grimoire

Grimoire is a library that extends Phantom.js for development of bots to automate website usage. Major additions to Phantom include:

• Rock-Solid DOM Interaction - A lot of automating website usage is waiting for the DOM to reflect a state, sending mouse/keyboard events to DOM elements. Grimoire adds a robust getSelector method that is used to retrieve structured data from the DOM, test for DOM states (or the negation of states), and reliably identify element viewport locations for mouse / keyboard actions.

• Event Emitter - Grimoire adds Node.js-style event emitter functionality. Phantom allows for the assignment of single handlers to events on a webpage (i.e. onPageResourceRequested, onLoadError). Grimore extends this by allowing the assignment of multiple handlers to these events.

• Web Server API - Phantom ships with a barebones web server module. Grimoire extends this into an RPC API, allowing easy interprocess driven browser sessions, turning Phantom scripts into server daemons, with the ability to webdrive multiple automation sessions in the same process.

• Visual Debugging - Grimoire adds ability to inspect an active browser session at any time, both with a full data dump of state and a visual rendering of the viewport.

• DOM Action Helpers - utility methods for clicking on elements, and setting values on a page.

Setup

Grimoire is an external JS file that needs to be loaded into a Phantomjs context with the injectJS method. It depends on loading several native Phantom modules and third-party JS libraries. Here is an example setup. For now, it's probably easiest just to include this at the top of your Phantom scripts using Grimoire, and suggestions for better design are welcome:

var Webpage = require('webpage')
  , System = require('system')
  , Phantom = phantom
  , CP = require('child_process')
  , FS = require('fs')
  , Server = require('webserver')
;
 
Phantom.injectJs('../node_modules/underscore/underscore-min.js');
Phantom.injectJs('../node_modules/async/dist/async.min.js');
Phantom.injectJs('../node_modules/jsbelt/lib/belt.js');
Phantom.injectJs('../node_modules/moment/min/moment.min.js');
 
var M = {}
  , Async = async
  , Moment = moment;
 
Phantom.injectJs('../node_modules/grimoire/grimoire.js');
M = Grimoire.call(M);

Note that this script uses NPM to manage dependencies. Bower or no dependency manager would be fine as well. After this setup code, add any script-specific code.

M basically represents a singleton instance of the module. It has inherited all the extensions Grimoire provides.

API Reference

createPage(options) - create a new Phantom page (essentially a new session). The page will include a uuid property which can be passed to the web server to retrieve this specific page/session at any point. Options:

• user_agent - pass a specific user agent. Otherwise, the page picks a user agent from a list accessible at UserAgents
• viewport: Optionally pass a viewport size object. i.e. {'width': 2048, 'height': 16000}.
• ttl: Set a time to live in ms for the page. If set, page will close automatically after this amount of time has passed.

Returns new page that was created

loadURL(options, callback(error, page)) - load a URL for a page, optionally waiting until all resources have loaded before invoking a callback Options:

• page - pass in the Phantom page to use
• url - URL to load
• immediate - if true, invoke callback as soon as URL begins to load, rather than waiting for all resources to load
• timeout - time in ms to wait before passing an error to the callback, indicating the page never loaded

getSelector(options, callback(error, result)) - find element or elements in the DOM, meeting different criteria and pass them to callback, or timeout if no such elements are found and pass error to callback. Options:

• selector - CSS selector to look for

• content - look for elements including innerText that matches content, where content is a string, or stringified regex

• page - Phantom page to test

• visible - if true, only look for elements visible in the viewport (Note: some DOM elements do not have coordinates using the getBoundingClientRect method, and will not work with this option.)

• multiple - if true, return all elements fitting these criteria. Otherwise, return first one found in DOM.

• count - if set, invoke callback after count number of elements are found to match criteria

• filter - A function or stringified function that is used to filter elements based on whether function returns truthfully or falsy. Function is passed a DOM element as the only parameter. Note that this function is run the in the Javascript context of the browser page, no the Phantom context. It cannot include variables not otherwise available in the browser context.

