Time: Wed 10/4/17 3 pm
Similar to the elementary data abstraction, just that we can change values in the encapsulation now somehow.
What is nonlocal? And why is it useful in this case? Can we just bind some variable in the parent frame to some new values without using this keyword? Do we need to have something in a nonlocal frame with the name beforehand?
Here's something really bad to do as an example:
def make_bike():
# The color of the bike
color = "red"
def get_color():
return color
def change_color(new_color):
# Change the color of the bike
color = new_color
# However now you thought you forgot to use nonlocal
# But note that it doesn't hoist
nonlocal color # Invalid syntax
# Python will just give up with your code :(
# ... and literally halt here
return [get_color, change_color]
[get_color, change_color] = make_bike() # Make a bike
change_color("blue") # Make it blue, don't like it being redAnd a slightly better example (in which there is no syntax error, but logically incorrect):
def make_factory():
# The color of the bike by default
factory_color = "green"
def make_bike():
# Use the factory color as default
color = factory_color
def get_color():
return color
def change_color(new_color):
# Someone hacked into the script and changed the following lines
nonlocal factory_color
factory_color = new_color
return [get_color, change_color]
return make_bike
make_bike = make_factory()
# Make a new bike
[get_0_color, change_0_color] = make_bike()
# Inspect its color
get_0_color() # green
# And now we want to repaint the bike
change_0_color("blue")
# And when we try to inspect the bike again...
get_0_color() # green
# Uhm it is still green :/
# Make another bike
[get_1_color, change_1_color] = make_bike()
# Try to inspect its color... which should be green
get_1_color() # blue
# Oh well, the factory settings are modified... uh ohCan we still implement something with a same interface that does exactly the same without nonlocal, using list, dict etc.?
There's something called global; however, Python doesn't act as picky as it does to nonlocal. When it sees something undefined paired up with global, it just doesn't complain. More tears...
Can we use global rather than nonlocal for everything? Please don't.
Primitively, they are just syntactical sugars for data abstraction. What are methods, bound methods and static methods?
A quick example of each:
class balloon():
@staticmethod
def describe():
print("Balloon!")
def grow(self):
print("The balloon grew larger...")
b = balloon()
balloon # <class '__main__.balloon'>
b # <__main__.balloon object at ...>
# Static method
balloon.describe # <function balloon.describe at ...>
balloon.describe() # Balloon!
b.describe # <function balloon.describe at ...>
b.describe() # Balloon!
# Unbound method
balloon.grow # <function balloon.grow at ...>
balloon.grow(b) # The balloon grew larger...
# Bound method
b.grow # <bound method balloon.grow of <__main__.balloon object at ...>>
b.grow() # The balloon grew larger...Class variables and instance variables are different. Class variables are shared between instances; however, instance variables are just different from instance to instance.
class sandbag():
weight = 10
# Someone goes to buy a sandbag
s = sandbag()
# He weighs the sandbag...
s.weight # 10
# And finds that it is of 10 kilos
# And another person comes into the story and does something terrible
sandbag.weight = 5
# The next day, whoever that bought the sandbag tries to re-weigh it
s.weight # 5
# And somehow... It became lighter :(How do you make custom constructors for a class?
There are some goodies and complications that come with objects too. Multiple inheritance is a beast--check out the diamond problem--and it can bring ambiguities in programming. There are ways to probe into the method resolution order.
What's super, and why do we use it?
They are naturally sustainable! Oh well, think about how trees
are conceptually related to pairs.
Define the terminologies around trees--tree, root, label, branch, leaf (the language we use may vary in different places).
Consider the following implementation:
def tree(label, branches=[]):
return [label] + branches
def label(tree):
return tree[0]
def branches(tree):
return tree[1:]There are multiple ways to get started on playing with such tree structure! As a few examples: