issues with the new apis (compared to boost bindings)

[ensonic]

= Button =

Nov 05 12:30:28 ev3dev openrobertalab.py[716]: File "/usr/lib/python2.7/threading.py", line 810, in __bootstrap_inner
Nov 05 12:30:28 ev3dev openrobertalab.py[716]: self.run()
Nov 05 12:30:28 ev3dev openrobertalab.py[716]: File "/usr/bin/openrobertalab.py", line 111, in run
Nov 05 12:30:28 ev3dev openrobertalab.py[716]: if self.service.hal.isKeyPressed('back'):
Nov 05 12:30:28 ev3dev openrobertalab.py[716]: File "/usr/lib/python2.7/dist-packages/roberta/ev3.py", line 140, in isKeyPressed
Nov 05 12:30:28 ev3dev openrobertalab.py[716]: return getattr(self.key, key).pressed
Nov 05 12:30:28 ev3dev openrobertalab.py[716]: AttributeError: type object 'Button' has no attribute 'back'

Button.back vs. Button.backspace
I renamed this for now

Nov 05 12:34:03 ev3dev openrobertalab.py[742]: File "/tmp/NEPOprog.py", line 44, in <module>
Nov 05 12:34:03 ev3dev openrobertalab.py[742]: main()
Nov 05 12:34:03 ev3dev openrobertalab.py[742]: File "/tmp/NEPOprog.py", line 40, in main
Nov 05 12:34:03 ev3dev openrobertalab.py[742]: while not hal.isKeyPressed('any'): hal.waitFor(500)
Nov 05 12:34:03 ev3dev openrobertalab.py[742]: File "/usr/lib/python2.7/dist-packages/roberta/ev3.py", line 127, in isKeyPressed
Nov 05 12:34:03 ev3dev openrobertalab.py[742]: if getattr(self.key, key).pressed:
Nov 05 12:34:03 ev3dev openrobertalab.py[742]: AttributeError: 'property' object has no attribute 'pressed'

Button.pressed vs. Button.check_buttons ?
but that'd be repeating the name, since what else should be checked in a Button class.
Same issues with 'buttons_pressed' property on subclasses.

Also why is Buttons.any a property?

Nov 05 13:11:28 ev3dev openrobertalab.py[887]: Traceback (most recent call last):
Nov 05 13:11:28 ev3dev openrobertalab.py[887]: File "/usr/lib/python2.7/threading.py", line 810, in __bootstrap_inner
Nov 05 13:11:28 ev3dev openrobertalab.py[887]: self.run()
Nov 05 13:11:28 ev3dev openrobertalab.py[887]: File "/usr/bin/openrobertalab.py", line 111, in run
Nov 05 13:11:28 ev3dev openrobertalab.py[887]: if self.service.hal.isKeyPressed('back'):
Nov 05 13:11:28 ev3dev openrobertalab.py[887]: File "/usr/lib/python2.7/dist-packages/roberta/ev3.py", line 134, in isKeyPressed
Nov 05 13:11:28 ev3dev openrobertalab.py[887]: return self.keys.check_buttons([key])
Nov 05 13:11:28 ev3dev openrobertalab.py[887]: TypeError: unbound method check_buttons() must be called with Button instance as first argument (got list instance instead)

ev3dev.button became ev3dev.Buttons

= Sound =
since this changed from sync (system) to async.
tone(a,b) becomes tone(a,b).wait()

= Motors =
.state is always []

I expect this to work (works with the boost bindings)

m=ev3dev.LargeMotor(ev3dev.OUTPUT_B)
m.run_to_rel_pos(speed_regulation_enabled='on', position_sp=1000)
while (m.state):
  print m.state

If you change while(m.state) to while(True), it prints [] even though the motor is running.

= Sensors =
.value() return 0

I expect this to work (works with the boost bindings)

s = ev3dev.TouchSensor(ev3dev.INPUT_1)
s.value()

Whether I hold the touch sensor button pressed ot not, it returns 0. Or another one:

s = ev3dev.UltrasonicSensor(ev3dev.INPUT_4)
while(True):
  print s.value()
  time.sleep(0.5)

This prints a static value (191) for me. If instead for any of the sensors I do:

watch -n0.1 cat /sys/class/lego-sensor/sensor*/value0

I see the values changing.

At this point I wonder how you verify the bindings?

[ensonic]

Could this be an issue with the file-descriptor caching?

[ensonic]

All the motor and sensor issues are solved with a 1 liner fix - #68

[ddemidov]

Thanks for testing, I'll look into these. Do you still plan to use both versions of the bindings? I mean, how important is API compatibility at this point?

[ddemidov]

why is Buttons.any a property?

Button implementation has changed significantly from boostc version, now you need to instantiate it in order to use the class.

EDIT: I think I misread your message. Why Button.any being a property is an issue?

[ensonic]

@ddemidov: I have switched over. I think compatibility is hopeless, people will need to do some editing.

Re Button.any, it reads weird, don't you think.

[ddemidov]

Does Button.any_pressed sound better?

[ensonic]

I think it should be a function since the property is computed anyway. The normal methods seems to e Button.check_pressed(), so maybe Button.check_any_pressed().

[rhempel]

I like to think of attributes as "things that you can access as if they were instance variables" - you read and write them. If an attribute accepts a value to write, it does not return a value. The only sideeffect is that the attribute gets the value.

Button.any is a property - returns True/False is any platform specific button is pressed
Button.buttons_pressed is a prpperty, returns a list of button names that are pressed.
Button.check_buttons() is a function that accepts an optional list of buttons to check. True if ALL buttons in the list are checked.

We can go around and around on semantics - we decided Button.any is the Pythonic way to return if any buttons are pressed.

I am happy to discuss alternatives, but at some point we need to make a decision and stick to it - I am no Python expert, and yes Button.any read wierd, but so does:

return any([value.find(pat) >= 0 for pat in pattern])

so I'm OK with Button.any() if we have to...

@ensonic - is there a missing function or attribute in the Button class that is a showstopper at this point?

[ensonic]

Lets then keep the Buttons api as it is as I don't wan't to change it further. The main issue was that it was not documented but different to the previous api. I still prefer functions for things that are computed as calling a function makes you aware that it might 'take a while'. For me window.width is a property, since that has a physical representation. Buttons.any is a predicate.

[rhempel]

I agree with @ensonic re Button.any() - it's not a property and it would be better, I think, if it worked like the Python any() function.
@ensonic - OK if I change this part of the API?

[ensonic]

@rhempel yes please and thanks :)