Re: [Tutor] Mixing in and Mixing out classes in python
Tomaz Bevec [EMAIL PROTECTED] wrote I'm working on a simulation of cellular growth patterns ...and cells can potentially gain and lose these behaviors over the course of the simulation. OK, That might be a valid scenario. But personally I'd probably implement that as an attribute of the cell - a list of behaviours and build a Behaviour class with subclasses. The methods of which take a Cell class as their argument. It is then trivial to add remove behaviours. And simply call them in a loop for b in self.behavious: b.doit(self) Or maybe have a dictionary of behaviours if you even want to call specific ones. eg def myMeth(self): # do sometjing try: self.behaviours['behaviourMyMeth'](self) except KeyError: pass # do the rest I have however thought of a number of ways to work around this, and per your suggestion I think I'll just try something else, I'd come back to the old OOP adage, inheritance is for Is-A relationships. Is you Cell a behaviour or does it Have-A behaviour? With mixins that adage tends to get bent since mixins are often functional in nature but if the function is not intrinsic to the class I tend to use delegation. any idea on how to dynamically mix out I'd be glad to here it. I don't know but I'm sure its possible if you understand the class internals well enough. Regards, Alan G. ___ Tutor maillist - Tutor@python.org http://mail.python.org/mailman/listinfo/tutor
Re: [Tutor] Mixing in and Mixing out classes in python
Tomaz Bevec [EMAIL PROTECTED] wrote I am using the following function to mixin in classes into specific object instances ... Is there an analogous way to Mixout classes from an instance at runtime? I'm sure it is possible but... Are you just asking out of interest? Or do you have a real world need for this? Using multiple inheritence is a non trivial aspect of OOP (albeit powerful) that is fraught with difficulty and potential side effects. Adding a mixin at run time is difficult enough, removing one would bend my brain way too far I suspect. (Many coding standards insist on limiting MI to two classes or even banning it outright as being too bug prone.) Dynamic mixins add a whole new complexity (it's somewhat akin to the FAQ here about dynamically naming variables) all you generic code has to take account of the new method introduced and any potential side-effects that may not be handled in your code. Unless your mixin classes are ultra clean and stateless in their implementation you run a risk of breaking things in ways that are almost impossible to debug. I've never come across a need for dynamic mixin additions and I'm interested in how you think you might use this. (CLOS is the only other place I've even seen this discussed and it's possible there with some under the covers work but I've never seen it used!) Curious, -- Alan Gauld Author of the Learn to Program web site http://www.freenetpages.co.uk/hp/alan.gauld ___ Tutor maillist - Tutor@python.org http://mail.python.org/mailman/listinfo/tutor
Re: [Tutor] Mixing in and Mixing out classes in python
Thanks for your reply Alan, I am partially asking out of interest, but I also have a potential application. I'm working on a simulation of cellular growth patterns (basically cell instances interacting stochastically on a lattice). Anyway, there are many different cell behaviors that I have to simulate, and cells can potentially gain and lose these behaviors over the course of the simulation. It would be too much to put every behavior function in the cell class, so I'm writing each behavior as a mixin and mixing it in (and initializing it) when necessary or when biological function is gained by the cells. In addition some of the behaviors would be too big (as in lines of code) if they were implemented in one function so splitting functionality up in my mixin class makes it conceptually easier for me. Also the behaviors sometimes need access to the internals of the cell class. After I mixin a class every method that starts with the word 'behavior.*' in that class is called when I update the simulation. All of my mixins inherit from object only. So far my approach is working well, but I don't have a clean way for cells to lose functionality. Hence my question about mixin out. I tried to just delete the 'behavior.*' functions with the del operator (like for an attribute) but it didn't work. I have however thought of a number of ways to work around this, and per your suggestion I think I'll just try something else, but if you have any idea on how to dynamically mix out I'd be glad to here it. Thanks, Tomaz --- On Thu, 7/31/08, Alan Gauld [EMAIL PROTECTED] wrote: From: Alan Gauld [EMAIL PROTECTED] Subject: Re: [Tutor] Mixing in and Mixing out classes in python To: tutor@python.org Date: Thursday, July 31, 2008, 3:16 PM Tomaz Bevec [EMAIL PROTECTED] wrote I am using the following function to mixin in classes into specific object instances ... Is there an analogous way to Mixout classes from an instance at runtime? I'm sure it is possible but... Are you just asking out of interest? Or do you have a real world need for this? Using multiple inheritence is a non trivial aspect of OOP (albeit powerful) that is fraught with difficulty and potential side effects. Adding a mixin at run time is difficult enough, removing one would bend my brain way too far I suspect. (Many coding standards insist on limiting MI to two classes or even banning it outright as being too bug prone.) Dynamic mixins add a whole new complexity (it's somewhat akin to the FAQ here about dynamically naming variables) all you generic code has to take account of the new method introduced and any potential side-effects that may not be handled in your code. Unless your mixin classes are ultra clean and stateless in their implementation you run a risk of breaking things in ways that are almost impossible to debug. I've never come across a need for dynamic mixin additions and I'm interested in how you think you might use this. (CLOS is the only other place I've even seen this discussed and it's possible there with some under the covers work but I've never seen it used!) Curious, -- Alan Gauld Author of the Learn to Program web site http://www.freenetpages.co.uk/hp/alan.gauld ___ Tutor maillist - Tutor@python.org http://mail.python.org/mailman/listinfo/tutor ___ Tutor maillist - Tutor@python.org http://mail.python.org/mailman/listinfo/tutor
Re: [Tutor] Mixing in and Mixing out classes in python
On Thu, Jul 31, 2008 at 8:09 PM, Tomaz Bevec [EMAIL PROTECTED] wrote: Thanks for your reply Alan, I am partially asking out of interest, but I also have a potential application. I'm working on a simulation of cellular growth patterns (basically cell instances interacting stochastically on a lattice). Anyway, there are many different cell behaviors that I have to simulate, and cells can potentially gain and lose these behaviors over the course of the simulation. It would be too much to put every behavior function in the cell class, so I'm writing each behavior as a mixin and mixing it in (and initializing it) when necessary or when biological function is gained by the cells. Perhaps you could keep the behaviors in a dictionary? You could override __getattr__() in the cell class to look up attributes whose names start with 'behavior' in the dict. Kent ___ Tutor maillist - Tutor@python.org http://mail.python.org/mailman/listinfo/tutor
