. 07 Designing Classes
. Previous --introducing objects and classes
.See ./06.html
.Open Prepare
. 3.10 Validating Data with set Functions
A class must always protect its objects from abuse.  A class defines
a collection of functions that can change the data in an object.  You
must be sure that they leave the object's data in a valid state.  In
"setter"s this means you must check the values of the parameters to see
if they are valid before you do anything else.
. 3.11 Software Engineering Case Study: Identifying the Classes in the ATM Requirements Specification 
When we think about how a program will be used, in terms that a user
understands, we discover the names of many of the classes inside our software.
Similar we can discover some of the methods (member functions) for these
classes in scenarios describing what the user wants to do.
.Close Prepare
.Open Answered Questions
. Is there a way to limit decimal places without using substrings
The double length computations are always done to maximum accuracy.

You can remove the decimal part of a double by assigning it to an integer.
.As_is 		int rounded_down = unrounded;

Thestandard technique for propper rounding to the nearest is to add 0.5 first
.As_is 		int rounded = unrounded + 0.5;

To round to, say two paces, you shift the decimal point before you round
and then shift it back again:
.As_is 		int rounded = (unrounded * 100.0)+0.5;
.As_is 		double result = rounded / 100.0;
For a running example, see
.See ./rounding.cpp
in this web site.

Their are several more advanced techniques in C++.  The `cmath` library
has two or three helpfule function for rounding.  I hope to cover these later.

There are also some advanced nput/output `manipulator` functions for 
accurately formatting output to a specific number of places.  This part of CS202
but here is a sample for formatting doolars and cents:
.As_is 		#include <iomanip>
.As_is 		. . .
.As_is 		//Suitable for a check protected monetary field
.As_is 		double dollars = 1234.45;
.As_is 		cout << '$';
.As_is 		cout << setfill('*'); cout <<setw(10); cout << setprecision(2);
.As_is 		cout << setiosflags(ios::fixed | ios:: showpoint);
.As_is 		cout << dollars <<endl;
.See ../examples/giomanip.cc
(old C++ code).

. What does separating interfaces do to the program
It splits the program into pieces of code that can be worked on 
independently. If you are doing a team project (the norm in practice) you
can give the separated parts to different people.  The interface lays down
the rules for their parts of the program to work together.  They can then 
implement their parts in any way that works.

Library classes are 90% hidden behind an interface.  For example here
is a simplified list of the "string interface" taken from our library:
.See ./string
(Plain text file).  You can use these functions without worrying about
what goes on inside the library.

. How does separating interface from implementation make my life easier
If you work alone then it means that you don't have to edit and recompile
massive files all the time.  You work on one file and if needed recompile it,
and then switch to another part of the problem and work on that.  

It is Ceasar's method: "Divide, and conquer". 
. In a .h file why are function definitions replaced by prototypes
So that you don't know the details of the function implementation.

A prototype tells you how to call a function but hides (and delays) the
details of what happens.  

A header file (like GradeBook.h) is shared, published information.  By only
putting prototypes in a header file you expose the right info for a client to
use it without swamping the client with details.
. Do I always need a void displayMessage()
No.  Only when you need an object to `display a message`.  Indeed many would
argue that instead of `display....` you should write `string getMessage`
and use `cout` to do the output.  They like to separate the user interface from
the logic of the class.  Then the class will not change if you get
new input/output devices.
. Why do we always need an object to be in a consistent state
Other methods in the object rely on the state being consistent.

Any object which has member functions that leave the state in a strange state
can not be trusted. 
. Which functions are void functions
The `void` functions may change the data in the object, but
the do not return any inormation about the object's state/data.

They are the "setters" or "set functions".

Non-void member fucntions return data and so are report on the state
of the object.  They are also called "getters" or "get functions".

Get functions should not change the state since this can lead to surprises.
Good design always follows the Principle of
.Key Least Surprise.
. Why is there null character placed at the end of string literals
The compiler doesn't store the quotation marks.  It stores the characters
with an extr `null` at the end and remembers where the string starts.  Programs
use the `null` to spot the end of the string.

Strings in the `string` library are more complex internally, but they hide
their techniques behind an interface.
. Should you use a set function to check some data
A `setter` changes the data inside an object so this sounds like overkill.
If you expect to need data validation without storing the result then you should
code member functions that test the data:
.As_is 		bool validData(....){....}
for instance.  A `bool` is a value that is either true or false and can be used
in conditions.
. Why do class names always start with a capital
Mathematicians always represent sets using capitals.... and a class defines a
set of objects.

It is not part of the C++ language but it is part of the $OO culture.
. Are if statements the only way to validate data
No.  There is a function `assert` in the <cassert> library that makes it
very simple to reject bad data.  Here is a sample of rejecting an `index` that
is negative or too big:
.As_is 	#include <cassert>
.As_is 	. . .
.As_is 	   assert (index >= 0 ); assert (index <= SIZE-1);
In general `assert(condition)` does nothing if the condition is true, and
aborts the whole program if it is false.... generating diagnostics.

Another, modern tool, for rejecting bad data is to use 'if' plus `raising an exception`.
This is advance stuff....

. If you know string data will be longer than 25 characters why not make it bigger
The question is:  where do you stop?  255 characters? 1023 characters? As the limit
grows the program needs more and more space for its data.

The logic is the same with just the number changing:
.As_is 		if( p.size() > 25 )....
.As_is 		if( p.size() > 255 )....

This example in the book goes back to the bad old days when you couldn't store
a string without defining its maximum size.  Running over this size leads te
program into very bad behaviors including crashing the operating system
and taking control of the computer!  If you use the sandard string you don't
have to worry about a maximum size: it adjusts.

However, the point that you must always vaildate the data you have been given
is still true.
. Why validate data
Fewer bugs.  In fact -- an easier life for the programmer.
. Are there more UML diagrams
Yes - thirteen!  In CS201 I'm only going to require you to draw class diagrams.

. Are there other kinds of class than the elided and normal variety
Yes.  "Templates" and "Stereotyped classes".  We won't do much with these...
. How to compile and link programs on UNIX
This command
.As_is 		g++ -c name.cpp
will compile a C++ program called `name.cpp` into an object code.  

You do this for each ".cpp" file you have...

You can link them into an executable program like this
.See 		g++ -o executable *.o
The "*" is a wild card.  "*.o" is replaced by a list of all your object files.
. Why are role names important in UML diagrams
A well chosen role name can make an association much clearer.  It also
hints at data members in the class.... more later.

.Close
. Exercises -- classes and how objects work
.See ./07ex.cpp
. Next -- control structures
.See ./08.html
. Lab
.See ./lab04/