Where we left off

const int SIZE = 64;
char sentence[SIZE];

cout << "Enter a sentence: ";
cin.getline(sentence, SIZE);

int i = 0;
int words = 0;
while (sentence[i] != '\0') {
    if (sentence[i] == ' ')
        words++;
}

Today's topics

• Grouping data with structures
• Functions + structures
• Arrays + structures
• Assignment 2

Textbook Chapter 10

But first, some getline confusion

• C++ has a function in the <string> header called getline:

  getline(istream& is, string& str);
  

• Do not use this - the string class handles all the low-level memory stuff that I want you to learn about
• Instead, use the getline member function of the input stream:

  char str[SIZE];
  cin.getline(str, SIZE);
  

More clarification: the const modifier

• So far we've used const in two places:
  • Defining a named constant, e.g. const double GST = 0.05
• Defining a function parameter as const, e.g.:

  int calc_average(const int values[], int n_vals);
  

• Arguments assigned to a const parameter do not need to be const
• This simply says that the function cannot modify the array
• Similarly, a const int can be assigned to the n_vals parameter, as the value is copied into the function parameter

Separate Compilation

• Typical lab structure:
  • lab.h
  • lab.cpp - #include "lab.h"
  • main.cpp - #include "lab.h"
• Prevents duplication of the code in lab.h, keeps main logic clear
• We could compile in multiple steps:
  • g++ -c lab.cpp - compiles lab.cpp into lab.o
  • g++ -c main.cpp - compiles main.cpp into main.o
  • g++ -o main main.o lab.o - links the two object files

What happens when you run make?

Compiling in multiple steps is annoying, so we dump it in a makefile

# This is "Makefile". Notice that comments begin with "#"
program: lab.o main.o
    g++ main.o lab.o –o program
main.o: main.cpp
    g++ -c main.cpp
lab.o: lab.cpp
    g++ -c lab.cpp

• Important: the indentation is a tab, not spaces! (emacs knows this)

Protecting against multiple #includes

• Most projects have many different modules (a somewhat random example)
• For example, in assignment 2 (not yet released):
  • main.cpp includes applicant.h and score.h
  • score.h includes applicant.h
• Problem: #include means "copy and paste" so we're defining stuff twice!
• Solution: header guards
  • Wrap your header file in #ifndef and #endif directives

Header guards

#ifndef APPLICANT_H
#define APPLICANT_H

... // contents of applicant.h

#endif // APPLICANT_H

• #ifndef checks if the macro APPLICANT_H is defined
• If it is, the preprocessor skips to the #endif
• By convention, the macro name is the header file name in all caps
• Also conventional to put a comment after the #endif

Separate Compilation check-in 1/2

Which of the following are good reasons to use separate compilation? Select all that apply.

Separate Compilation check-in 2/2

The #include directive is a preprocessor directive that means:

Moving on to structures

Functions with long lists of parameters are painful:

// Calculates the amount owed by the customer based on usage and account limits
void calculate_bill(double base_charge, double usage_limit, double maxMB_used,
                    double endMB, double& over_charge, double& penalty_charge,
                    double& gst_owed, double& total);

// Displays the final bill. If no surcharges are owing, these are not shown.
void print_bill(int account_number, double usage_limit, double beginMB,
                double maxMB_used, double endMB, double base_charge,
                double over_charge, double penalty_charge, double gst_owed,
                double total);

• Wouldn't it be nice if we could bundle all that stuff into a single variable?

Structure syntax

• General form:

  struct <type name> {
      <field1 declaration>;
      <field2 declaration>;
      ...
      <fieldn declaration>;
  };
  

• This says "define a new type named <type name> with the given fields

A structure for bill calculations

struct BillInfo {
    int account_num;
    double base_charge;
    double usage_limit;
    double maxMB_used;
    ...
    bool valid;
}

• This defines a new type called BillInfo with the given fields
• This does not declare a variable of type BillInfo!

Using your new type

• Once you've defined a new type, you can use it to declare variables:

  BillInfo user_bill; // a BillInfo instance
  BillInfo another_bill; // another BillInfo instance
  

• This is now allocating memory for all the fields in the structure
• Common practice:
  • define structures globally so all functions are aware of the new type
  • name structures using UpperCamelCase (PascalCase)

Accessing structure fields

• Like class objects, structure fields can be accessed with dot syntax:
  • user_bill.account_num = 12345;
  • user_bill.base_charge = 10.00;
  • cout << "Account: " << user_bill.account_num << endl;
• Once you've accessed via ., fields behave just like normal variables
• The fields of a given instance are not related to another instance
• Memory for each field is allocated in order

Initializing structure fields

• You can initialize structure fields at declaration time:

  BillInfo user_bill = {12345, 10.00, 1000, 100, true};
  

• But this requires remembering the order of fields and can be error prone
• Like arrays, missing values are initialized to a 0 value of their data type

  BillInfo user_bill = {};
  

Operations on structures

• You can pass structures to functions:

  void print_bill(BillInfo bill);
  

• You can return structures from functions:

  BillInfo read_and_process();
  

• You can even copy structures:

  BillInfo bill1 = {12345, 10.00, 1000, 100, true};
  BillInfo bill2 = bill1;
  bill2.valid = false; // What happens to bill1.valid?
  

• But you can't compare them with == or !=

Structures and functions

• Unlike arrays, structures are passed by value by default
• You can (and usually should) pass structures by reference:

  void read_bill(BillInfo& bill);
  

• What happens in memory with the following function prototype?

  void print_bill(BillInfo bill);
  

• Instead of passing by value, good idea to pass by const reference

Returning structures from functions

• Unlike arrays, structures can be declared in a function and returned
• The structure is copied into the caller's memory:

    BillInfo read_and_process() {
        BillInfo bill;
        // read data into bill
        return bill;
    }
  
    // in main
    BillInfo user_bill = read_and_process();
  

• For small structures this is fine, but for large structures this passing a reference is more efficient (visualization)

Structures vs arrays 1/2

Which of the following is false?

Structures vs arrays 2/2

What can you infer from the function prototypes shown?

void a(int arr[]);
void b(BillInfo bill);

Structures with array fields

• Structures can contain arrays (including C-strings) as fields

  struct Student {
      char name[64];
      int number;
      double gpa;
  };
  

• Oddly, this now allows for whole array operations like copying!

  Student a = {"Bob", 12345, 3.5};
  Student b = a;
  strcpy(b.name, "Alice");
  

Arrays of structures

• You can also declare arrays of structures:

  BillInfo bills[10];
  
  for (int i = 0; i < 10; i++) {
      read_bill(bills[i]);
  }
  

• This allocates memory for all fields of all instances in the array
• Standard array rules apply for passing to/returning from functions:

  void read_and_process(BillInfo bills[]); // passed by reference
  void print_bills(const BillInfo bills[]); // mark as read-only
  

Arrays of structures continued

• To access a field of a structure in an array:
  • First, use indexing to access the element in the array
  • Then, use dot notation to access the field of the element
• For example, to set the ith bill's account number:

  bills[i].account_num = 12345;
  

• You can have arrays of structures that have arrays as fields...
• Or even arrays of structures that have arrays of structures as fields...

Coming up next

• Lab: Structures
• Lecture: File I/O and command line arguments
• Assignment 2 now available