python classes getters and setters static method and class method - python

class Spam(object):
#a_string = 'candy'
def __init__(self, sold=0, cost=0):
self.sold = sold
self.cost = cost
#staticmethod
def total_cost():
return True
#classmethod
def items_sold(cls, how_many):
#property
def silly_walk(self):
return print (self.a_string)
#silly_walk.setter
def silly_walk(self, new_string):
self.a_string = new_string.upper()
def do_cost(self):
if self.total_cost():
print('Total cost is:', self.cost)
.
from spam import Spam
def main ():
cost = 25
sold = 100
a_string = 'sweets'
sp = Spam(100, 25)
sp.do_cost()
sw = Spam.silly_walk(a_string)
sw.silly_walk()
if __name__ == '__main__':
main()
so im new to python and i don't understand how to use the setters and getters in this. so what i want to do is:
use #property to create a setter and getter for a property named silly_walk. Have the setter upper case the silly_walk string.
Show example code that would access the static method.
Show example code that would use the silly_walk setter and getter.
im getting very confused with what "self" does in the class and im not sure if what im doing is correct
update:
problem was the #classmethod not having a return and indentation error, so everything is fixed thanks everybody

self is convention. Since you're inside a class, you don't have functions there you have methods. Methods expect a reference to the object calling them as the first argument, which by convention is named self. You can call it anything you like.
class Foo(object):
def __init__(itsa_me_maaaario, name):
itsa_me_maaario.name = "Mario"
That works just as well.
As for the rest of your code -- what's your QUESTION there? Looks like your setter is a bit weird, but other than that it should work mostly okay. This is better:
class Spam(object): # inherit from object in py2 for new-style classes
def __init__(self, a_string, sold=0, cost=0) # put the positional arg first
...
#staticmethod
def total_cost():
# you have to do something meaningful here. A static method can't access
# any of the objects attributes, it's really only included for grouping
# related functions to their classes.
#classmethod
def items_sold(cls, how_many):
# the first argument to a classmethod is the class, not the object, so
# by convention name it cls. Again this should be something relevant to
# the class not to the object.
#property
def silly_walk(self):
return self.a_string
# don't call itself.
#silly_walk.setter
def silly_walk(self, new_string):
self.a_string = new_string
# it really just hides the attribute.
For instance I have a class I built to abstract a computer system I'm in charge of. It might be something like:
class System(object):
type_ = "Base system"
def __init__(self, sitenum, devicenum, IP):
self._sitenum = sitenum
self._devicenum = devicenum
self._IP = IP
# the leading underscores are a flag to future coders that these are
# "private" variables. Nothing stopping someone from using it anyway,
# because System()._IP is still that attribute, but it makes it clear
# that they're not supposed to be used that way.
#staticmethod
def ping_system(IP):
subprocess.call(["ping",IP], shell=True) # OH GOD SECURITY FLAW HERE
# group this with Systems because maybe that's how I want it? It's an
# aesthetic choice. Note that this pings ANY system and requires an
# argument of an IP address!
#classmethod
def type_of_system(cls):
return cls.type_
# imagine I had a bunch of objects that inherited from System, each w/
# a different type_, but they all inherit this....
#property
def description(self):
return "Site {}, Device {} # {}".format(self._sitenum,
self._devicenum,
self._IP)
#description.setter
def description(self, *args):
if len(args) == 3:
self._sitenum, self._devicenum, self._IP = args
elif len(args) == 1 and len(args[0]) == 3:
self._sitenum, self._devicenum, self._IP = args[0]
else:
raise ValueError("Redefine description as Sitenum, Devicenum, IP")
Example:
computer = System(1, 1, '192.168.100.101')
System.ping_system('192.160.100.101') # works
computer.type_of_system # "Base system"
computer.description # "Site 1, Device 1 # 192.168.100.101"
new_description = [1, 2, '192.168.100.102']
computer.description = new_description
# invokes description.setter
computer._devicenum # is 2 after the setter does its magic.

Related

What is the proper way to write getter and setter in Python?

