cs3101-2 programming languages – c++ lecture 5

CS3101-2, Lecture 5

CS3101-2Programming

Languages – C++Lecture 5

Matthew P. JohnsonColumbia University

Fall 2003

CS3101-2, Lecture 5

Agenda hw3 was due last night Today:

Templates Exceptions Other odds and ends The STL Grading and the final

hw4 TBA tonight

CS3101-2, Lecture 5

Templates Often want to do basically the same thing with

different things functions work on variables

only types specified algorithmic thinking computer science “functionalism” in phil. of mind abstraction human = “the rational animal” (Aristotle)

Sometimes want to do basically the same thing with different types of things Queue of ints, queue of widgets “abstract data types”

CS3101-2, Lecture 5

max functions Suppose want the max of two numbers What kind of numbers?

ints chars floats doubles

All! How?

CS3101-2, Lecture 5

max functions Soln 1: Write one, maxly general function double max(double a, double b) {

return a > b ? a : b;

} double x = max(2.5, 3.5); char c = (char)max(‘A’,’B’); This works but it’s not nice

All four types can widen to doubles but must be cast back

CS3101-2, Lecture 5

max functions Soln 2: Write one function for each type int max(int a, int b) {

return a > b ? a : b;

} double max( … etc. Is allowed in C++ (though not in C) But manually duplicating code

for nontrivial ftns – bad hard to maintain

CS3101-2, Lecture 5

max functions Soln 3: Use the C preprocessor macros #define max(a,b) (a > b ? a : b) C source code is preprocessed

#includes replaced with header files #ifndef, etc. macro calls replaced with macro content

int c = max(2,3); int c = (2 > 3 ? 2 : 3); Works too, but complications, e.g.:

z = max(x++, y++) z = (x++ > y++ ? x++ : y++)

x, y inc-ed twice Need many parens – sq(a+b), etc.

CS3101-2, Lecture 5

max functions Soln 4: Use the CPP in a more sophisticated way Don’t use the CPP to generate expressions but to

generate functions #define define_max(t) \t max(t a, t b) { \

return a > b ? a : b;\}

define_max(char) define_max(int) etc. – no ; Avoids prev. CPP problems But reqs code for all poss types

Done manually

CS3101-2, Lecture 5

Templates template <class T> result ftn(param-list) {…} The place-holder for the substituted type is t template and class are used as keywords can use typename in place of class

T is a type Primitive or class All occurrences in ftn replaced with real type

CS3101-2, Lecture 5

max functions Soln 5: use templates

parameterized function expands per type as necessary

template<typename T>T max(T a, T b) {

return a > b ? a : b;}

Now can simply call the ftn: x = max(2.5,3.5);

Compiler autoly creates only theftn specializations needed

CS3101-2, Lecture 5

Sorting things Consider problem of sorting:

Sorting ints Sorting doubles Sorting strings Sorting widgets

Point of sorting: put list in order Q: What does “in order” mean? A: Given an ordering relation < on the

members For x and y, tells whether x < y Reorder s.t. x is before y iff x < y

CS3101-2, Lecture 5

Generic sorting Sort alg doesn’t depend of element type

Merge sort, quick sort, etc. Need only give means to compare to elms How? In C we pass in a pointer to a compare ftn: void qsort(void *base, int n, int size, int (*cmp)(const void *, void *));

Pass in pointer to ftn: int cmp(const void *a, void *b) {

Widget* w1 = (Widget*)a; …}

Works, but very awkward

CS3101-2, Lecture 5

Generic sorting In Java, we pass a Comparable

implementer In the sort ftn we say if (a.compareTo(b) < 0) …

// means a < b Objects must implement this interface

compare with ftn call Primitives can’t implement

Compared with ops could put in wrappers…

CS3101-2, Lecture 5

Generic sorting C++ soln 1: Define our own Comparable analog:

abstract class Has virtual compareTo Or, better: has virtual < == > operators

Any class extending our class can now be sorted Pass in array of Comparable-extending objects Sort uses polymorphism to treat as (mere) Comparables

Downside: can only sort objects if they extend Comparable

Mult inher: can always add Comp parent,but must do so

To sort primitives must create wrapper classes

CS3101-2, Lecture 5

Generic sorting C++ soln 2: use templates! Let sort take an array of some arb. kind

Don’t need Comparable Don’t need compareTo

In sort, just say if (a < b) …

If these are numbers, this works If these are objects that overload

ops, this works Only requirement:

kind supports < == > ops

CS3101-2, Lecture 5

Templates: swapping Remember our swap-with-ptrs ftn? void swap(int &a, int &b) {

int temp c = a;

a = b;

b = a;

