Thread: Numbers to words

  1. #16
    Registered User
    Join Date
    Mar 2009
    Posts
    102
    Quote Originally Posted by anon View Post
    Does the problem persist if you replace calls to pow with a user-defined function that works with integers?

    Code:
    int pow10(int n)
    {
       int r = 1;
       while (n > 0) {
          r *= 10;
          --n;
       }
       return r;
    }
    Yeah it does. No change, still the same.

    Strange really, no idea what the problem is. A colleague using Linux (32 bits) tested and it seems to work all very well. I can't figure out what the problem is. Is any one of you using windows vista 32 bits?
    Last edited by Dontgiveup; 03-24-2011 at 04:18 PM.

  2. #17
    Registered User
    Join Date
    Mar 2011
    Posts
    546
    add this to your 'do' loop

    do {
    double c = (double)(counter - 1);
    printf("\n--- %f ",c);
    double p = pow(10,c);
    printf("%f ",p);
    int i = int(p);
    printf("%d ",i);
    int j = x / i;
    printf("%d ",j);
    int k = j % 10;
    printf("%d ",k);
    int z = k;
    printf("%d ",z);
    int y = (x/(int(pow(10,(double)(counter-1)))))%10;//Here is probably where
    printf("%d ---\n",y);

    ...


    here's the result i get
    Please enter a number
    75767

    --- 4.000000 10000.000000 10000 7 7 7 7 ---
    seven-
    --- 3.000000 1000.000000 1000 75 5 5 5 ---
    five-
    --- 2.000000 100.000000 100 757 7 7 7 ---
    seven-
    --- 1.000000 10.000000 10 7576 6 6 6 ---
    six-
    --- 0.000000 1.000000 1 75767 7 7 7 ---
    seven-

    the OS doesn't matter but the compiler/library version might. are you using visual C++? which version. i am testing with VC2010
    Last edited by dmh2000; 03-24-2011 at 04:26 PM.

  3. #18
    Registered User
    Join Date
    Mar 2009
    Posts
    102
    I am using DEVC++ version 4.9.9.2. Here is what I got:



    Please enter a number
    75767

    --- 4.000000 10000.000000 10000 7 7 7 7 ---
    seven-
    --- 3.000000 1000.000000 1000 75 5 5 5 ---
    five-
    --- 2.000000 100.000000 100 757 7 7 5 ---
    five-
    --- 1.000000 10.000000 10 7576 6 6 6 ---
    six-
    --- 0.000000 1.000000 1 75767 7 7 7 ---
    seven-

  4. #19
    Registered User
    Join Date
    Mar 2011
    Posts
    546
    you should change pow(10,... to pow(10.0,... because an integer in the first position means it is a little harder to understand which overload is being used. but that won't fix the problem.

    and i used 'printf' instead of 'cout' because I am too old school. that can be cout of course

    what the hell. i call compiler/library error. it is almost never that but 'y' should have been == to 'z'.

    what the hell. i call compiler/library error. it is almost never that but 'y' should have been == to 'z'.
    Last edited by dmh2000; 03-24-2011 at 04:43 PM.

  5. #20
    Registered User
    Join Date
    Mar 2009
    Posts
    102
    So I wonder where the problem is.

  6. #21
    Registered User
    Join Date
    Mar 2011
    Posts
    546
    cmath vs math.h - C++ Forum

    change the pow(10, to pow(10.0 and see what happens

  7. #22
    Registered User
    Join Date
    Mar 2009
    Posts
    102
    Changed to pow(10.0 still, the problem is there.

  8. #23
    Registered User
    Join Date
    Mar 2011
    Posts
    546
    ok, the problem is in the precision of the floating point in the computation of y. i installed dev-c++ and reproduced the problem. I looked at the assembly output of the two versions of computing 'z' and 'y' and they are slightly different.

    the 'y' version computes pow(10.0,(counter-1)) in a way that comes out slightly less than the exact value. so in the erroneous computation '5', the pow(10,(counter-1) computes 99.9999999 (instead of 100) and then the cast to int generates 99 instead of 100. when 75767 was divided by 99 it came out as 765 instead of 767 so the digit it output is 5. this is to be expected in mixed floating point and integer that the precision issues cause unexpected results.

    the right fix is to round the floating point value up to the next integer before dividing. this will fix the problem.

    try this:

    int y = (x/(int(pow(10,(counter-1))+0.5)))%10;//Here is probably where


    edit : don't use ceiling, it might cause a problem if pow return 100.000000001 whereas the +0.5 won't.
    xxxxxxxxor use the ceiling function

    xxxxxxxxint y = (x/(int(ceil(pow(10,(counter-1))))))%10;

    p.s. someone came in my office while I was working on it that delayed me.

    i can't resist a challenge lol
    Last edited by dmh2000; 03-24-2011 at 05:42 PM.