I am from the C# background.
I am trying to write getter and setters methods. In other words, I am trying to create properties for a class.
class ParamDefinition:
def __init__(self, type_name):
self.__type_name = type_name
#property
def get_type_name(self):
return self.__type_name
#get_type_name.setter
def set_type_name(self, type_name):
self.__type_name = type_name
def printf(self):
print(self.__type_name)
def test():
compDef = ParamDefinition(None)
compDef.type_name = "int"
compDef.printf()
if __name__ == "__main__":
test()
Output:
None
What is incorrect with my property-definition?
Both the getter and setter need to have the same name:
class ParamDefinition:
def __init__(self, type_name):
self.__type_name = type_name
#property
def type_name(self):
return self.__type_name
#type_name.setter
def type_name(self, type_name):
self.__type_name = type_name
def printf(self):
print(self.__type_name)
def test():
compDef = ParamDefinition(None)
compDef.type_name = "int"
compDef.printf()
if __name__ == "__main__":
test()
Output:
int
EDIT:
While this isn't technically part of the question I thought I should add some more thoughts. You say you are from a C# background, and in C# it's fairly normal to make every field a property by default since you can add extra logic later without breaking the interface, and allows reflection.
These advantages don't really exist with Python and as you can see the boilerplate needed to turn a field into a property is much more significant than in C#. Therefore I'd recommend not using properties in Python until it's really what you need.

What's the best approach to defer class attribute initialization?

I have the following OO use-case:
class SomeClass(object):
def __init__(self, arg1):
self.attr_now = arg1
self.attr_later = None
def compute(self):
# do some lengthy computation ...
self.attr_later = x # some result
# continue work ...
Is this an ideal approach to defer initialization of attribute attr_later or is there a better way? What I see is the type unsafety of the attr_later as it may change type during the lifespan of the enclosing instance ...
Since you are talking about the OO use case, I think you must know about the access modifications you can do to protect the attribute for its value and type.
Think of this
class SomeClass(object):
def __init__(self, arg1):
self.__attr_now = arg1
self.__attr_later = None
#property
def attr_later(self):
return self.__attr_later
def compute(self):
# do some lengthy computation ...
self.__attr_later = x # some result
# continue work ...
The instance variables can be prefixed with a __ for making it private to that instance. So you are sure, the variable cannot be changed from outside. More over, create a property for the variable, so that you can access the value from outside.
If you think, you want to set the value from outside, define a setter and control the incoming assignment - something like
#attr_later.setter
def attr_later(self, attr_val):
if isinstance(attr_val,str):
self.__attr_later = attr_val
else:
raise
# or do damage control
You are in control of your instance, if you properly encapsulate the properties. Hope this makes some sense

