CS100J Lecture 22

1. CS100J Lecture 22 • Previous Lecture • MatLab • Learn MatLab • Implement (a piece of MatLab) in Java • Get an appreciation for what MatLab does for you • Learn to implement a class of some complexity • Handout: The CS100 MatLab Syllabus • Java Concepts • Overloading • this as a constructor • The Next Several Lectures • MatLab and its implementation, continued. Lecture 22

2. Sample Client Code // ones(2,3) + ones(2,3) ML.add( ML.ones(2,3), ML.ones(2,3) ) // 5 + ones(2,3) ML.add( new ML(new ml(5)), ML.ones(2,3) ) • The second example suggest that it might be useful to add an additional overloaded constructor for ML: // Construct scalar matrix v. public ML(int v) { this(1,1); values[0][0] = new ml(v); } • so the example could be: // 5 + ones(2,3) ML.add( new ML(5), ML.ones(2,3) ) • etc. Lecture 22

3. Critique of ML Method add • Most of the code (all but the 3 invocations of ml.add) has nothing to do with the fact that this is the method for addition. • To implement subtraction, almost all code (except for the 3 invocations of ml.add) would be identical. • We could implement each of the other element-by-element matrix operations in about 30 seconds each: • Copy the definition of ML.add • Replace “add” with “sub”, or “mult”, or “div”, etc. • Write ml.add, ml.mult, ml.div, etc. • Terrible: • Grotesque duplication of code. • Loss of “single point of change”. • No abstraction. • Surely, there must be a way to avoid duplicating the common code. • Coming later. Lecture 22

4. MatLab • Matrices with the same number of rows can be concatenated horizontally (comma optional). >> [ 1 2 3 ] ans = 1 2 3 >> [ ones(2,3) , (2 .* ones(2,2))] ans = 1 1 1 2 2 1 1 1 2 2 • Matrices with the same number of columns can be concatenated vertically. >> [ 1 ; 2 ; 3] ans = 1 2 3 >> [ ones(2,3) ; (2 .* ones(1,3))] ans = 1 1 1 1 1 1 2 2 2 Lecture 22

5. MatLab • A constant matrix can be written with a combination of horizontal and vertical matrix concatenations. >> [ 1, 2, 3; 4, 5, 6; 7, 8, 9] ans = 1 2 3 4 5 6 7 8 9 Lecture 22

6. Implementation Strategy • ML.row(x,y) implements binary row concatenation, i.e., [x, y] • ML.col(x,y) implements binary column concatenation, i.e., [x ; y] • Defer concatenation of more than 2 rows or columns at a time. Until then: /* [a, b, . . ., y, z] */ ML.row(a, ML.row(b, ... ML.row(y, z)) ... ) /* [a; b; . . .; y; z] */ ML.col(a, ML.col(b, ... ML.col(y, z)) ... ) • (Actually ML.row and ML.col are associative, so the grouping is arbitrary.) Lecture 22

7. class ML: Method row // [x, y] static ML row(ML x, ML y) { if (x.h != y.h) thrownew RuntimeException ("operands must have same height"); else { ML result = new ML(x.h, x.w+y.w); for (int r = 0; r < x.h; r++){ for (int c = 0; c < x.w; c++) result.values[r][c] = x.values[r][c]; for (int c = 0; c < y.w; c++) result.values[r][x.w+c] = y.values[r][c]; } return result; } } Lecture 22

8. class ML: Method col // [x; y] static ML col(ML x, ML y) { if (x.w != y.w) thrownew RuntimeException ("operands must have same width"); else { ML result = new ML(x.h+y.h, x.w); for (int c = 0; c < x.w; c++){ for (int r = 0; r < x.h; r++) result.values[r][c] = x.values[r][c]; for (int r = 0; r < y.h; r++) result.values[x.h+r][c] = y.values[r][c]; } return result; } } Lecture 22