Assorted C/C++ tips and tricks.
If compairing to a constant, the constant should be the lvalue (on the left).
This erroneous code will compile without errors (good compilers will warn you):
while (foo = NULL)
{ ... }
The pointer foo will be set to NULL and the loop will never be entered. On the other hand, trying to build the following will always generate an error:
while (NULL = foo) { ... } test.c: In function `main': test.c:7: error: invalid lvalue in assignment
Whenever you need to define a set of mutually exclusive
states, always use an enumeration. Don't use an int and a
bunch of #defines. If space is really a premium and you only
need a few things, you could settle for a char and a
few letters as a code. If you need to represent some states that
aren't mutually exclusive, use bitflags (see the next section).
An enumeration is basically an integer type associated with a bunch
of special tokens. These tokens can be used for assignment and
is-equal or not-equal checks - you can think of them as a sort of
special #define.
//declare the enumeration _type_
enum BirdState {FLYING, LANDING, STANDING, TAKEOFF, EATING, SLEEPING};
//create a variable of it (in c or c++)
enum BirdState bs1 = SLEEPING;
//create a variable of it (in c++ only)
BirdState bs2 = SLEEPING;
//compare state
if (bs1 == EATING)
bird1.hungry--;
//set and compare state
if (bs1 == TAKEOFF && bird1.velocity > 0.3)
bs1 = FLYING;
There are some differences between enums in C and C++.
First, in C++ you do not need the enum keyword when
declaring a variable. Second, in C++ you cannot use general integer
operators (such as arithmetic and bitwise operators) on
enumerations
enum BirdState bs2; // legal in C and C++
BirdState bs3; // legal in C++ only
typedef enum BirdState BirdState_t
BirdState_t bs4; // can use typedef like this
// to make it legal in C as well
bs2++; // Illegal in C++
bs3 = 2; // Illegal in C++
bs4 = bs2 | bs3; // Illegal in C++
Bitflags are a method of storing multiple values, which are not
mutucally exclusive, in one variable. You've probably seen them
before. Each flag is a bit position which can be set on or off. You
then have a bunch of bitmasks #defined for each bit
position so you can easily manipulate it:
#define LOG_ERRORS 1 // 2^0, bit 0
#define LOG_WARNINGS 2 // 2^1, bit 1
#define LOG_NOTICES 4 // 2^2, bit 2
#define LOG_INCOMING 8 // 2^3, bit 3
#define LOG_OUTGOING 16 // 2^4, bit 4
#define LOG_LOOPBACK 32 // and so on...
// Only 6 flags/bits used, so a char is fine
unsigned char flags;
// initialising the flags
// note that assignming a value will clobber any other flags, so you
// should generally only use the = operator when initialising vars.
flags = LOG_ERRORS;
// sets to 1 i.e. bit 0
//initialising to multiple values with OR (|)
flags = LOG_ERRORS | LOG_WARNINGS | LOG_INCOMING;
// sets to 1 + 2 + 8 i.e. bits 0, 1 and 3
// setting one flag on, leaving the rest untouched
// OR bitmask with the current value
flags |= LOG_INCOMING;
// testing for a flag
// AND with the bitmask before testing with ==
if ((flags & LOG_WARNINGS) == LOG_WARNINGS)
...
// testing for multiple flags
// as above, OR the bitmasks
if ((flags & (LOG_INCOMING | LOG_OUTGOING))
== (LOG_INCOMING | LOG_OUTGOING))
...
// removing a flag, leaving the rest untouched
// AND with the inverse (NOT) of the bitmask
flags &= ~LOG_OUTGOING;
// toggling a flag, leaving the rest untouched
flags ^= LOG_LOOPBACK;
WARNING: DO NOT use the equality operator (i.e. bitflags == bitmask) for testing if a flag is set - that expression will only be true if that flag is set and all others are unset. To test for a single flag you need to use & and == :
if (flags == LOG_WARNINGS) //DON'T DO THIS
...
if ((flags & LOG_WARNINGS) == LOG_WARNINGS) // The right way
...
if ((flags & (LOG_INCOMING | LOG_OUTGOING)) // Test for multiple flags set
== (LOG_INCOMING | LOG_OUTGOING))
...
Stolen from the FreeBSD errno manpage:
extern int * __error();
#define errno (* __error())
(Background: This allows __error() to return a errno for a specific thread instead of using a single global errno that could be wiped out by another thread)
(non-portable)
First of all, make sure you're checking terminal types or at least providing a monochrome option so people who use terms that don't support this don't suffer excape characters all over their text.
The codes themselves have a starting sequence, (Escape-[ or \033[) then a set of numbers seperated by semicolons which may contain a text attribute normal, bold, underlined, blinking...) and foreground and background colours. The code is closed with an m.
Properties Properties Foreground Background
(disable)
0 normal 30 black 40 black
1 bold 22 unbold 31 red 41 red
4 underline 24 ununderline 32 green 42 green
5 blink 25 unblink 33 yellow 43 yellow
7 inverse 27 uninverse 34 blue 44 blue
9 strike 29 unstrike 35 magenta 45 magent
36 cyan 46 cyan
37 white 47 white
39 default 49 default
Attributes are not as widely supported as simple colours. Often in an X environment the user has chosen a bold font and bold will have no effect; underlined, blinking and inverse text sometimes fail for no particular reason. The codes for cancelling an attribute (unbold etc) sometimes do something completely different. Strikethrough is "standard" but I've never seen it actually work.
