Sonic is a customisable workflow engine with some nice out of the box features.
npm install --save sonic
var Sonic = require('sonic')
var sonic = new Sonic(workflow)
var data = {attr1: 'val1', attr2: 'val2'}
sonic.run(data, function(err) {
// data as modified by sonic
})
The workflow can be seen as a decision tree. An object is passed to the top of the tree through
sonic.run(object, ...)
. Each node can take the decision to modify the object, then reject or pass it to a child
node.
When the data has passed over all the branches (except those from which it were rejected), and if the data was modified,
it is passed again to the top of the node. This makes sure that the data ends in a 'stable' state where all nodes agree
that the data does not need to be modified anymore.
For example, let's say that we have 2 different paths in our workflow. The top node is just a 'fork' node that will make sure our data is injected in the children 'branches'.
Let's assume we have a business logic that deals with tickets.
- the first branch has a filter that checks if the ticket is in 'open' state. Therefore all subsequent actions of this branch will apply to a ticket that is 'opened'. Another filter of this branch only applies if the ticket is marked as 'spam'. If so, the ticket status is changed to 'closed'.
- the second branch only applies to 'closed' tickets and then archive the ticket.
The decision tree would look like this:
[fork]
/ \
/ \
[open] [closed]
/ \
| |
[is spam] [archive]
|
|
[close]
if the [close]
action marks the ticket as 'modified' (using this.next(true)
), the ticket will make it back
to the top of the decision tree, be recognised as [closed]
and end up being archived.
This is the behaviour that you want because it makes sure that the data exits the workflow in a 'stable' state (meaning there is no work left to do)
a custom action can be registered
sonic.register(actionName, function(data, options) {
// data is the data object
// options contain the workflow options:
// options.action => the action name
// options.... => any attribute as defined in the workflow
// pass to the next handler. optional boolean argument if the data has been modified by the current action.
this.next(true|false)
// exit the current 'branch'. optional error object. Passing an error will cause the whole workflow to stop (not
// only the current branch)
this.reject(err)
})
and invoked in the workflow
var workflow = [
{
action: actionName // our custom action name
...
}
]
Only one of this.next()
or this.reject()
can be called in a given action. Calling both will result in an
error thrown and the workflow exited. Also, the data may end up in a corrupted state (unfinished branch execution).
A number of actions are defined by sonic. All these actions contain the sc-
prefix to prevent any collision with
your own defined actions.
Although you can override these actions by registering them yourself (eg. sonic.register('sc-filter', ...)
) it is
recommended that you register your actions with your own prefix so that it can act as a namespace.
Example:
{
"action": "sc-filter",
"name": "entity3, secondPass",
"attributes": {
"name": "entity3",
"secondPass": true,
"success": null
}
}
will filter out (ie. reject from the current workflow branch) any data for which one of the following apply:
- attribute
name
is not strictly equal to"entity3"
- attribute
secondPass
is not strictly equal totrue
- attribute
success
is "truthy"
Attributes can also contain a regular expression. for example:
{
"action": "sc-filter",
"attributes": {
"name": "/^caramel/",
}
}
will filter out any data for which the name
attribute is not a string which starts with caramel
.
Attributes match can be nested. For example:
{
"action": "sc-filter",
"attributes": {
"info.nested": true,
}
}
will filter out any data for which does not have attribute info
which is an object with attribute nested
strictly equal to true
.
Attributes match can use a query like language for matching:
{
"action": "sc-filter",
"attributes": {
"price" : {$gt: 34},
"discount" : {$lt: 0.1}
"status" : {$in: ['open', 'new']}
}
}
Available query parameters are:
$gt
strictly greater than$gte
greater or equal to$lt
strictly less than$lte
less or equal to$in
which value strictly match one of these (array)$nin
which value strictly match none of these (array)
It is also possible to combine multiple query parameters:
{
"action": "sc-filter",
"attributes": {
"price" : {$gt: 34, $lt: 100}
}
}
Example:
{
"action": "sc-set",
"name": "set closed status on spam tickets",
"set": {
"status": "closed",
"spam": true
}
}
Will set the status
to "closed"
and the attribute spam
to true
.
Example:
{
"action": "sc-fork",
"forks": [
[... workflow branch 1 ...],
[... workflow branch 2 ...]
]
}
splits the execution into 2 independent branches. Both branches will share the same data
object and are executed
sequentially:
- first branch 1 is fully executed
- then branch 2 is fully executed
- then the flow resume the parent branch where it left it (except if both branches rejected the data).