} Suppose we want to swap other types templates

CS3101-2, Lecture 5

Generic swapping template <class T>void swap(T &a,

T &b) {

T temp c = a;

a = b;

b = a;

} Now can swap any prim Can also swap any objects

As long as = op is public

CS3101-2, Lecture 5

Fancier swapping Remember our fancier swap ftn? void swap(int &a, int &b) {

a ^= b ^= a ^= b;}

Fancier template function: template <class T>void swap(T &a,

T &b) {a ^= b ^= a ^= b;

} Now can swap ints, chars, longs But: cannot swap objects

Unless their ^= is overloaded – unlikely

CS3101-2, Lecture 5

Template specialization string s,t … max(s,t); works But max(“hi”,”there”) doesn’t: if (“hi” < “there”) … compares two pointers -

where the char[]s start Not what we mean

Soln: create a specialization special version for this case We check for spec. before template

char *max(char *a, char *b) {

return strcmp(a,b) > 0 ?

a : b;

}

CS3101-2, Lecture 5

Class templates Couple weeks ago: wrote a stack class

supported only integers We’ll abstract element type away

“Abstract data types” Only changes to declar:

1. prepend on class dfn: template <class T>class className {…}2. Replace int T

For ftn implems, we1. prepend the same2. replace className with className<T>3. Replace int T

template <class T>void Stack<T>::push(const T elm){}

To instantiate: Stack<string> strStack;

CS3101-2, Lecture 5

Class specialization Similarly, can specialize member functions

of class templates: void stack<char*>::push(const char

*const item) {

data[count++] = item;

}

CS3101-2, Lecture 5

Templates & statics Review: static data members

one inst shared by all class insts What about statics in templates classes? Q: Could one inst be shared by all insts? A: No – consider: template <class T> class C {

static T mem; …} mem couldn’t me shared by all insts shared by all insts

But: for C<int>, mem shared byall C<int> insts

CS3101-2, Lecture 5

Templates & friends Given class, can declare some outside ftn or class

its friend We have a stack class Suppose: want to declare external sort ftn its

friend Before: had stack with ints

could use sort ftn based on ints Now: have Stack<T>

friend is template too template <class T>class Stack {friend void C<T>::f5(X<T>); …

CS3101-2, Lecture 5

Odds and ends: Forward declarations Suppose classes Cat and Dog each depend on each

other class Cat {

void look(Dog d) { cout << “Meow!\n”; } }; class Dog {

void look(Cat c) { cout << “Bark!\n”; } }; Q: Will this compile? A: No - Dog is referenced before declared Soln: a forward declaration Put class Dog; before Cat def

Dog not yet complete but Dog will now be recognized, w/o Cat depend.

CS3101-2, Lecture 5

Namespaces int i; namespace Example {

double PI = 3.14; int i = 8;void printVals();namespace Inner { int i = 9; }

} // no semi-colon! Can access:

Example::i, Example::Inner::i printVals implementation: void Example::printVals() {

i is Example::i::i is the global I

}

CS3101-2, Lecture 5

Namespaces Now: can use

Example Example::Inner Example::Inner::i Example::i

Nested namespaces ~ Java packages Unfortly: #include (CPP) / using (C++) independent In general, use maximally narrow ranges

Prevent ambiguity Don’t say using namespace std;

Or can fully specify reference: std::std << std::endl;

CS3101-2, Lecture 5

assert Old days: bad thing happens

writing to bad memory address divide by 0, etc. core dump, maybe don’t notice, etc.

void Stack::push(const int item) {data[count++] = item;

} no room overwrite wrong data, crash, etc.

Somewhat better: assert that everything is okay assert(count >= 0 && count < sizeof(data)/sizeof(data[0])); “Everything’s okay, right?”

If false, we quit with message offalse expression

CS3101-2, Lecture 5

Exceptions Now: to some extent

bad behavior is prevented attempt “exception”

If bad things happen we halt, tell calling ftn maybe it halts, tells its calling ftn eventually, either

someone responds accordingly or main ftn passes to OS

try – throw – catch try to do something

maybe an exception gets thrown if so, we may catch it, and go on

CS3101-2, Lecture 5

Exception handling void Stack::push(const int item)

throws BoundExp {

if (count < 0 || count >=

sizeof(data)/sizeof(data[0]))

throw BoundExp(“stack overflow”);

data[count++] = data; //ok if here

} What is BoundExp? A class we define

CS3101-2, Lecture 5

Our exception class BoundExp: exception {

public:

BoundExp(const string &s) exception(s) {}

}; NB: It’s just a class

Its parent is exception but needn’t be Exception has what()

maybe other info

CS3101-2, Lecture 5

Throwing and catching try {

Stack s; …s.push(25); …

}catch (BoundExp &exp) {

cout << “Error: “ << exp.what() << ‘\n’;} catch (ExpType2 &exp) {

// can catch mult kinds// only catch <= 1…

} catch (…) { // … is a wildcard!cout << “Unknown exception

caught.\n”;}

CS3101-2, Lecture 5

Exception classes <exception> exception <stdexcept>

runtime_error, logic_error bad_alloc: new failed bad_cast: dynamic_cast failed

Can throw non-exception objs And even primitives But handling easer if don’t

CS3101-2, Lecture 5

STL Ceteris paribus, libraries are good

Hard, subtle problems many mistakes don’t re-invent the wheel

Unless we’re wheel artists better to commodify the wheel Use an “off-the-shelf” wheel like everyone else

The standard wheel is reliable and efficient STL == Starbucks of programming

Lots of important algorithms,data structures in CS

Barring good reason use std versions

CS3101-2, Lecture 5

Standard Template Library Many template classes, functions Abstract data types Three general categories:

1. Containers2. Iterators3. Algorithms

Three kinds of containers:1. Sequences2. Associative3. Adapted

CS3101-2, Lecture 5

STL: “first-class” containers Sequences: vector: Dynamic-array-backed

const-time random-access const-time insert/delete at back

deque: double-ended queue fast random-access - how? fast insert/delete at front and back

list: doubly-linked list fast insert/delete anywhere

Associative: set: non-sequential, unique multiset: non-sequential, non-unique map: maps from keys to unique values multimap: maps to non-unique values

CS3101-2, Lecture 5

STL: containers Container adapters:

use “first-class” containers by composition stack: LIFO queue: FIFO priority_queue

Near-containers: arrays string bitset valarray

CS3101-2, Lecture 5

Container member ops & ftns copy constructor empty() size() swap First-class: begin() end() rbegin() rend() erase clear()

NB: These are allow very simple - little more than getters

CS3101-2, Lecture 5

STL Iterators Standard way to traverse through

container: iteration Abstraction of both “index” and “pointer”

just: means of iterating forward, back, etc.

Iterator direction types:1. Forward iterator2. Reverse iterator

both supported by vector, list, etc.

3. Random-access iterator supported by vector

Also: its can be const or not

CS3101-2, Lecture 5

Types of iterators I/O iterators are “one-pass”

can only move in one direction can only traverse once – p++

Other types: bidirectional: p++, p-- random-access: p + i, p - i, p[i] *(p+i), p1 < p2

vector: random-access deque: random-access list: bidirectional set/multiset: bidirectional map/multimap: bidirectional

CS3101-2, Lecture 5

vector class Most commonly used container class

Fast random access random-access iterators

Can access mems with []s like arrays – unsafe with at(i) – checks bounds, throws exception – safer

Essentially: dynamic array hidden in obj add to/delete from back: const time unless run out of space autoly copy to larger array insert/del from middle: linear time

must move half of mems forward/back

CS3101-2, Lecture 5

Vectors <vector> Similar to Java’s Vector in that:

dynamic-array-backed list same complexities

Different in that: takes insts of specified type

vector<int> nums; vector<double> vals(20);

size-20 vector of doubles vector<Base> objs;

takes Base objects vector<Base*> ptrs;

takes Base*s or Extended*s

CS3101-2, Lecture 5

Template errors can be illegible Consider this ftn: template <class T>

void printReverse(const vector<T> &vect) {

for (vector<T>::reverse_iterator curr = vect.rbegin();

curr != vect.rend(); curr++)

cout << *curr << ",";

} Slightly different from before

how?