For these reasons you're probably better off just sticking with the colours. A C example follows:
#define ANSI_NORMAL "\033[0m" /* these are all normal text with black background (40) */ #define ANSI_RED "\033[0;31;40m" #define ANSI_GREEN "\033[0;32;40m" #define ANSI_YELLOW "\033[0;33;40m" #define ANSI_BLUE "\033[0;34;40m" #define ANSI_MAGENTA "\033[0;35;40m" #define ANSI_CYAN "\033[0;36;40m" #define ANSI_WHITE "\033[0;37;40m" #define ANSI_BOLD_RED "\033[1;31;40m" #define ANSI_UNDERLINE_RED "\033[4;31;40m" #define ANSI_BLINK_RED "\033[5;31;40m" ... printf("%sI am blue!%s\n", ANSI_BLUE, ANSI_NORMAL);
Always, always close the text with the \033[0m code (normal text) or else the text after your program ends will be affected...
(advanced, non-portable)
Bit masks are where you specify the number of bits to use in a integral member of a class or struct. C has more restrictions than C++ on this. In particular, only "int", "signed int" or "unsigned int" can have bit masks in C. Many C compilers ignore bit masks altogether.
First rule of bit masks: Don't. Don't unless you have a really bad need to save a tiny amount of memory, and slowing down your code a lot isn't a problem (the compiler will have to work around things as they are not properly aligned). Bit masks are also inherently non-portable and differ on different compilers on different machines. Some compilers will ignore them totally.
If you really want to go ahead with it, all you do is put a colon after the member followed by the number of bits it should occupy. Here's an example:
struct TwelveHourTime
{
unsigned hour : 4; //4 bits (16), enough for 1-12
unsigned minute : 6; //6 bits (64), enough for 0-59
unsigned am : 1; //1 bit, on for am off for pm
};
// 0 1 2 3 4 5 6 7 8 9 A
// |_ _ _ _|_ _ _ _ _ _|_|
// hour minute am
You can use bit masks with a blank identifier to add some padding:
struct TwelveHourTime
{
unsigned hour : 4; //4 bits (16), enough for 1-12
unsigned minute : 6; //6 bits (64), enough for 0-59
unsigned : 5; //5 unused bits
unsigned am : 1; //1 bit, on for am off for pm
};
// 0 1 2 3 4 5 6 7 8 9 A B C D E F
// |_ _ _ _|_ _ _ _ _ _|_ _ _ _ _|_|
// hour minute (unused) am
You can a blank identifier with bit mask of 0 to force the next element to align to the boundary of the given type:
struct TwelveHourTime
{
unsigned hour : 4; //4 bits (16), enough for 1-12
char : 0; //push minute to the next byte boundary
unsigned minute : 6; //6 bits (64), enough for 0-59
unsigned am : 1; //1 bit, on for am off for pm
};
// 0 1 2 3 4 5 6 7 8 9 A B C D E
// |_ _ _ _|_ _ _ _|_ _ _ _ _ _|_|
// hour (unused) minute am
Again, bit masks are very non-portable and the exact layout depends on the compiler and platform. The diagrams are a guide only.
(advanced)
This can be achieved with a little creative union and struct work. The basic idea is to have a structure representing a superclass, which contains a union of all the other subclasses.
You will need another entity (preferably an enumeration) in the superclass and all subclasses to determine what type of struct it is.
Example:
typedef struct{
int type;
float ypos;
float xpos;
float length;
float height;
} Shape_Triangle
typedef struct{
int type;
float ypos;
float xpos;
float length;
} Shape_Square
typedef struct{
int type;
float ypos;
float xpos;
float radius;
} Shape_Circle
typedef union{
int type;
Shape_Square square;
Shape_Triangle triangle;
Shape_Circle circle;
} Shape;
...
Shape s = getShape();
switch(s.type)
{
case SHAPE_SQUARE:
s.Shape_Square.length=3;
break;
case SHAPE_TRIANGLE:
s.Shape_Triangle.height=4;
break;
case SHAPE_CIRCLE:
s.Shape_Circle.radius=5;
break;
}
A drawback of this method is that you need to duplicate the members in the substructure, and you must make sure the type variables are all in the same physical position (first is best) in the struct/union. Alternatively, use a struct for the superclass and subclass:
typedef struct{
float length;
float height;
} Shape_Triangle
typedef struct{
float length;
} Shape_Square
typedef struct{
float radius;
} Shape_Circle
typedef union{
Shape_Square square;
Shape_Triangle triangle;
Shape_Circle circle;
} Shape_Union;
typedef struct{
int type;
float xpos;
float ypos;
Shape_Union subshape;
}
...
Shape s = getShape();
switch(s.type)
{
case SHAPE_SQUARE:
s.subshape.Shape_Square.length=3;
break;
case SHAPE_TRIANGLE:
s.subshape.Shape_Triangle.height=4;
break;
case SHAPE_CIRCLE:
s.subshape.Shape_Circle.radius=5;
break;
}
This requires extra typing "through" the union.