signed vs unsigned

Discussion in 'C++' started by LuB, Feb 21, 2006.

  1. LuB

    LuB Guest

    This isn't a C++ question per se ... but rather, I'm posting this bcs I
    want the answer from a C++ language perspective. Hope that makes sense.

    I was reading Peter van der Linden's "Expert C Programming: Deep C
    Secrets" and came across the following statement:

    "Avoid unnecessary complexity by minimizing your use of unsigned types.
    Specifically, don't use an unsigned type to represent a quantity just
    because it will never be negative (e.g., "age" or "national_debt")."

    Admittedly, I minimize comments in my code. I use them when necessary
    but try to limit them to one or two lines. I just don't like their
    aesthetic affect in the source. I find the code much harder to read
    littered with paragraphs of explanations.

    I find the code easier to read when broken up with appropriate newlines
    and small, short comments acting as headings.

    So - that effectually means that I try my darndest to write
    self-describing code. Shorter functions, self-explanatory names for
    functions and variable names, etc. Sometimes exceessive commenting is
    necessary, but as a whole, I tend to avoid it.

    I'm really enjoying Peter's book, but I find this comment hard to
    swallow considering that if an age or array index can never be negative
    - I would want to illustrate that with an apporpriate choice of type -
    namely, unsigned int.

    Am I in the minority here? Is my predilection considered poor style?

    I guess, from the compiler's standpoint ... using int everywhere is
    more portable ... since comparison's against unsigned int can vary
    between K&R and ANSI C.

    I'm not incurring some type of performance penalty for such decisions
    am I?

    Thanks in advance,

    -Luther
     
    LuB, Feb 21, 2006
    #1
    1. Advertisements

  2. Hopefully not!
    Absolutely not!

    The key to dealing with signed vs. unsigned is not to mix them, if
    possible, in comparisons. Be especially wary of comparing an unsigned
    counter value to 0 while decrementing the counter.
     
    Bob Hairgrove, Feb 21, 2006
    #2
    1. Advertisements

  3. LuB

    Cory Nelson Guest

    Like you said, marking a type as "unsigned" if for nothing else is an
    easy documentation for other developers that it will never be negative.
    I havn't read his stuff but I can't imagine what complexities he is
    talking about.

    Performance differences depend largly on your architecture. On AMD64,
    division/modulus are faster when using unsigned types. Conversion to
    floating point is faster with signed types. Addition/subtraction are
    the same.
     
    Cory Nelson, Feb 21, 2006
    #3
  4. LuB

    roberts.noah Guest

    The only reason that I can think of that you might want to not use
    'unsigned' is if the assumption that it will never be negative might
    change. However, most times you are better off with the assumption and
    the protection of the unsigned type and then change it later if you
    have to. Minimize dependencies on the type so that it is easier to do.

    The great thing about unsigned is if the value can't be negative and
    you use signed then you always have to check it. Simply defining the
    type as unsigned gets rid of all that as well as documenting your
    domain.
     
    roberts.noah, Feb 21, 2006
    #4
  5. LuB

    Tomás Guest

    LuB posted:

    My first rule is to write "const" wherever I can.
    (except for return types).

    My second rule is to write "unsigned" wherever I can.


    Thus I'll write:

    unsigned GetDogAge(unsigned const age)
    {
    return age * 7;
    }

    Or alternatively re-use the parameter variable:

    unsigned GetDogAge(unsigned age)
    {
    return age *= 7;
    }


    -Tomás
     
    Tomás, Feb 21, 2006
    #5
  6. LuB

    Daniel T. Guest

    And the bad thing about unsigned is that if code would otherwise make
    the value negative, you can't check it. IE

    void foo( unsigned s ) {
    // at this point s == 4294966272
    // is it an error (s came in as -1024) or
    // does the client really want us to deal with that number?
    }

    Bjarne Stroustrup says, "The unsigned integer types are ideal for uses
    that treat storage as a bit array. Using an unsigned instead of an int
    to gain one more bit to represent positive integers is almost never a
    good idea. Attempts to ensure that some values are positive by declaring
    variables unsigned will typically be defeated by the implicit conversion
    rules."
     
    Daniel T., Feb 21, 2006
    #6
  7. LuB

    roberts.noah Guest

    Performs an unnecissary and unused assignment as age is automatic and
    will be gone after return.
     
    roberts.noah, Feb 21, 2006
    #7
  8. LuB

    forkazoo Guest

    Well, it is a style issue, so many people will probably disagree with
    me. They are probably all right, and I would still be able to respect
    them :)

    That said, I personally use signed types almost all the time, even if
    something should never be negative. Every once in a while, I make a
    horribly stupid error in something like a file loading function.
    Suppose an object can have zero or more child objects. If my loader
    returns an object that claims to have -1024 children, then I know I've
    done something wrong, and it can be easier to track down bugs that
    compile and don't crash. If it just claims that there are a large
    positive number of children, it's not as easy to catch the error
    condition. There is even the wildly unlikely possibility that there is
    an error due to a hardware fault like bad RAM. (Though, the problems
    are almost always caused by me! Many people who are better programmers
    than me may find that hardware accounts for a greater percentage of
    problems than I do :)

    Also, I use negative values as intentional error codes and special
    cases. Some people consider this horrible style. I can respect that.
    But, for a lot of the things I do, it is the quickest, easiest,
    simplest, and clearest way to have a recoverable error. So, my
    hypothetical object loading function might return an object which
    claims -1 children if it failed to open the file, -2 if it failed to
    parse the file, -3 if the file contained invalid information, etc. I
    know, exceptions are probably better for most of these things, but my
    personal style is to write C++ that looks a lot like C, and uses
    occasional C++ features. It's just more inline with the way I think
    about the problems.

