Talk:cpp/types/numeric limits

Layout change proposal

How about we change the layout from tables to the list of members/constants, similar to most other pages? It gives us the ability to reference specific fnctions/constants individually from elsewhere, gives us a neat sidebar, and lets us expound on the details of any important member with examples and references within its own page, instead of being limited to a cell in a table. Who can tell from the table blurb what in the world does "tinyness_before" mean or how "digits10" is different from "max_digits10"?

Here's how it would look, right after the list of standard specializations,

Member constants

is_specialized [static]	identifies types for which std::numeric_limits is specialized (public static member constant)
is_signed [static]	identifies signed types (public static member constant)
is_integer [static]	identifies integer types (public static member constant)
is_exact [static]	identifies exact types (public static member constant)
has_infinity [static]	identifies floating-point types that can represent the special value "positive infinity" (public static member constant)
has_quiet_NaN [static]	identifies floating-point types that can represent the special value "quiet not-a-number" (NaN) (public static member constant)
has_signaling_NaN [static]	identifies floating-point types that can represent the special value "signaling not-a-number" (NaN) (public static member constant)
has_denorm [static]	identifies the denormalization style used by the floating-point type (public static member constant)
has_denorm_loss [static]	identifies the floating-point types that detect loss of precision as denormalization loss rather than inexact result (public static member constant)
round_style [static]	identifies the rounding style used by the type (public static member constant)
is_iec559 [static]	identifies the IEC 559/IEEE 754 floating-point types (public static member constant)
is_bounded [static]	identifies types that represent a finite set of values (public static member constant)
is_modulo [static]	identifies types that handle overflows with modulo arithmetic (public static member constant)
digits [static]	number of radix digits that can be represented without change (public static member constant)
digits10 [static]	number of decimal digits that can be represented without change (public static member constant)
max_digits10 [static]	number of decimal digits necessary to differentiate all values of this type (public static member constant)
radix [static]	the radix or integer base used by the representation of the given type (public static member constant)
min_exponent [static]	one more than the smallest negative power of the radix that is a valid normalized floating-point value (public static member constant)
min_exponent10 [static]	the smallest negative power of ten that is a valid normalized floating-point value (public static member constant)
max_exponent [static]	one more than the largest integer power of the radix that is a valid finite floating-point value (public static member constant)
max_exponent10 [static]	the largest integer power of 10 that is a valid finite floating-point value (public static member constant)
traps [static]	identifies types which can cause arithmetic operations to trap (public static member constant)
tinyness_before [static]	identifies floating-point types that detect tinyness before rounding (public static member constant)

Member functions

min [static]	returns the smallest finite value of the given type (public static member function)
lowest [static] (C++11)	returns the lowest finite value of the given type (public static member function)
max [static]	returns the largest finite value of the given type (public static member function)
epsilon [static]	returns the difference between 1.0 and the next representable value of this type (public static member function)
round_error [static]	returns the maximum possible ISO 10967 rounding error for this type (public static member function)
infinity [static]	returns the positive infinity value of the given floating-point type (public static member function)
quiet_NaN [static]	returns the quiet NaN value of the given floating-point type (public static member function)
signaling_NaN [static]	returns the signaling NaN value of the given floating-point type (public static member function)
denorm_min [static]	returns the smallest positive denormalized value of the given floating-point type (public static member function)

Helper classes

float_round_style	indicates floating-point rounding modes (class)
float_denorm_style	indicates floating-point denormalization modes (class)

And then within each page, we could list the default value for non-specialized numeric_limits, and the specified values for fundamental types (although the big at-a-glance table in the end of this page is useful too and should be kept, perhaps split between integer and floating-point though), give examples, etc.

--Cubbi 07:40, 5 October 2011 (PDT)

Looks good to me. Something like this should've been implemented from the start.P12 11:05, 5 October 2011 (PDT)

Yup, much better. Nate 15:19, 7 October 2011 (PDT)

specialization for user-defined types

I think that what's in the page now may be incorrect, and that std::numeric_limits cannot be extended to library-defined types, but can be extended by implementations to cover arithmetic types like __int128.

http://stackoverflow.com/questions/16122912/is-it-ok-to-specialize-stdnumeric-limitst-for-user-defined-number-like-class