#Vim(for Chrome)

Vim for Chrome is a project to bring the vim editor to Google Chrome as an app.

##Commands

The crucial bit, at least with respect to central functionality, is how commands are executed and defined.

To move to insert mode:

vim.exec('i');

To type hello:

vim.exec(['h', 'e', 'l', 'l', 'o']);

Obviously, this provides a simple way to route keystrokes into the application.But it also facilitates creating new commands:

Capital "O" might be defined like:

vim.exec(['k', 'o']);

##Extensibility

The actual extension API is like:

vim.extend({
	'insert': { //name of mode you are extending
		'O': function() {
			vim.exec(['k', 'o']);
		}
	},
	on: { //event-based actions
		'insert': function() { //event fired by vim on switch to insert
			vim.notify('-- INSERT --');
		}
	}
});

A little more dry:

###Extension object:

extensionObject = {
	{
		mode: {
		 	{
			'command': function(value) {
				//run when that command is called in that mode
			},
		...
	},
	on: {
		'eventName': function(subject) { ... };
		...
	}
};

• mode is a string, but can be any mode.
• command is a string, BUT will be executed as a regular expression if a literal match is not found. The command function is given the command value in any case.

###What about syntax + highlighting ?

Maybe the question is: how does this structure allow for syntax formatting and highlighting ?

I think that the answer is pretty straightforward : create a listener for the "stroke" event(doesn 't exist), which will get passed the actual line that changed. Check the line for your keywords, then when you find them draw out the element associated with each character and add a class ('function ') if the keyword is 'function ' for later handling by a stylesheet. That seems about as easy/efficient as it could be, though obviously that would then be exposed in a more friendly way.

##Status

Just revisiting this (spring 2013) since starting it last winter. Need to evaluate whether (1) the architecture will actually scale to the needs of vim (hunch: yes) and (2) need to find a better solution for rendering. The backbone paradigm, I think, breaks down with this. We end up with a trillion 'models' for characters when fundamentally we're just trying to map a string of text to the screen with some extra meta information. So that needs some thought, and a rewrite based on those conclusions.