Software Engineering I Notes 
Sept 9, 2013

Black box
	Don't know what's going on inside code
	Can't see source code or internal data
White/glass box
	Can see code
	Can do more analysis
	Can attempt to ensure certain coverage of statements

	E.g. 90 % of all statements executed
		80% of all branches taken

	White box strategies
		Cover all possible paths
			Infeasible
		Cover all possible edges
			Design a pattern to make sure all outgoing nodes of a flowchart

	Could test all values
		Impossible!

	Equivalence clases
		Divide possible inputs into groups that can be treated the same
		One test per eq class
		
		Month validation

		Vehicle information
			Units
			Maximum/minimum speed

			Fule types
			Fuel efficiency
			time to accelerate

			Range

		Boundary testing


Defects

		Incorrect logical conditions
			Logic condition for if-then-else/loop is incorrectly formed
			e.g. use and instead of or
				< instead of >
				>= instead of >
			Parentheses in wrong place

			See Storn's Chebyshev function

		Perf. calc in wrong part of control construct

			while(j < maximum)
			{
				k = someOp(j)
				j++
			}
			if(k == -1)  s.o.pl("Error in computation")

			Use equivalence class and boundary testing

			Execute loops zero, 1, and more than one times

		Not terminating a loop
			i = 0; j = N;
			while( i < j)
				doSomething()

		Not setting up correct preconditions 
			test all preconditions, esp values beyond what algorithm can expect

		Not handling null conditions
			Find all null conditions possible and run test cases

		Singleton/non singleton
			More than one or only one of something
			Testing strategies may need brainstorming

			What if you are recoding someone's program from C into Java

			What if there was an expectation that there was only one global
			representing what is now a object?

		Off-by-one errors
			for(int i = 1; i < arrayName.length; i++)
				doOperation(arrayName[i]);

			Use boundary testing!!!

		Operator precedence errors
			x +( y * z) vs (x+y)*z

			x*y/z  what if they are integers?

			Is (x*y)/z == x*y/z?

			what about y/z*x?
			

		Using wrong standard algorithms
			inefficient sorting examples?
				insertion sort vs quicksort vs mergesort

			inefficient search algorithm
				binary search vs linear search

			Unstable sorting algorithms

Defects in numerical algorithms

	Not enough precision
		short vs int vs long
		float vs double
		BigNumber

		What if you are building a bridge and it ends up misaligned because of improper precision

		what if you can't support large enough numbers?

	Poor order of operations

		3.78*10^24 + 1

	Floating point comparisons with equality

		is 10.0 == 9.9999999999999?
		is 10.0 = (2.0+2.0+2.0+2.0+2.0) ?

		Are these valid comparisons?

		NO!

		Using two double values, x and y

		compare as follows:  abs(x - y) < 1e-16