Table Of Contents
Kivy is able to handle most types of input: mouse, touchscreen, accelerometer, gyroscope, etc. It handles the native multitouch protocols on the following platforms: Tuio, WM_Touch, MacMultitouchSupport, MT Protocol A/B and Android.
The global architecture can be viewed as:
Input providers -> Motion event -> Post processing -> Dispatch to Window
The class of all input events is the MotionEvent. It generates 2 kinds of events:
- Touch events: a motion event that contains at least an X and Y position. All the touch events are dispatched across the Widget tree.
- No-touch events: all the rest. For example, the accelerometer is a continuous event, without position. It never starts or stops. These events are not dispatched across the Widget tree.
A Motion event is generated by an Input Provider. An Input Provider is responsible for reading the input event from the operating system, the network or even from another application. Several input providers exist, such as:
- TuioMotionEventProvider: create a UDP server and listen for TUIO/OSC messages.
- WM_MotionEventProvider: use the windows API for reading multitouch information and sending it to Kivy.
- ProbeSysfsHardwareProbe: In Linux, iterate over all the hardware connected to the computer, and attaches a multitouch input provider for each multitouch device found.
- and much more!
When you write an application, you don’t need to create an input provider. Kivy tries to automatically detect available hardware. However, if you want to support custom hardware, you will need to configure kivy to make it work.
Before the newly-created Motion Event is passed to the user, Kivy applies post-processing to the input. Every motion event is analyzed to detect and correct faulty input, as well as make meaningful interpretations like:
- Double/triple-tap detection, according to a distance and time threshold
- Making events more accurate when the hardware is not accurate
- Reducing the amount of generated events if the native touch hardware is sending events with nearly the same position
After processing, the motion event is dispatched to the Window. As explained previosuly, not all events are dispatched to the whole widget tree: the window filters them. For a given event:
Motion event profiles¶
Depending on your hardware and the input providers used, more information may be made available to you. For example, a touch input has an (x,y) position, but might also have pressure information, blob size, an acceleration vector, etc.
A profile is a string that indicates what features are available inside the motion event. Let’s imagine that you are in an on_touch_move method:
def on_touch_move(self, touch): print touch.profile return super(..., self).on_touch_move(touch)
The print could output:
Many people mix up the profile’s name and the name of the corresponding property. Just because 'angle' is in the available profile doesn’t mean that the touch event object will have an angle property.
For the 'pos' profile, the properties pos, x, and y will be available. With the 'angle' profile, the property a will be available. As we said, for touch events 'pos' is a mandatory profile, but not 'angle'. You can extend your interaction by checking if the 'angle' profile exists:
def on_touch_move(self, touch): print 'The touch is at position', touch.pos if 'angle' in touch.profile: print 'The touch angle is', touch.a
You can find a list of available profiles in the motionevent documentation.
A touch event is a specialized MotionEvent where the property is_touch evaluates to True. For all touch events, you automatically have the X and Y positions available, scaled to the Window width and height. In other words, all touch events have the 'pos' profile.
You must take care of matrix transformation in your touch as soon as you use a widget with matrix transformation. Some widgets such as Scatter have their own matrix transformation, meaning the touch must be multiplied by the scatter matrix to be able to correctly dispatch touch positions to the Scatter’s children.
You must use one of them to scale coordinates correctly to the context. Let’s look the scatter implementation:
def on_touch_down(self, touch): # push the current coordinate, to be able to restore it later touch.push() # transform the touch coordinate to local space touch.apply_transform_2d(self.to_local) # dispatch the touch as usual to children # the coordinate in the touch is now in local space ret = super(..., self).on_touch_down(touch) # whatever the result, don't forget to pop your transformation # after the call, so the coordinate will be back in parent space touch.pop() # return the result (depending what you want.) return ret
If the touch has a shape, it will be reflected in the ‘shape’ property. Right now, only a ShapeRect can be exposed:
from kivy.input.shape import ShapeRect def on_touch_move(self, touch): if isinstance(touch.shape, ShapeRect): print 'My touch have a rectangle shape of size', \ (touch.shape.width, touch.shape.height) # ...
A double tap is the action of tapping twice within a time and a distance. It’s calculated by the doubletap post-processing module. You can test if the current touch is one of a double tap or not:
def on_touch_down(self, touch): if touch.is_double_tap: print 'Touch is a double tap !' print ' - interval is', touch.double_tap_time print ' - distance between previous is', touch.double_tap_distance # ...
A triple tap is the action of tapping thrice within a time and a distance. It’s calculated by the tripletap post-processing module. You can test if the current touch is one of a triple tap or not:
def on_touch_down(self, touch): if touch.is_triple_tap: print 'Touch is a triple tap !' print ' - interval is', touch.triple_tap_time print ' - distance between previous is', touch.triple_tap_distance # ...
Grabbing touch events¶
It’s possible for the parent widget to dispatch a touch event to a child widget from within on_touch_down, but not from on_touch_move or on_touch_up. This can happen in certain scenarios, like when a touch movement is outside the bounding box of the parent, so the parent decides not to notify its children of the movement.
But you might want to do something in on_touch_up. Say you started something in the on_touch_down event, like playing a sound, and you’d like to finish things on the on_touch_up event. Grabbing is what you need.
When you grab a touch, you will always receive the move and up event. But there are some limitations to grabbing:
- You will receive the event at least twice: one time from your parent (the normal event), and one time from the window (grab).
- You might receive an event with a grabbed touch, but not from you: it can be because the parent has sent the touch to its children while it was in the grabbed state.
- The touch coordinate is not translated to your widget space because the touch is coming directly from the Window. It’s your job to convert the coordinate to your local space.
Here is an example of how to use grabbing:
def on_touch_down(self, touch): if self.collide_point(*touch.pos): # if the touch collides with our widget, let's grab it touch.grab(self) # and accept the touch. return True def on_touch_up(self, touch): # here, you don't check if the touch collides or things like that. # you just need to check if it's a grabbed touch event if touch.grab_current is self: # ok, the current touch is dispatched for us. # do something interesting here print 'Hello world!' # don't forget to ungrab ourself, or you might have side effects touch.ungrab(self) # and accept the last up return True