Show Lecture.Limits as a slide show.

CS253 Limits

C integer types in <limits.h> or <climits>

Type	Range	Type	Range
signed char	SCHAR_MIN…SCHAR_MAX	int	INT_MIN…INT_MAX
unsigned char	0…UCHAR_MAX	unsigned	0…UINT_MAX
char	CHAR_MIN…CHAR_MAX	long	LONG_MIN…LONG_MAX
short	SHRT_MIN…SHRT_MAX	unsigned long	0…ULONG_MAX
unsigned short	0…USHRT_MAX	long long	LLONG_MIN…LLONG_MAX
		unsigned long long	0…ULLONG_MAX

Also, CHAR_BIT is how many bits in a char.

C floating-point types in <float.h> or <cfloat>

Type	Mantissa digits₁₀	Exponent range	Value range
float	FLT_DIG	FLT_MIN_10_EXP…FLT_MAX_10_EXP	FLT_MIN…FLT_MAX
double	DBL_DIG	DBL_MIN_10_EXP…DBL_MAX_10_EXP	DBL_MIN…DBL_MAX
long double	LDBL_DIG	LDBL_MIN_10_EXP…LDBL_MAX_10_EXP	LDBL_MIN…LDBL_MAX

Pop Quiz

Quick, now:

What symbol says how many decimal digits are in the mantissa of a long double?

LDBL_DIG

What symbol gives the maximum value of a short?

SHRT_MAX

What symbol gives the maximum value of a size_t?

There isn’t one. It’s probably ULLONG_MAX, because size_t is often an alias for unsigned long long, but it doesn’t have to be.

C++ way

• <limits> defines the template class numeric_limits, which is specialized for all built-in types.
• It has many static constexpr methods and constants, including:
  • .min()
  • .max()
  • .digits (number of bits, excluding sign/exponent)
  • .digits10 (number of decimal digits, excluding sign/exponent)
  • .min_exponent10
  • .max_exponent10
• Why are .min() & .max() methods, not variables like .digits?

Simple Example

What is the range of values for a char?

constexpr int smallest = numeric_limits<char>::min();
constexpr int largest = numeric_limits<char>::max();
cout << "Plain char: " << smallest << "…" << largest;

Plain char: -128…127

• Look—negative characters! At least, on this computer. A char can be signed or unsigned.
• Sure, on today’s computers, a char is almost always eight bits, but not always.
• Guessing is a poor way to know—use numeric_limits!

Integer Example

cout << numeric_limits<int>::digits10 << '\n'
     << numeric_limits<int>::min()    << '\n'
     << numeric_limits<int>::max()    << '\n';

9
-2147483648
2147483647

• We use :: to we call a static method. We don’t need an instance for that.
• We’d use . if we had an instance of this class.
• digits10 is a constant, not a method, so it’s digits10, not digits10().

Floating-Point Examples

using fl = numeric_limits<float>; // a type alias
cout << fl::digits       << '\n'
     << fl::digits10     << '\n'
     << fl::min_exponent << '\n'
     << fl::max_exponent << '\n'
     << fl::epsilon()    << '\n'
     << fl::lowest()     << '\n'
     << fl::min()        << '\n'
     << fl::max()        << '\n';

24
6
-125
128
1.19209e-07
-3.40282e+38
1.17549e-38
3.40282e+38

numeric_limits<long double> ld; // an instance
cout << ld.digits       << '\n'
     << ld.digits10     << '\n'
     << ld.min_exponent << '\n'
     << ld.max_exponent << '\n'
     << ld.epsilon()    << '\n'
     << ld.lowest()     << '\n'
     << ld.min()        << '\n'
     << ld.max()        << '\n';

64
18
-16381
16384
1.0842e-19
-1.18973e+4932
3.3621e-4932
1.18973e+4932

Aliases work, too

size_t is an alias for an unsigned integer type. Which one? We don’t care. Since size_t is an alias for another type, it works with numeric_limits:

cout << boolalpha
     << numeric_limits<size_t>::is_signed  << '\n'
     << numeric_limits<size_t>::is_integer << '\n'
     << numeric_limits<size_t>::digits     << '\n'
     << numeric_limits<size_t>::digits10   << '\n'
     << numeric_limits<size_t>::min()      << '\n'
     << numeric_limits<size_t>::max()      << '\n';

false
true
64
19
0
18446744073709551615

boolalpha is an I/O manipulator. It says to display bool values as false and true, not 0 and 1.

Types

cout << numeric_limits<long>::max()  << '\n'
     << numeric_limits<int>::max()   << '\n'
     << numeric_limits<short>::max() << '\n'
     << numeric_limits<char>::max()  << '\n';

9223372036854775807
2147483647
32767
␡

• That last line sure is disappointing—I expected 127 or 255.
• The result type of ::max() and ::min() is the same type as you’re asking about.
  • Therefore, the first line of output is a value of type long.
  • Similarly, the last line is a char. When we write a char, we don’t get a numeric value; we get a character:

  cout << 'A';
  

  A

• See a few slides back, where we put the result of ::max() into an int.

A Mystery

cout << numeric_limits<size_t>::min   << '\n'
     << numeric_limits<size_t>::min() << '\n';

1
0

Why did that compile? One line must be wrong!

• The smallest value for size_t, an unsigned type, is 0. The second line is correct: .min() is a function.
• How does the first line compile?
• A function name is a pointer to the function, just like an array name is a pointer to its first element.
• A pointer-to-method can’t be displayed with <<; but bool can.
• A built-in conversion from any pointer to bool exists.
• The pointer is non-null, so converts to true, which displays as 1, without boolalpha.