Also has not been updated for a long time, the main thing to write today is about the object-oriented properties of the problem.
The explanation of the property is that the method can be simply understood as a method but a field is called. In addition, the other is no adorner, using Python to provide the way to write, according to your different objects to use, in fact, is also the definition of the method, to use the object call.
1 classObject:2 defF1 (self):3 return '123'4 defF2 (self, value):5 Print(value)6 defF3 (self):7A = 1 + 18 Print(a)9Foo = property (Fget=f1, Fset=f2, fdel=f3)Ten Onep =Object () Aresult =P.foo - Print(Result) -P.foo ='Week' the delP.foo
View Code
One is to define the method first, followed by the Fget, Fset, Fdel of the adorner in the way that corresponds to the method. Then it is called with the object. It's just that this way feels more elegant.
PS: Parallel and distributed issues, discussed today under the main grasp a bit of parallel results of mutual influence, distributed results do not affect each other.
This is similar to doing a eight-step thing, parallel is the work required to divide the workload into eight parts, and distributed is equivalent to divided into eight sections, each to do, and then collide with the results. For example, gender (male and female), body (height), body shape (fat thin), eyesight (wearing glasses, not wearing) this if the non-distributed look is 2*2*2*2, the training set is not enough may not be able to determine whether the person wearing glasses. It is equivalent to the message is passed back to the past, but there is loss or distortion in the middle, and not enough to verify may not be allowed. Or, judging whether to wear glasses is dependent on the last condition of the body of the combination of things down. Did not do beyond or cross the previous node condition, according to the given algorithm sequence down, can not directly and gender "communication" or judgment. And distributed is, directly divided into (male, female, fat, thin, high, short, wearing, not wearing) These eight attributes, free to learn collision. This is, for example, the equivalent of direct communication, not completely rely on the middle of the transfer, do not need a lot of training set to fix (quite a few times we need to communicate with each other I trust you this person, you bring the upper information is true, distributed directly can communicate with the upper layer collision). But to learn from each other, to verify (direct ask), and ultimately to judge this wear without wearing glasses. Eight attributes all have information that the other person does not know, touches each other to complement. In the past, the workload was evenly divided, and the results were combined. This understanding, the assembly of a car, need eight parts, distributed is equivalent to let eight people each responsible for one, their own collision how to assemble the car better. Non-distributed is equivalent to the amount of work divided into eight people, step by step, the middle of the need for a large amount of data to correct the previous step to convey the accuracy. For example, there is a 1 flaw that you might want to find out from a lot of data corrections, but distributed is the flaw that asks 1 directly in a collision.
Python Object-oriented properties