valence.js
is an interface for controlling github's electron GUI toolkit from another process. It is inspired by the Thrust project.
Essentially, electron is a node.js process connected to a chromium process. The idea is that you write node.js javascript code to control a chromium process in order to build applications that look and feel like native applications.
Valence is a protocol for communicating between this node.js process and another controller process, and valence.js
is the node.js component of the implementation. valence.js
runs in the electron main process as depicted in this diagram:
|------------| stdio |--------------| IPC |----------------|
| Your app |--------->| electron |---->| electron |
| |<---------| main process |<----| render process |
|------------| |--------------| |----------------|
perl/whatever valence.js chromium
valence.js
is compatible with electron versions 0.25.1 through 1.4.1 (and probably later).
Why have a separate controller process at all? Why not just write the controller logic in javascript and run it in the electron main process?
First of all, the obvious reason is that not everybody wants to write substantial controller logic in javascript. Different languages have different strengths and libraries that may not be available in javascript/node.js. For example, when communicating with SQL databases it's hard to beat perl's DBI module (see AnyEvent::DBI or AnyEvent::Task for how to use DBI in an async program).
Secondly, sometimes we already have significant existing programs written in another language that we would like to add a GUI front-end to. Rather than re-write such apps in javascript, valence.js
provides a "glue" option for other environments to use electron.
Finally, even if your app is written in javascript, in order to use electron directly, your app needs to support the exact version of node
that electron is currently compiled with. This can especially be an issue with native modules that depend on older node
APIs. With valence.js
you can use any node
environment that is applicable to your application -- well, once we have a javascript driver that is :).
The reference implementation of the app-side of the valence protocol is written in perl 5.
It can be installed with the following cpan minus command (use --sudo
if you with to install it for all users on your machine):
$ cpanm Valence --sudo
After it is installed, see the Valence documentation for how to use it. To work on the code itself, please fork it on github.
valence.js
is an electron app and you can run it as you would any other:
/path/to/electron /path/to/valence/app/
After this, the app will wait until it receives messages on standard input, and will subsequently emit messages over its standard output. All of your app's communication with valence.js
is done over electron's stdio pipes.
All messages (both input and output) are minified JSON, followed by a new-line. This means there is exactly one message per line.
Every method has a cmd
parameter which indicates the "command" to be invoked by the message. The other parameters depend on which command was sent.
APP -> VALENCE
Indicates that valence should call a method on a particular object.
Parameters:
-
obj
: The object id (seesave
below) corresponding to the object that the method should be called on. Optional. If omitted, assumed to benode.js
'smodule
object. -
method
: The name of the method to be invoked. Required. -
save
: The object id to assign to whatever is returned from the method call. Typically this is a counter that is maintained by the app which is incremented for every method call. Optional. If omitted, the return result is discarded. -
args
: An array of arguments that should be passed to the method. Required (but can be empty). -
args_cb
: An array of arrays. Each array represents a callback insertion that should be applied toargs
above. The first element in each sub-array is the offset in args where a callback should be inserted, and the second element is the callback id (see thecb
command). Optional.
APP -> VALENCE
Indicates that the app is finished with an object and valence can discard its reference to it.
Parameters:
obj
: The object id of the object to be destroyed. This should have been passed in via a previoussave
parameter. Required.
APP -> VALENCE
Used to lookup an attribute from an object and save it in a new object.
Parameters:
-
obj
: The object id of which to lookup an attribute from. Required. -
attr
: The name of the attribute. Required. -
save
: The object id to assign to the result. Optional (but pointless without).
APP -> VALENCE
Retrieves a value stored in an object id by having it passed to a callback.
Parameters:
-
obj
: The object id of the object to retrieve. Required. -
cb
: The callback id (see thecb
command) to invoke with the result. Required.
VALENCE -> APP
Sent by valence.js
when it wishes to invoke a callback in your application.
Parameters:
-
cb
: The callback id to be invoked. This was passed in from acall
orget
command. Required. -
args
: An array of arguments to be passed to the callback. Required (but may be empty).
Here is the detailed description of a few selected messages.
Note that the JSON in these examples has been "prettified" and in the actual valence protocol they must be minified and one-per-line.
When experimenting, if you are using the perl module you can set the VALENCE_DEBUG
environment variable to sniff the traffic between a perl app and valence.js
(see the Valence docs).
This example is of an app calling a method which contains a callback. It is roughly equivalent to executing this code:
get_object(3).on('closed', function() {
notify_callback(4);
});
Here is what the message might look like:
{
"args" : [
"closed",
null
],
"args_cb" : [
[
1,
4
]
],
"cmd" : "call",
"method" : "on",
"obj" : "3"
}
The obj
above is 3
which presumably was the object resulting from the creation of a BrowserWindow object. We are attaching a callback to listen for closed
events. The callback id is 4
, and its location in the args array is 1
. Note that the callback has been stubbed out with a null
value and will be replaced by an actual javascript function by valence.js
, and how there is no save
parameter because we aren't interested in the returned result of the on
method.
This is what valence.js
might send when the closed
event was triggered:
{
"args" : [
{}
],
"cb" : 4,
"cmd" : "cb"
}
The callback id is 4
(see the previous example). There is an empty hash in the args
list. This corresponds to the javascript event, but has been converted into an empty hash because the node.js
JSON.stringify()
method could not serialize this value.
-
Currently the
args_cb
values can only reference the first level of the arguments and cannot replace a callback inside a nested structure like an object or array. Eventually the protocol should support location specifiers such as[1][3]['field']
-
It needs to spec out behaviour for what happens when exceptions are thrown.
-
The protocol should expose a way to eval raw javascript in the main process (you can already do it in the render process with
WebContents.executeJavaScript()
). -
The special
new
method can currently only support 1 parameter because of a limitation in javascript. -
valence.js
currently doesn't know when a callback it has installed has been garbage collected. If it did then it could send a "callback destroy" command to the app. This may be possible with the npm weak module.
Some examples that use Valence: DB-Browser, HackRF-RCCar
Doug Hoyte, [email protected]
Copyright 2015-2016 Doug Hoyte.
This project is licensed under the 2-clause BSD license.
Electron itself is Copyright (c) 2014-2016 GitHub Inc. and is licensed under the MIT license.