How To Call A Class Function From A Variable

I need a class router (for lack of a better word). The router needs to instantiate a class & call a function of that class instance based on variables passed to it.
How do I properly define the class function?
How do I properly call the class function?
Example code:
class ClassWorker1:
def function_1(self):
print('1a')
def function_2(self):
print('2a')
def function_3(self):
print('3a')
class ClassWorker2:
def function_1(self):
print('1b')
def function_2(self):
print('2b')
def function_3(self):
print('3b')
class ClassRouter(object):
def __init__(self, class_name, class_function):
self.class_instance = class_name()
self.class_function = class_function
self.main()
def main(self):
# how should I call the class function here?
self.class_instance.class_function()
return
a = 1
b = 1
if a == 1:
class_name = ClassWorker1
else:
class_name = ClassWorker1
if b == 1:
# Strings won't work as class function definition
# I won't know the class at this point. I will only know
# the shared function name at this point.
# how could this class function be defined directly?
class_function = 'function_1'
elif b == 2:
class_function = 'function_2'
else:
class_function = 'function_3'
ClassRouter(class_name, class_function)
I need a class router (for lack of a better word).
Are you sure you need a class for this ?
The router needs to instantiate a class & call a function of that class instance
When it belongs to a class or instance, a function is usually named a "method". Not really important but it makes things clearer. Also, an "instance" is obviously always, by definition, an instance of a class ;)
How do I properly define the class function?
How do I properly call the class function?
Does the router really have to be a class ? But anyway...
There are a couple distinct issues here (I of course assume you need something that's generic enough).
The first one is that your class (the one that will be instanciated by the "router") constructor may need some args - position or named or both. If it's the router's duty to instanciate the class (but should it be ?), you'll have to pass those args (both position and named) to the router. And since your router has to be generic (else it's useless) you cannot explicitely name these args in your router's constructor.
Hopefully, Python has a way to "unpack" tuples (for position args) and dicts (for named args) when calling a function, using respectively the * and ** operators at call time, ie:
def somefunc(arg1, arg2, arg3="foo", arg4=None):
print arg1, arg2, arg3, arg4
args = ("one", "two", "three")
kwargs = {"arg4": "four"}
somefunc(*args, **kwargs)
This let you pass arguments to a function in a generic way.
So if you want your router to be in charge of instanciating the "target" class, you'll have to support this:
class Router(object):
def __init__(self, cls, clsargs=None, clskwargs=None):
if clsargs is None:
clsargs = ()
if clskwargs is None:
clskwargs = {}
self._obj = cls(*clsargs, **clskwargs)
class Worker(object):
def __init__(self, name):
self.name = name
print self.name
r = Router(Worker, clsargs=("foo",))
# or
r = Router(Worker, clskwargs={"name":"foo"})
Now note that at this point you really don't gain anything (except for more code) from having the router instanciating the Worker - since you need to have the Worker class and it's constructor's args to instanciate the router, you could as well just instanciate the Worker yourself and pass the Worker instance to the router:
Since you must have a reference to the class passed to the router (else you can't pass it ), you could as well
class Router(object):
def __init__(self, obj):
self._obj = obj
class Worker(object):
def __init__(self, name):
self.name = name
print self.name
r = Router(Worker("foo"))
# or
r = Router(Worker(name="foo"))
The cases where it would make sense to have the router instanciate the worker are:
1/ if the Worker's constructor arguments are not known when the router is instanciated and are to be passed later (which requires a distinct router method to pass those args)
2/ if the Worker's instanciation is very costly and you're not even sure you'll really need it, in which case you want to wait until the router's "main" method is called to instanciate the worker.
The second issue is "how do I get the worker's method by name". This one has already been answered by Lukas: you use getattr(obj, attrname).
The third issue is "if my worker method needs arguments, how do I pass them". This is the same problem as with the worker's constructor arguments, so the solution is obviously the same. Depending on the concrete use case, you'll have to pass those args either when instanciating the router or when calling it's "main" method.
wrt/ this "main" method, remember that you can define your own callable types by implementing the __call__ method, ie
class NotAFunc(object):
def __init__(self, wot):
self.wot = wot
def __call__(self, count):
print self.wot * count
notafunc = NotAFunc("wot ? ")
notafunc(42)
So it might make sense to use this as your router's "main" method
Now do you really need a router class at all ? Python functions are object on their own (so a function can take a function and/or return a function), and moreover act as closures (a closure is a function that "captures" part of the environment where it's been defined):
def route(instance, methodname, methargs=None, methkwargs=None):
method = getattr(instance, methodname)
if methargs is None:
methargs = ()
if methkwargs is None:
methkwargs = {}
def func():
return method(*methargs, **methkwargs)
return func
class Worker(object):
def __init__(self, name):
self.name = name
def work(self, count):
return [self.name for i in range(count)]
r = route(Worker("foo"), "work", (42,))
print r()
Note that while I kept your "router" term, most of what I described above are known patterns. You may want to search for "proxy", "proxy method", and (for the last exemple) "partial evaluation".
You are looking for dynamic attribute lookup.
class C:
def c1(self, x):
return 2*x
instance = C()
method_name = 'c1'
method = getattr(instance, method_name)
print(method(1)) # call method and print result
You'll need to override the __new__ method of your (new-style!) class.
class ClassRouter(object):
def __new__(self, class_name, *args):
if arg=="Class1":
new_instance = ClassWorker1(*args)
new_instance.method()
return new_instance
elif arg=="Class2":
return ClassWorker2(*args)

