I moved the Sage simulation of the can crusher to an object oriented implementation today. A few days ago,I was worried this might have been a bit of overkill and just a subconcious desire on my part to place the project in a code format I'm used to seeing things in. I hit an example yesterday that convinced me otherwiser, and only a few short, OK, and somewhat grueling, hours later, I had a much easier to use OO simulator.
Prior to yesterday, my usage mode of the can crusher code was as follows:
1. Evaluate the cell that contained the initializaiton code. There were some declarations of global varaibles and a little bit of code that atually ran on evaluation to place values in these variables.
2. Evaluate the cells that contained the current calculating function and the can moving function separately.
3. Evaluate the cell that contained the simulation code.
This, as far as I knew had to be done every time I wanted to change any values and run a simulation. I was constantly worried that I might not evaluate the cells in the correct order or that a rogue global variable might escape initialization.
The clincher came when I wondered if there was a difference between the simulated current through the driving coil when the can was allowed to crush and when it was not. In order to do this, I had to execute the above procedure, then, copy the results of the global current array into a placeholder array, then modify the simulation code, then run the above procedure again, and finally graph the new global current array and the placeholder array on the same plot to get the following result
Contemplating the stress of wondering what I had called my placeholder variables and whether or not I'd run through the initialization procedure correctly every time led to this morning's object oriented re-write. What I wound up with was a simulator object that encapsulates all the initialization procedures and its own set of member variables that correspond to the previous set of globally available variables Now, to create a comparison like the one shown above, I can just create two simulations, passing both of them the same limit for the number of simulation steps to run, but tell one of them to turn the can moving portion of the code off, like so, (my simulator class is named Crusher):
nm_crushtest = Crusher()
crushtest = Crusher()
After the simulations have run, the graph is easy to generate:
nomove = list_plot(nm_crushtest.coilOutTime[0:372, 0:2], color='red')
move = list_plot(crushtest.coilOutTime[0:372, 0:2], color='blue')
show(nomove + move)
The Grueling Bit
If you're coming from a C++ background, the only part that will drive you nuts is having to put 'self.' in front of every member variable every time it is used in the class definition. I still think I must have been doing this wrong and hope that I kind Python expert will correct my gruesomely bad coding style soon.
1. The Python object oriented documentation
2. Upcoming useful docs: How to handle large +Sage Mathematical Software System programs
3. The can crusher simulator in it's many revision controlled incarnations