CS3101-2, Lecture 5

Template errors can be illegible

When compiled: Error E2034 c:\Borland\Bcc55\include\rw/iterator.h 442:

Cannot convert 'const int *' to 'int *' in function reverse_iterator<int *>::reverse_iterator(const reverse_iterator<const int *> &)

Error E2094 vect.cpp 19: 'operator!=' not implemented intype 'reverse_iterator<int *>' for arguments of type 'reverse_iterator<const int *>' in function printReverse<int>(const vector<int,allocator<int> > &)

Error E2034 c:\Borland\Bcc55\include\rw/iterator.h 442: Cannot convert 'const int *' to 'int *' in function reverse_iterator<int *>::reverse_iterator(const reverse_iterator<const int *> &)

Warning W8057 c:\Borland\Bcc55\include\rw/iterator.h 442: Parameter 'x' is never used in function reverse_iterator<int *>::reverse_iterator(const reverse_iterator<const int *> &)

*** 3 errors in Compile *** Why? reverse_iterator not const_reverse_iterator

CS3101-2, Lecture 5

Vectors e.g. – vect.cpp template <class T>

ostream& op<<(ostream& out, const vector<T> &vect) {out << "(";for (int i = 0; i < vect.size(); i++)

out << vect[i] << ",";out << ")"; return out; }

template <class T>void printReverse(const vector<T> &vect) {

cout << "(";for (vector<T>::const_reverse_iterator curr = vect.rbegin(); curr != vect.rend(); curr++) cout << *curr << ",";cout << ")";

}

CS3101-2, Lecture 5

Vectors e.g. – vect.cpp void main() {

srand(time(NULL));vector<int> ints;cout << "Initial size == " << ints.size()

<< "\nInitial capacity == " << ints.capacity();for (int i = 0; i < 5; i++)

ints.push_back(rand() % 20);cout << "\nNow, size == " << ints.size() << "\nCapacity == " << ints.capacity();cout << "\nvector: " << ints;cout << "\nvector reversed: ";printReverse(ints);

CS3101-2, Lecture 5

Vectors e.g. – vect.cpp try {

ints.at(100) = 20;} catch (out_of_range oor) {

cout << "\nTried to set mem 100,";

cout << "\nbut caught exception: " << oor.what();

}

sort(ints.begin(), ints.end());cout << "\nAfter sort, vect: “

<< ints;

}

CS3101-2, Lecture 5

Vectors e.g. – vect.cpp Initial size == 0

Initial capacity == 0

Now, size == 5 Capacity == 256

vector: (7,3,16,14,17,)

vector reversed: (17,14,16,3,7,) Tried to set mem 100 to 20, but caught exception: index out of range in function: vector:: at(size_t) index: 100 is greater than max_index: 5

After sort, vector: (3,7,14,16,17,)

CS3101-2, Lecture 5

STL interators – iterio.cpp Access set of values from one place Usually, place is a container But: input stream may be construed as a place #include <iostream> #include <iterator> using namespace std; void main() {cout << “Enter two nums: “;istream_iterator<int> intIn(cin);int x = *intIn;intIn++; x += *intIn;ostream_iterator<int> intOut(cout);cout << “The sum is: “;*intOut = x; cout << endl;

}

CS3101-2, Lecture 5

I/O iterators – ioiter.cpp Code: int x = *intIn;

intIn++; x += *intIn; Output: C:\3101-2\lec5>ioiter

Enter two nums: 5 6The sum is: 11

But if code: int x = *intIn;

/*intIn++;*/ x += *intIn; Then output: C:\3101-2\lec5>ioiter

Enter two nums: 5 6The sum is: 10

CS3101-2, Lecture 5

copy function – vect2.cpp Another way to print container:

use copy function if (!vect.empty()) {

ostream_iterator<T> out(cout, " ");copy(vect.begin(), vect.end(), out);

} copy(src begin it, src end it, dest it);

src begin it: vect.begin() src end it: vect.end()

dest it: ostream_iterator<T> out(cout, " ") it’s an ostream_iterator it’s wrapping around cout it’s outputting Ts it’s printing “ “ between the Ts

CS3101-2, Lecture 5

shuffle, sort, search, min – vect2.cpp

void sort(begin it, end it) it-s must be random-access members must support ==, <

void random_shuffle(begin it, end it) same req’s

bool binary_search(begin, end, target) same req’s also: assumes sorted

min_element(v.begin(), v.end()) returns iterator

CS3101-2, Lecture 5

shuffle, sort, search, min – vect3.cpp

All ftns translate automatically to strings

transform ftn – vect4.cpp transform(begin it, end it, dest it, ftn)

transform(v.begin(), v.end(), v.begin(), square);

cout << "\nAfter squaring, vector: " << v << endl;

CS3101-2, Lecture 5

for_each ftn – vect4.cpp Another way to print container:

use for_each function for_each(begin it, end it, ftn) Our subroutine: template<class T>void print(T val) {

cout << val << "/";}

if (!vect.empty()) { for_each(vect.begin(),

vect.end(), print<T>);}

NB: print<T> is a function pointer

CS3101-2, Lecture 5

Other containers list: doubly linked list

insert/delete anywhere: const time access: linear time bidirectional iterators

deque: double-ended queue insert/delete at front/back: const time insert/delete in middle: linear time access: constant time random-access iterators