correct class inheritance

I've written some code that allows the term 'job' to be used universally to perform a unique task. The specific jobs can be chosen through setting an initial variable "job_type". From that initial variable a particulay subclass is chosen to perform the appropriate job. Maybe the code will make more sense :)
if __name__=='__main__':
# these variables would normally be called in from a config file
job_type = 'job1'
uni_var = 10
job_select = superClass(job_type, uni_var)
job_select.job()
class superClass(object):
def __init__(self, job_type, uni_var):
self.job_type = job_type
self.uni_var = uni_var
if self.job_type == 'job1':
self.jobChoice = option1()
else:
self.jobChoice = option2()
# This is the definition called by the main function it then
# redirects the request to the appropriate job sub class
def job(self):
self.jobChoice.job()
class option1(superClass):
def __init__(self):
pass
def job(self):
print 'job option 1'
print uni_var
class option2(superClass):
def __init__(self):
pass
def job(self):
print 'job option 2'
print uni_var
The thought behind this code was to allow a single/constant 'main' function, to action a variety of unique tasks based purely on the variable 'job_type'. Which it seems to be doing fine.
My question (as a very inexperienced coder) is, have I gone about this the right way or is there a better way to do things?
Also, have I set up the variable 'uni_var' correctly in the superClass to be correctly shared across all/any superClass subclasses?
Thanks.
I suspect that what you really want is to use the Factory Method Pattern here.
You could change your code to something like this:
if __name__=='__main__':
# these variables would normally be called in from a config file
job_type = 'job1'
uni_var = 10
job_select = superClass.optionFactory(job_type, uni_var)
job_select.job()
class superClass(object):
def __init__(self, job_type, uni_var):
self.job_type = job_type
self.uni_var = uni_var
# This is the definition called by the main function it then
# redirects the request to the appropriate job sub class
def job(self):
raise NotImplementedError()
#staticmethod
def optionFactory(job_type, uni_var):
"Return an instance of superClass based on job_type and uni_var."
if job_type == "job1":
return option1(job_type, uni_var)
else:
return option2(job_type, uni_var)
class option1(superClass):
def __init__(self, job_type, uni_var):
super(option1, self).__init__(job_type, uni_var)
def job(self):
print 'job option 1'
print uni_var
class option2(superClass):
def __init__(self, job_type, uni_var):
super(option2, self).__init__(job_type, uni_var)
def job(self):
print 'job option 2'
print uni_var
However, notice that this implementation will require that superClass be changed every time a new subclass is created. Another alternative would be to make the optionFactory method a standalone function (rather than a method of superClass). Like this:
if __name__=='__main__':
# these variables would normally be called in from a config file
job_type = 'job1'
uni_var = 10
job_select = optionFactory(job_type, uni_var)
job_select.job()
class superClass(object):
def __init__(self, job_type, uni_var):
self.job_type = job_type
self.uni_var = uni_var
# This is the definition called by the main function it then
# redirects the request to the appropriate job sub class
def job(self):
raise NotImplementedError()
class option1(superClass):
def __init__(self, job_type, uni_var):
super(option1, self).__init__(job_type, uni_var)
def job(self):
print 'job option 1'
print uni_var
class option2(superClass):
def __init__(self, job_type, uni_var):
super(option2, self).__init__(job_type, uni_var)
def job(self):
print 'job option 2'
print uni_var
def optionFactory(job_type, uni_var):
"Return an instance of superClass based on job_type and uni_var."
if job_type == "job1":
return option1(job_type, uni_var)
else:
return option2(job_type, uni_var)
Don't use a class as a factory, it's silly. You just need a consistent interface across implementations.
class JobA(object):
def do_job(self, arg):
print 'job a', arg
class JobB(object):
def do_job(self, arg):
print 'job b', arg
job_types = {
'job1': JobA, 'job2': JobB
}
job_type = 'job1'
uni_var = 10
job = job_types[job_type]()
job.do_job(uni_var)
Hell, if the jobs don't keep state, they shouldn't be classes, either, but functions instead.
def job_a(arg):
...
def job_b(arg):
...
job = job_types[job_type]
job(uni_var)
You are actually not using inheritance at all in your original code, and your code contains a severe bug, but masks it from appearing.
You create an instance of superClass, which stores job_type and uni_var, and then instantiates either option1 or option2 and stores a reference to that. The "subclasses" are initialised with no data (so they have no job_type or uni_var attributes, and you have to override __init__ to do-nothing methods to avoid error).
When you call job_select.job(...), superClass explicitly delegates to its job_choice attribute. So there's no useful inheritance going on; you've overridden everything about superClass in your subclasses, and instead having job_select be an instance of different sub-classes depending on the job_type and using method resolution to call the right code, job_select is always a superClass which contains an option1 or option2 and explicitly delegates to it.
The serious bug I mentioned: neither option1 nor option2 actually contains any information about the job, so their job methods can't do anything interesting. You call print uni_var, which doesn't work (it would normally be print self.uni_var to get the uni_var of this job), but seems to work because you have a global called uni_var. As soon as you start doing anything more complicated this scheme will fall over badly.
#srgerg's answer is a good example of what you could do to actually use inheritance and factory functions to solve your problem. #CatPlusPlus's answer is a good example of how you can use more appropriate tools for the very simple code in this example (may not be what you need if your real requirements are more complex what you've got written at the moment).

