I am working through the traits
presentation from PyCon 2010. At about 2:30:45 the presenter starts covering trait event notifications, which allow (among other things) the ability to automatically call a subroutine any time a trait
has changed.
I am running a modified copy of the example he gave... In this trial, I am trying to see whether I can fire a static event whenever I make a change to volume
or inputs
.
from traits.api import HasTraits, Range, List, Float
import traits
class Amplifier(HasTraits):
"""
Define an Amplifier (a la Spinal Tap) with Enthought's traits. Use traits
to enforce values boundaries on the Amplifier's objects. Use events to
notify via the console when the volume trait is changed and when new volume
traits are added to inputs.
"""
volume = Range(value=5.0, trait=Float, low=0.0, high=11.0)
inputs = List(volume) # I want to fire a static trait event notification
# when another volume element is added
def __init__(self, volume=5.0):
super(Amplifier, self).__init__()
self.volume = volume
self.inputs.append(volume)
def _volume_changed(self, old, new):
# static event listener for self.volume
if not (new in self.inputs):
self.inputs.append(self.volume)
if new == 11.0:
print "This one goes to eleven... so far, we have seen", self.inputs
def _inputs_changed(self, old, new):
# static event listener for self.inputs
print "Check it out!!"
if __name__=='__main__':
spinal_tap = Amplifier()
spinal_tap.volume = 11.0
print "DIRECTLY adding a new volume input..."
spinal_tap.inputs.append(4.0)
try:
print "NEGATIVE Test... adding 12.0"
spinal_tap.inputs.append(12.0)
except traits.trait_errors.TraitError:
print "Test passed"
When I run this script, I can see This one goes to eleven... so far, we have seen [5.0, 11.0]
in the console output, so I know that _volume_changed()
gets fired when I assign 11.0
to spinal_tap.volume
.
However, I never see any events from _inputs_changed()
. No matter what example I cook up, I can't get a List
to fire an event.
This is the output I am seeing... note that there is no evidence that _inputs_changed()
ever fires.
[mpenning@Bucksnort ~]$ python spinaltap.py
This one goes to eleven... so far, we have seen [5.0, 11.0]
DIRECTLY adding a new volume input...
NEGATIVE Test... adding 12.0
Test passed
[mpenning@Bucksnort ~]$
I have run this both under Python2.6 / Cygwin / Windows 7 and Python 2.5 / Linux (all using traits
version 4.0.0 that I easy_install
directly off Enthought's site). The results are the same no matter what I have tried so far.
Should a List
be able to fire a static event when using traits? If so, am I doing something wrong?
After browsing their unit tests, I found a test for Dict
traits in enthought's event unittest coverage... it looks like when you have a container like a Dict
or List
that you need to set up the magic event listener method like this:
## Broken method definition: def _inputs_changed(self, old, new):
# container event static listeners must be in the form of _foo_items_changed()
def _inputs_items_changed(self, old, new):
# static event listener for self.inputs
if len(new.added) > 0:
print "Check it out, we added %s to self.items" % new.added
elif len(new.removed) > 0:
print "Check it out, we removed %s from self.items" % new.removed
Likewise, I also discovered that the on_trait_change
decorator (used for dynamic traits
event notification) requires similar nomenclature if you are calling it with a traits.api.List
or traits.api.Dict
... so I could also write the code above as:
from traits.api import on_trait_change
# ...
@on_trait_change('inputs_items')
def something_changed(self, name, new):
# static event listener for self.inputs
if len(new.added) > 0:
print "Check it out, we added %s to self.items" % new.added
elif len(new.removed) > 0:
print "Check it out, we removed %s from self.items" % new.removed
Either way, when I run the code, I get expected output:
[mpenning@Bucksnort ~]$ python spinaltap.py
Check it out, we added [5.0] to self.items
Check it out, we added [11.0] to self.items
This one goes to eleven... so far, we have seen [5.0, 11.0]
DIRECTLY adding a new volume input...
