo": "^1.0.0"
}
}
and run:
$ wget http://getcomposer.org/composer.phar
$ php composer.phar install
or do a quick php composer.phar require phpflo/phpflo
from within your existing project.
Autoloading
To use PhpFlo, you need a PHP Standards Group -compatible autoloader. You can use the Composer-supplied autoloader:
<?php
require 'vendor/autoload.php';
Running the examples
File line count using embedded PhpFlo:
$ ./examples/linecount/count.php somefile.txt
Terminology
- Component: individual, pluggable and reusable piece of software. In this case a PHP class implementing
PhpFlo\Common\ComponentInterface
- Graph: the control logic of a FBP application, can be either in programmatical or file format
- Inport: inbound port of a component
- Network: collection of processes connected by sockets. A running version of a graph
- Outport: outbound port of a component
- Process: an instance of a component that is running as part of a graph
Components
A component is the main ingredient of flow-based programming. Component is a PHP class providing a set of input and output port handlers. These ports are used for connecting components to each other.
PhpFlo processes (the boxes of a flow graph) are instances of a component, with the graph controlling connections between ports of components.
Structure of a component
Functionality a component provides:
- List of inports (named inbound ports)
- List of outports (named outbound ports)
- Handler for component initialization that accepts configuration
- Handler for connections for each inport
A minimal component would look like the following:
<?php
use PhpFlo\Component;
class Forwarder extends Component
{
protected $description = "This component receives data on a single input port and sends the same data out to the output port";
public function __construct()
{
// Register ports
$this->inPorts()->add('in', ['datatype' => 'all']);
$this->outPorts()->add('out', ['datatype' => 'all']);
// Forward data when we receive it
$this->inPorts()->in->on('data', array($this, 'forward'));
// Disconnect output port when input port disconnects
$this->inPorts()->in->on('disconnect', array($this, 'disconnect'));
}
public function forward($data)
{
$this->outPorts()->out->send($data);
}
public function disconnect()
{
$this->outPorts->out->disconnect();
}
}
Alternatively you can now use PhpFlo\Common\ComponentTrait
together with ```PhpFlo\Common\ComponentInterface````if you're more into composition.
This example component register two ports: in and out. When it receives data in the in port, it opens the out port and sends the same data there. When the in connection closes, it will also close the out connection. So basically this component would be a simple repeater.
You can find more examples of components in the lib/PhpFlo/Component
folder.
Please mind that there's an mandatory second parameter for the "add" command. This array receives the port's meta information and has following defaults:
$defaultAttributes = [
'datatype' => 'all',
'required' => false,
'cached' => false,
'addressable' => false,
];
This is but a subset of the available attributes, a noflo port can take.
- datatype defines the "to be expected" datatype of the dataflow. Currently all datatypes from noflo are implemented
- required not implemented yet
- cached not implemented yet
- addressable decides if a port needs to be either an instance of Port (false) or ArrayPort (true)
Defining the datatype is mandatory, since there is a port matching check during graph building, according to this matrix:
out\in | all | bang | string | bool | number | int | object | array | date | function |
---|---|---|---|---|---|---|---|---|---|---|
all | x | x | ||||||||
bang | x | x | ||||||||
string | x | x | x | |||||||
bool | x | x | x | |||||||
number | x | x | x | |||||||
int | x | x | x | x | ||||||
object | x | x | x | |||||||
array | x | x | x | |||||||
date | x | x | x | |||||||
function | x | x | x |
These types are only implicitly checked. There is no data validation during runtime!
Some words on component design
Components should aim to be reusable, to do one thing and do it well. This is why often it is a good idea to split functionality traditionally done in one function to multiple components. For example, counting lines in a text file could happen in the following way:
- Filename is sent to a Read File component
- Read File reads it and sends the contents onwards to Split String component
- Split String splits the contents by newlines, and sends each line separately to a Count component
- Count counts the number of packets it received, and sends the total to a Output component
- Output displays the number
This way the whole logic of the application is in the graph, in how the components are wired together. And each of the components is easily reusable for other purposes.
If a component requires configuration, the good approach is to set sensible defaults in the component, and to allow them to be overridden via an input port. This method of configuration allows the settings to be kept in the graph itself, or for example to be read from a file or database, depending on the needs of the application.
The components should not depend on a particular global state, either, but instead attempt to keep the input and output ports their sole interface to the external world. There may be some exceptions, like a component that listens for HTTP requests or Redis pub-sub messages, but even in these cases the server, or subscription should be set up by the component itself.