    Also, by always using signed types, I can avoid comparisons between
    signed and unsigned. It is also generally easier to catch overflow.
    After all, the national debt might never be negative, but it certainly
    might exceed MAX_INT on some systems! :)

    I tend to work alone on personal projects. So, the most important
    thing for me is that my personal coding style is readable - to me.
    Feel free to strongly disagree with my style.
     
    forkazoo, Feb 21, 2006
    #8
  9. LuB

    Neil Cerutti Guest

    I don't agree with that rule.
    What do you expect to happen if a client passes in an negative int?

    The argument is not in the domain, yet the error has been rendered
    impossible to detect or recover from.
     
    Neil Cerutti, Feb 21, 2006
    #9
  10. * Neil Cerutti:
    I don't disagree with your viewpoint regarding using or not using
    unsigned whenever possible; I think both are valid viewpoints, and as
    with indentation the main thing is to be consistent in one's choices.

    However, I disagree with your reason!

    With the unsigned argument a validity test might go like

    assert( age < 200 ); // unsigned validity test.

    With a signed argument the test might go like

    assert( age >= 0 ); // signed validity test.
    assert( age < 200 ); // more signed validity test.

    Now, first of all that demonstrates the "impossible to detect" is simply
    incorrect, and second, in my view it demonstrates a slight superiority
    for unsigned in this particular case, with respect to validity testing.

    Cheers,

    - Alf
     
    Alf P. Steinbach, Feb 21, 2006
    #10
  11. LuB

    roberts.noah Guest

    I believe he is talking about something like this:


    #include <iostream>
    using namespace std;

    unsigned int GetDogAge(unsigned int age)
    {
    return age * 7;
    }

    int main(int argc, char* argv[])
    {
    int t = -5;
    unsigned int x = GetDogAge(t);

    cout << x;

    int y; cin >> y;

    return 0;
    }

    ------ Build started: Project: Playground, Configuration: Debug Win32
    ------

    Compiling...
    Playground.cpp
    Linking...

    Build log was saved at
    "file://c:\src\Playground\Playground\Debug\BuildLog.htm"
    Playground - 0 error(s), 0 warning(s)


    ---------------------- Done ----------------------

    Build: 1 succeeded, 0 failed, 0 skipped

    Output is, "4294967261."



    It is an interesting point.
     
    roberts.noah, Feb 21, 2006
    #11
  12. LuB

    andy Guest

    Sounds like a good excuse to write an unsigned_int class:

    #include <boost/utility/enable_if.hpp>
    #include <boost/numeric/conversion/converter.hpp>
    #include <boost/type_traits/is_unsigned.hpp>

    class unsigned_int{
    unsigned int m_value;
    public:
    template <typename T>
    unsigned_int(T const & rhs,
    typename boost::enable_if<
    )
    {
    // further features
    // throws on overflow when T is bigger eg unsigned long
    boost::numeric::converter<unsigned int,T> convert;
    m_value = convert(rhs);
    }
    };

    int main()
    {
    unsigned_int n1 = 1U;
    unsigned_int n2 = static_cast<unsigned short>(1);
    unsigned_int n3 = 1UL;
    unsigned_int n4 = static_cast<unsigned int>(1);

    unsigned_int n5 = 1; // Error!
    }

    reagrds
    Andy Little
     
    andy, Feb 21, 2006
    #12
  13. LuB

    andy Guest

    Sounds like a good excuse for an unsigned_int class ...

    #include <boost/utility/enable_if.hpp>
    #include <boost/numeric/conversion/converter.hpp>
    #include <boost/type_traits/is_unsigned.hpp>

    class unsigned_int{
    unsigned int m_value;
    public:
    template <typename T>
    unsigned_int(T const & rhs,
    typename boost::enable_if<
    )
    {
    // further features
    // throws on overflow when T is bigger eg unsigned long
    boost::numeric::converter<unsigned int,T> convert;
    m_value = convert(rhs);
    }
    };

    int main()
    {
    unsigned_int n1 = 1U;
    unsigned_int n2 = static_cast<unsigned short>(1);
    unsigned_int n3 = 1UL;
    unsigned_int n4 = static_cast<unsigned int>(1);

    unsigned_int n5 = 1; // Error!
    }

    reagrds
    Andy Little
     
    andy, Feb 21, 2006
    #13
  14. LuB

    andy Guest

    Sounds like a good excuse to write an unsigned_int class:

    #include <boost/utility/enable_if.hpp>
    #include <boost/numeric/conversion/converter.hpp>
    #include <boost/type_traits/is_unsigned.hpp>

    class unsigned_int{
    unsigned int m_value;
    public:
    template <typename T>
    unsigned_int(T const & rhs,
    typename boost::enable_if<
    )
    {
    // further features
    // throws on overflow when T is bigger eg unsigned long
    boost::numeric::converter<unsigned int,T> convert;
    m_value = convert(rhs);
    }
    };

    int main()
    {
    unsigned_int n1 = 1U;
    unsigned_int n2 = static_cast<unsigned short>(1);
    unsigned_int n3 = 1UL;
    unsigned_int n4 = static_cast<unsigned int>(1);

    unsigned_int n5 = 1; // Error!
    }

    reagrds
    Andy Little
     
    andy, Feb 21, 2006
    #14
  15. LuB

    andy Guest

    wrote:

    Sorry
    Sorry
    Sorry for posting 3 times. Browser hung up

    regards
    Andy Little
     
    andy, Feb 21, 2006
    #15
  16. * :
    Yes, that would be caught by the validity test for unsigned formal arg,

    assert( age < 200 );

    Cheers,

    - Alf
     
    Alf P. Steinbach, Feb 21, 2006
    #16
  17. LuB

    wkaras Guest

    It's like driving on the left or driving on the right. It's basically
    an
    arbitrary choice (now that we drive cars), it's just good if everyone
    makes the same arbitrary choice. That is if there's more than
    just you on the road, which is analagous to maintaining
    existing code or heavily using libraries written by others.

    Many compilers generate (to me helpful) warnings if you do a
    Many also generate (to me useless) warnings if you do a
    == or != comparison between a signed and unsigned. So,
    you can end up with alot of static_casts to get rid of the
    warnings.

    For some reason, if an unsigned type is used in an expression
    where it has to be implicitly converted to a larger type, it
    will be converted to a signed larger type. As far as I can
    see, this rule has no practical effect, other than to give
    some compilers an excuse to generate unnecessary
    warnings. (I guess it could change the results when
    using bitwise ops on variables of different sizes, which
    to me is a bad idea anyway.) So, that's one reason to prefer
    signed. But, like you, I prefer to use unsigned for variables that
    should always be nonnegative, to make the intent clear.
     
    wkaras, Feb 21, 2006
    #17
  18. LuB

    Ben Pope Guest

    Surely overflow and underflow are a problem with both signed and unsigned?

    These are often situations where you are unable to check validity
    without domain specific knowledge - I don't see how signed unsigned vary
    in that regard.

    Ben Pope
     
    Ben Pope, Feb 22, 2006
    #18
  19. LuB

    Gavin Deane Guest

    Self-explanatory names are good.
    Why would you want to illustrate it with an appropriate choice of type?
    I would want to illustrate it with a self-explanatory name.
    This question comes up regularly. Search the history of the newsgroup
    and you will find plenty of discussion. But I am slightly surprised to
    see this thread get to nearly 20 posts (as far as I can see so far)
    without anyone mentioning the subtraction issue.

    Personally, I do not accept that argument for unsigned as documentation
    to mean "negative values make no sense here". The variable name (age,
    buffer_size, radius) serves that purpose. Comparison with const is not
    valid. const is useful because it is enforced by the compiler. unsigned
    is not enforced. -1 years does not make sense for age, but neither does
    4294967295 years. At best, using unsigned has only documented one end
    of your range. And since the variable name has already achieved that,
    you haven't gained anything.

    But that's not the only reason I don't use unsigned for these sort of
    quantities. In general, subtracting one person's age from another in
    the real world yields a signed result but in C++ subtracting one
    unsigned quantity from another yields an unsigned result.

    <quote Kaz Kylheku (in post number 28 as I write this)>
    http://groups.google.co.uk/group/co...f8?q="Thus+code+like"&rnum=1#0c236f41f9de66f8

    Thus code like if (buffer_space - requested_payload > header_size)
    is broken if unsigned integers are used, and there was no prior check
    that the requested_payload isn't larger than the buffer.

    </quote>

    Both signed and unsigned arithmetic in C++ behave just like real world
    integer arithmetic until you reach the minimum and maximum values for
    those types. The difference is that for signed, the correlation between
    the real world and C++ breaks down at + or - a very big number, whereas
    for unsigned, the correlation between the real world and C++ breaks
    down at zero and a (different) very big number.

    In other words, if I use unsigned, I am very close to the edge of the
    domain in which C++ arithmetic corresponds to the real world. If I use
    signed, I am comfortably in the middle of that domain and the
    extremities where I have to start coding for special cases are far, far
    away from any number that I am likely to ever want to use for an age or
    a length or a size, or I am likely to ever end up with by doing
    arithmetic with those quantities.

    Gavin Deane
     
    Gavin Deane, Feb 22, 2006
    #19
  20. LuB

    Ben Pope Guest

    That's a very good point.
    In an ideal world, we would probably all do range checking more often
    than we do.

    I always say you make your own luck, with signed types you're more lucky!

    Ben Pope
     
    Ben Pope, Feb 22, 2006
    #20
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.