Virtual classes: doing it right?

I have been reading documentation describing class inheritance, abstract base classes and even python interfaces. But nothing seams to be exactly what I want. Namely, a simple way of building virtual classes. When the virtual class gets called, I would like it to instantiate some more specific class based on what the parameters it is given and hand that back the calling function. For now I have a summary way of rerouting calls to the virtual class down to the underlying class.
The idea is the following:
class Shape:
def __init__(self, description):
if description == "It's flat": self.underlying_class = Line(description)
elif description == "It's spiky": self.underlying_class = Triangle(description)
elif description == "It's big": self.underlying_class = Rectangle(description)
def number_of_edges(self, parameters):
return self.underlying_class(parameters)
class Line:
def __init__(self, description):
self.desc = description
def number_of_edges(self, parameters):
return 1
class Triangle:
def __init__(self, description):
self.desc = description
def number_of_edges(self, parameters):
return 3
class Rectangle:
def __init__(self, description):
self.desc = description
def number_of_edges(self, parameters):
return 4
shape_dont_know_what_it_is = Shape("It's big")
shape_dont_know_what_it_is.number_of_edges(parameters)
My rerouting is far from optimal, as only calls to the number_of_edges() function get passed on. Adding something like this to Shape doesn't seam to do the trick either:
def __getattr__(self, *args):
return underlying_class.__getattr__(*args)
What I am doing wrong ? Is the whole idea badly implemented ? Any help greatly appreciated.
I agree with TooAngel, but I'd use the __new__ method.
class Shape(object):
def __new__(cls, *args, **kwargs):
if cls is Shape: # <-- required because Line's
description, args = args[0], args[1:] #  __new__ method is the
if description == "It's flat": # same as Shape's
new_cls = Line
else:
raise ValueError("Invalid description: {}.".format(description))
else:
new_cls = cls
return super(Shape, cls).__new__(new_cls, *args, **kwargs)
def number_of_edges(self):
return "A shape can have many edges…"
class Line(Shape):
def number_of_edges(self):
return 1
class SomeShape(Shape):
pass
>>> l1 = Shape("It's flat")
>>> l1.number_of_edges()
1
>>> l2 = Line()
>>> l2.number_of_edges()
1
>>> u = SomeShape()
>>> u.number_of_edges()
'A shape can have many edges…'
>>> s = Shape("Hexagon")
ValueError: Invalid description: Hexagon.
I would prefer doing it with a factory:
def factory(description):
if description == "It's flat": return Line(description)
elif description == "It's spiky": return Triangle(description)
elif description == "It's big": return Rectangle(description)
or:
def factory(description):
classDict = {"It's flat":Line("It's flat"), "It's spiky":Triangle("It's spiky"), "It's big":Rectangle("It's big")}
return classDict[description]
and inherit the classes from Shape
class Line(Shape):
def __init__(self, description):
self.desc = description
def number_of_edges(self, parameters):
return 1
Python doesn't have virtual classes out of the box. You will have to implement them yourself (it should be possible, Python's reflection capabilities should be powerful enough to let you do this).
However, if you need virtual classes, then why don't you just use a programming language which does have virtual classes like Beta, gBeta or Newspeak? (BTW: are there any others?)
In this particular case, though, I don't really see how virtual classes would simplify your solution, at least not in the example you have given. Maybe you could elaborate why you think you need virtual classes?
Don't get me wrong: I like virtual classes, but the fact that only three languages have ever implemented them, only one of those three is still alive and exactly 0 of those three are actually used by anybody is somewhat telling …
You can change the class with object.__class__, but it's much better to just make a function that returns an instance of an arbitrary class.
On another note, all class should inherit from object unless you use using Python 3, like this, otherwise you end up with an old-style class:
class A(object):
pass

Categories