extension-poco is a native extension for Defold to enable test automation from a computer to a test device (desktop or mobile).

This extension implements the Poco api, using the Cocos2Dx-lua Lua server, with some small customization on top (see defold-poco.lua)

To use this library in your Defold project, add the following URL to your game.project dependencies:

https://github.com/defold/extension-poco/archive/master.zip

We recommend using a link to a zip file of a specific release.

The Poco api is compatible with both Python 2.7 and Python 3.3-3.6.

Install the Poco library (from documentation):

pip install pocoui

See main.script for an example.

Note that you can also use this in the gui script context as well.

Add a require to one of your Defold scripts

local poco = require('poco.lua.defold-poco')

Create a PocoManager instance at game startup:

function init(self)
    poco:init_server(15004) -- default port for Poco
end

Update the PocoManager instance continuously:

function update(self, dt)
    poco:server_loop()
end

You also need to set the view projection matrices (e.g. on resize event). The default values are the identity matrix.

poco:set_view_proj(get_view_proj()) -- view projection matrix from render script
poco:set_gui_view_proj(get_gui_view_proj()) -- view projection matrix from render script

Here's an example using rendercam:

local rendercam = require "rendercam.rendercam"

local function get_view_proj()
    local view = rendercam.calculate_view()
    local proj = rendercam.calculate_proj()
    return proj * view
end

local function get_gui_view_proj()
    local mat = vmath.matrix4_orthographic(0, rendercam.window.x, 0, rendercam.window.y, -1, 1)
    return mat
end

You can add a custom function that you can call from the client scripts:

Server:

poco:set_dispatch_fn("TestNewFunction", function(x, y, z)
        print("INSIDE NEW CLIENT TEST FUNCTION!", x, y, z)
        return {message= "" .. (x + y + z)}
    end)

Client:

cbk = poco.agent.rpc.call("TestNewFunction", 1, 2, 3)
cbk.wait()

You can also add a hook to an existing function, in order to manipulate the result before it's returned to the client.

poco:set_dispatch_callback_fn("Dump", function(scene)
    pprint("INSIDE NEW CALLBACK FUNCTION!", scene)
    return scene
end)

If you're using a predefined ipaddress, you can use a VirtualDevice:

device = VirtualDevice(ipaddress)

You create a StdPoco instance using an ip address and a port:

poco = StdPoco(addr=(ipaddress, port), device=device, use_airtest_input=False)

If you are using Android, you don't know the ipaddress of the device, you can use a device = None and ipaddress 0 0.0.0.0.

The default port for the poco server in the extension is 15004

See the examples for practical examples of how to create a Poco instance.

You can get the full scene graph:

ui = poco.agent.hierarchy.dump()

To click on an object named button_node:

poco("button_node").click()

Special keys are of the format {KEY_name}. Other characters are treated as text. (NOTE: Currently no unicode support)

poco.agent.input.keyevent('[email protected]{KEY_BACKSPACE}m')

Examples

This extension also contains a test project, which we can test against.

Launch the test project on your device and then run one of the test scripts.

How to test against a local build on the host computer:

$ python2 ./examples/example_dump.py

Connect to a certain ip (any device type):

$ python2 ./examples/example_dump.py -a <ip address>

Connect to an Android device via adb (if you don't know its ipaddress)

$ python2 ./examples/example_dump.py -p android

For list of all arguments:

$ python2 ./examples/example_dump.py -h

There are several resources on how to write a test script here:

• AirTestProject/Poco

• Poco - Examples and Tutorials

While the Defold scene graph inspection api strives to reconstruct the scene as it once was in the Defold Editor, it is not possible to accurately do so to 100%. Some of the data is not avaiable once in runtime format.

Each component type also has to implement its own hooks into the inspection api, so if the component type you're looking for isn't seen in the output graph, this might be the reason.