• transformer - A function or stringified function that transforms the element passed to it. The element passed is actually a composed object including the following properties:

  • el - the DOM element
  • rect - x, y, width, and height of bounding rectange
  • text - innerText of element
  • outerHTML - outerHtml of element

  As with filter, this function is executed in the browser's Javascript context

• negate - invoke callback only after no elements pass the tests specified

• repeat_interval - pause between polling DOM for this number of ms (default: 100)

• timeout - timeout and pass error to callback after this amount of time in ms (default: 10000)

inspectPage(options, callback) - debugging method used to visually inspect a page and log all page information, optionally pausing script execution Options

• page - page to inspect
• rect - specific rectange to render on inspection (default: entire viewport)
• image_path - path to save JPEG rendering of page to (default: temp path)
• json_path - path to save JSON dump of page content / data (default: temp path)
• pause - if true, pause invoking callback before image viewer is closed
• return_image - if true, return image_path at from function

elementClick(options, callback) - click on an element on the page. This method should be passed an element object from the getSelector method. Options

• element - objects passed to the callback of a getSelector method, must have a rect property that conveys the bounding rectangle
  • x - x centroid-coordinate within the viewport
  • y - y centroid-coordinate within the viewport
  • width - width of bounding rectangle
  • height - height of bounding rectangle
  • top - top of bounding rectangle
• elements - optional, array of elements, all of which will be clicked.
• page - page containing element(s)
• x_offset - by default, the centroid of the element is clicked. Adjusting x_offset shifts the x-coordinate of the specific point clicked
• y_offset - by default, the centroid of the element is clicked. Adjusting y_offset shifts the y-coordinate of the specific point clicked
• move_delay - the delay between moving the mouse to a given point and clicking on the point in ms (default: 0)
• delay - the delay between clicking on an element and invoking the callback. If multiple elements are passed to the method, this delay is invoked between each element click (default: 0)

elementClickCheckbox(options, callback) - click on a checkbox element on the page. Checkboxes are a little tricky as they do not have dimensions in the getBoundingClientRect method. This method should be passed an element object from the getSelector method, which is usually the immediate parent of the checkbox DOM element itself. Options

• element - objects passed to the callback of a getSelector method, must have a rect property that conveys the bounding rectangle
  • x - x centroid-coordinate within the viewport
  • y - y centroid-coordinate within the viewport
  • width - width of bounding rectangle
  • height - height of bounding rectangle
  • top - top of bounding rectangle
• elements - optional, array of elements, all of which will be clicked.
• page - page containing element(s)
• x_offset - by default, the left side of the element's x-axis is clicked. Adjusting x_offset shifts the x-coordinate of the specific point clicked
• y_offset - by default, the centroid of the element's y-axis is clicked. Adjusting y_offset shifts the y-coordinate of the specific point clicked
• move_delay - the delay between moving the mouse to a given point and clicking on the point in ms (default: 0)
• delay - the delay between clicking on an element and invoking the callback. If multiple elements are passed to the method, this delay is invoked between each element click (default: 0)
• click_margin - the offset to the right to click, in order to hit the checkbox element itself (default: 5)

elementEnterText(options, callback) - enter text into an element

elementSelectOption(options, callback) - choose option in select list

startServer(options, callback) - start a webserver for this script, which provides an RPC API for invoking script methods via HTTP request.

Options

• port - port to listen on
• host - accessible host for webserver. This is used to construct a debugging url for each request

Usage

The webserver method converts standalone scripts into modules capable of robust interprocess communication. To execute a method in the script remotely, make a request to /method?method=name-of-method&other_args=..... For example, if the script includes a method called clickRow which requires an index argument to specify the row, requesting /method?method=clickRow&index=5 will execute the clickRow method on the row with index 5.

Web requests will return a JSON object which includes an error key (including the message if any error occurred, and a data key, containing the results of the method. The data object will include page_uuid with a unique identifier for the page created or used with the given method execution. In future requests, this same UUID can be passed as page=[uuid] to retrieve an existing page in executing a given method. In this way, the webserver allows for session-style design patterns, where multiple methods can be executed over time on the same page, with multiple pages being managed by the same script at once.

Additionally, responses will include an inspect_url parameters, with a URL to inspect a specified page visually. When this endpoint is requested, an image is created with the specified page and returned as the response. This is useful in debugging script behavior.

main(options) - this method is called at the end of the script file to parse commandline arguments passed to the script and call methods at load. This method reads in command line arguments and parses matching the pattern --key=value. Passing an argument with the name of another method in the script will cause this method to be executed, passing in any command line arguments as options. For example, a command line call of:

phantomjs somescript.js --startServer --port=5555 --host="http://localhost" --something=else

will execute the startServer method, passing in the following arguments (note: all command line arguments are passed as string values):

{
  'port': '5555'
, 'host': 'http://localhost'
, 'something': 'else'
}