CS3101-2, Lecture 5

strings as containers Can traverse strings in the usual way: for (int i = 0; i < i.length(); i++)

cout << s[i]; Also: for (char::iterator curr = s.begin(); curr != s.end(); curr++)

cout << *curr;

CS3101-2, Lecture 5

STL Algorithms - <algorithm> binary_search sort count: count(list.begin(),

list:end(), val, num); equal: compares containers for_each: applies ftn to each element copy: copies container reverse min/max Some in <numeric>

CS3101-2, Lecture 5

Other algorithms Set-theoretic:

set_union set_intersection set_difference set_symmetric_difference

Sorting: sort_heap stable_sort

And many more…

CS3101-2, Lecture 5

Algorithms in STL Important observation: STL class live in own headers

- <vector>, <list>, but STL algs live in places like <algorithm> and <numeric>

To sort, we pass access (it) to our obj to the sort ftn We don’t call obj.sort() Why? The STL doesn’t use inheritance! Why not? virtual functions are slow(er) No inher would have to dup. ftns No inher no encapsulation algorithms on their own

CS3101-2, Lecture 5

STL e.g.: grades Goal: store grades for group of students

ordered set of assignments maybe students with same name

For each student, have grades want fast access to each grade use vector of chars typedef vector<int> Grades;

First, create map of students: map<string, Grades> roster;

CS3101-2, Lecture 5

STL: add a student to map map represents a function: key maps to value map, basically: set of ordered pairs

pair<x,y> template void Roster::addStudent(const string &name) {

//check if already existsif (roster.find(name) != roster.end())

return;//check for roomif (roster.size() == MAX) waitList.push_back(name);else { Grades grades; roster.insert( pair<string,Grades>(name,grades));

}

CS3101-2, Lecture 5

STL: add a student to map Notice find line: if (rost.find(name) != rost.end())

return; find function searches for an elm with our key if found, returns pointer to it if not, returns pointer to end()

points past, not to last member like: for (int i = 0; I < n; i++)

More precisely: these ptrs areiterators – inc/dec to step through seq

More precisely: these ptrs are iterators ops overloaded as though moving thru mem

CS3101-2, Lecture 5

STL: add a student to map Notice insert line: roster.insert( pair<string,Grades>(name,grades));

Dissection: Add a member to a roster The member is a pair of two things

member ftns: first(), second() The things are string and Grades pair<X,Y>(x,y) is a constr call:

passes x and y to constr for type pair<X,Y>

CS3101-2, Lecture 5

STL: drop a student void Roster::dropStudent(String &name) {

if (roster.find(name) == roster.end())return;

roster.erase(name);if (waitList.size() > 0) {

string wait = waitList.pop_front();

waitList.pop();Grades grades;roster.insert(

pair<string,Grades>(name,grades));}

}

CS3101-2, Lecture 5

STL: set a grade void Roster::setGrade(const &string name,

const int assign, const char grade) {

map<string,Grades> stud = roster.find(name);

if (stud == roster.end()) {cerr << “not found\n”;return;

}if (stud->second().size() <= assign) stud->second().resize(assign+1);

stud->second[assign] = grade;}

CS3101-2, Lecture 5

STL: print grades void Roster::print() { … } Already saw many print/<< functions…

That’s all we’ll cover of STL Many more classes, algs in STL Much more to C++ itself But: you now know enough

About the most import. Features To learn remaining details on own

CS3101-2, Lecture 5

Next time: final exam Closed books/notes/everything 2 hours/class time Questions:

“Vocab”: protected, static, etc. Find errors/read code/predict output Write code

See web for a semi-definitive list of topics Jake will be proctoring the exams But we’ll have OH next week(end)

Come in if you have questions!

CS3101-2, Lecture 5

Grading Final will be hard/challenging

Ave score probably near 70% Final grades are curved

Rule of thumb: mean ~ stdev Mean ~ B/B- +-stdev ~ one letter grade See mean/stdevs on web to estimate your

grade

CS3101-2, Lecture 5

The future Q: What happens when backward-compat is

removed from C++? A: C#

less complexity Microsoft approves

Better or worse than C++, Java? Find out in CS3101-4: C#, next spring

Tonight: hw4 TBA – due by final Please fill out course evals!

Link will be on classpage tonight Available until beginning of finals Get valuable extra credit on final

Sign in and good luck! Happy Thanksgiving!