Check it out, we added [4.0] to self.items
NEGATIVE Test... adding 12.0
Test passed
[mpenning@Bucksnort ~]$
As this also caught me out recently, I've just verified Mike Pennington's answer with Traits 4.2.1. There does seem to be a distinction between changes to the List trait itself (such as assigning a new list to it), and changes to the membership of the List (such as appending or setting by index). The former uses the same name as the trait (e.g. inputs
), whereas the latter uses the "_items" suffix. This example seems to demonstrate this:
from traits.api import Float, HasTraits, Instance, List
class Part(HasTraits):
costs = List(Float)
# called when the actual List trait changes:
def _costs_changed(self, old, new):
print("Part::_costs_changed %s -> %s" % (str(old), str(new)))
# called when the contents of the List trait changes:
def _costs_items_changed(self, old, new):
print("Part::_costs_changed %s -> %s" % (str(old), str(new)))
class Widget(HasTraits):
part = Instance(Part)
def __init__(self):
self.part = Part()
self.part.on_trait_change(self.update_costs, 'costs')
self.part.on_trait_change(self.update_costs_items, 'costs_items')
def update_costs(self, name, new):
print("update_costs: %s = %s" % (name, str(new),))
def update_costs_items(self, name, new):
print("update_costs_items: %s = %s" % (name, str(new),))
w = Widget()
w.part.costs = [ 1.0, 2.0, 3.0 ]
# Part::_costs_changed [] -> [1.0, 2.0, 3.0]
# update_costs: costs = [1.0, 2.0, 3.0]
w.part.costs = [ 1.0, 2.0, 3.1 ]
# Part::_costs_changed [1.0, 2.0, 3.0] -> [1.0, 2.0, 3.1]
# update_costs: costs = [1.0, 2.0, 3.1]
w.part.costs[0] = 5.0
# Part::_costs_changed <undefined> -> <traits.trait_handlers.TraitListEvent object at 0x1007bd810>
# update_costs_items: costs_items = <traits.trait_handlers.TraitListEvent object at 0x1007bd810>
w.part.costs.append(4.0)
# Part::_costs_changed <undefined> -> <traits.trait_handlers.TraitListEvent object at 0x1007bd810>
# update_costs_items: costs_items = <traits.trait_handlers.TraitListEvent object at 0x1007bd810>
This behaviour is hinted at in the documentation here.
However if an extended name is used it does seem possible to have the same handler called when the entire list or membership is changed:
from traits.api import Float, HasTraits, Instance, List
class Part(HasTraits):
costs = List(Float)
def _costs_changed(self, old, new):
print("_costs_changed %s -> %s" % (str(old), str(new)))
def _costs_items_changed(self, old, new):
print("_costs_items_changed %s -> %s" % (str(old), str(new)))
class Widget(HasTraits):
part = Instance(Part)
def __init__(self):
self.part = Part()
self.part.on_trait_change(self.update_costs, 'costs[]') # <-- extended name
def update_costs(self, name, new):
print("update_costs: %s = %s" % (name, str(new),))
w = Widget()
w.part.costs = [ 1.0, 2.0, 3.0 ]
# _costs_changed [] -> [1.0, 2.0, 3.0]
# update_costs: costs = [1.0, 2.0, 3.0]
w.part.costs = [ 1.0, 2.0, 3.1 ]
# _costs_changed [1.0, 2.0, 3.0] -> [1.0, 2.0, 3.1]
# update_costs: costs = [1.0, 2.0, 3.1]
w.part.costs[0] = 5.0
# _costs_items_changed <undefined> -> <traits.trait_handlers.TraitListEvent object at 0x1007c6f90>
# update_costs: costs_items = [5.0]
w.part.costs.append(4.0)
# _costs_items_changed <undefined> -> <traits.trait_handlers.TraitListEvent object at 0x1007c6f90>
# update_costs: costs_items = [4.0]
In this case, the name
parameter of the update_costs
handler can be used to differentiate between the container itself changing, or a single item within the container changing.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With