Ports and events
Being a flow-based programming environment, the main action in PhpFlo happens through ports and their connections. There are five events that can be associated with ports:
- Attach: there is a connection to the port
- Connect: the port has started sending or receiving a data transmission
- Data: an individual data packet in a transmission. There might be multiple depending on how a component operates
- Disconnect: end of data transmission
- Detach: A connection to the port has been removed
It depends on the nature of the component how these events may be handled. Most typical components do operations on a whole transmission, meaning that they should wait for the disconnect event on inports before they act, but some components can also act on single data packets coming in.
When a port has no connections, meaning that it was initialized without a connection, or a detach event has happened, it should do no operations regarding that port.
Graph file format
In addition to using PhpFlo in embedded mode where you create the FBP graph programmatically (see example), you can also initialize and run graphs defined using a FBP file. This format gives you the advantage of much less definition work, compared to the deprecated (but still valid) JSON files.
If you have older JSON definitions, you can still use them or convert then to FBP, using the dumper wrapped by the graph or directly from definition:
$builder = new \PhpFlo\Builder\ComponentFactory();
$network = new PhpFlo\Network($builder);
$network->boot(__DIR__.'/count.json', $builder);
file_put_contents('./count.fbp', $network->getGraph()->toFbp());
The PhpFlo FBP files declare the processes used in the FBP graph, and the connections between them. The file format is shared between PhpFlo and NoFlo, and looks like the following:
ReadFile(ReadFile) out -> in SplitbyLines(SplitStr)
ReadFile(ReadFile) error -> in Display(Output)
SplitbyLines(SplitStr) out -> in CountLines(Counter)
CountLines(Counter) count -> in Display(Output)
Other supported formats are YAML and JSON, but keep in mind that the FBP domain specific language (DSL) should be the way to go if you want to use something like the noflo ui.
JSON example:
{
"properties": {
"name": "Count lines in a file"
},
"processes": {
"ReadFile": {
"component": "ReadFile"
},
"SplitbyLines": {
"component": "SplitStr"
},
"CountLines": {
"component": "Counter"
},
"Display": {
"component": "Output"
}
},
"connections": [
{
"src": {
"process": "ReadFile",
"port": "out"
},
"tgt": {
"process": "SplitbyLines",
"port": "in"
}
},
{
"src": {
"process": "ReadFile",
"port": "error"
},
"tgt": {
"process": "Display",
"port": "in"
}
},
{
"src": {
"process": "SplitbyLines",
"port": "out"
},
"tgt": {
"process": "CountLines",
"port": "in"
}
},
{
"src": {
"process": "CountLines",
"port": "count"
},
"tgt": {
"process": "Display",
"port": "in"
}
}
]
}
YAML example:
properties:
name: 'Count lines in a file'
initializers: { }
processes:
ReadFile:
component: ReadFile
metadata: { label: ReadFile }
SplitbyLines:
component: SplitStr
metadata: { label: SplitStr }
Display:
component: Output
metadata: { label: Output }
CountLines:
component: Counter
metadata: { label: Counter }
connections:
-
src: { process: ReadFile, port: OUT }
tgt: { process: SplitbyLines, port: IN }
-
src: { process: ReadFile, port: ERROR }
tgt: { process: Display, port: IN }
-
src: { process: SplitbyLines, port: OUT }
tgt: { process: CountLines, port: IN }
-
src: { process: CountLines, port: COUNT }
tgt: { process: Display, port: IN }
To run a graph file, load it via the PhpFlow API:
<?php
$builder = new PhpFlo\Builder\ComponentFactory();
// create network
$network = new PhpFlo\Network($builder);
$network
->boot(__DIR__.'/count.fbp')
->run($fileName, "ReadFile", "source")
->shutdown();
Note that after this the graph is live, meaning that you can add and remove nodes and connections, or send new initial data to it. See example.
Since the network now also features the HookableInterface
, you can easily add callbacks on events for e.g. debugging purposes:
<?php
$builder = new PhpFlo\Builder\ComponentFactory();
// create network
$network = new PhpFlo\Network($builder);
$network
->hook(
'data',
'trace',
function ($data, $socket) {
echo $socket->getId() . print_r($data, true) . "\n";
}
)
->boot(__DIR__.'/count.fbp')
->run($fileName, "ReadFile", "source")
->shutdown();
As you can see, there's a lot of potential in the callbacks, since they can also use object references to store and/or manipulate data, but natively receive socket and data from the supported events.
This feature is also used in the upcoming phpflo/flowtrace library, which decorates the Network
class and adds PSR-3 compatible logging.
Development
PhpFlo development happens on GitHub. Just fork the main repository, make modifications and send a pull request.
To run the unit tests you need PHPUnit. Run the tests with in development:
$ bin/phpunit
Some ideas
- Use phpDaemon to make the network run asynchronously, Node.js -like