  9. #24
    Registered User
    Join Date
    Mar 2009
    Posts
    102
    Yeah you are right. Now it works flawlessly and is accurate, thanks a lot. Just out of curiosity, although I am a complete beginner, can you tell me from where you got the assembly and managed to find the "error"; just in case I get another problem and I maybe able to see what the problems is.

  10. #25
    Registered User
    Join Date
    Mar 2011
    Posts
    546
    Quote Originally Posted by Dontgiveup View Post
    Yeah you are right. Now it works flawlessly and is accurate, thanks a lot. Just out of curiosity, although I am a complete beginner, can you tell me from where you got the assembly and managed to find the "error"; just in case I get another problem and I maybe able to see what the problems is.

    when you use the Dev-C++ IDE you have a limited selection of compiler options. but in the 'bin' directory of Dev-C++, you can access the compiler directly from the command line and add more options. in the case I compiled with

    c:\dev-cpp\bin\g++ -g -c x.cpp
    <this produces x.o>
    c:\dev-cpp\bin\objdump -S x.o
    <this produces the assembler output intermixed with source>

    . looking at this, I could see that the assembly for the two computations was slightly different. I didn't figure out exactly what was happening but it gave me the clue to look closer at the floating point terms in the expression.

    and as usual I was wrong to call compiler bug. it wasn't. it is just the nature of floating point.

  11. #26
    Registered User
    Join Date
    Mar 2009
    Posts
    102
    Thank you so much. Now what if I would want to introduce numbers with a point. Example: Someone types in 1234.12 and I would like to print out the answer: one-two-three-point-one-two

  12. #27
    Gawking at stupidity
    Join Date
    Jul 2004
    Location
    Oregon, USA
    Posts
    3,218
    Quote Originally Posted by Dontgiveup View Post
    Now what if I would want to introduce numbers with a point. Example: Someone types in 1234.12 and I would like to print out the answer: one-two-three-point-one-two
    You mean: one-two-three-four-point-one-two.
    If you understand what you're doing, you're not learning anything.

  13. #28
    Registered User
    Join Date
    Mar 2009
    Posts
    102
    Quote Originally Posted by itsme86 View Post
    You mean: one-two-three-four-point-one-two.
    yeah that was a mistake, just wanted to know how I can still print out decimal numbers in words.

  14. #29
    The larch
    Join Date
    May 2006
    Posts
    3,573
    For that you'd be better off converting the value to string first.

    With floating point values what you see on the screen differs from the value in memory. For output, the value is rounded to some precision whereas the actual value is an approximation with a larger number of digits.

    Code:
    #with Python
    >>> a = 2.32
    >>> print a         #print value rounded to default precision
    2.32
    >>> print "%.16f" % (a,) #output value with a larger precision
    2.3199999999999998
    Therefore it may be quite hard to get meaningful results with just arithmetic manipulations. You can't round 2.3199999999999998 and store the result in the double, because it wasn't possible to represent the rounded value 2.32 exactly in the first place.

    All that is linked to the troubles you were having with the pow function earlier.
    I might be wrong.

    Thank you, anon. You sure know how to recognize different types of trees from quite a long way away.
    Quoted more than 1000 times (I hope).

Popular pages Recent additions subscribe to a feed

Similar Threads

  1. Seg Fault in Compare Function
    By tytelizgal in forum C Programming
    Replies: 1
    Last Post: 10-25-2008, 03:06 PM
  2. seg fault at vectornew
    By tytelizgal in forum C Programming
    Replies: 2
    Last Post: 10-25-2008, 01:22 PM
  3. converting numbers to words
    By dionys in forum C Programming
    Replies: 2
    Last Post: 05-08-2004, 09:34 AM
  4. the definition of a mathematical "average" or "mean"
    By DavidP in forum A Brief History of Cprogramming.com
    Replies: 7
    Last Post: 12-03-2002, 11:15 AM
  5. A (complex) question on numbers
    By Unregistered in forum C++ Programming
    Replies: 8
    Last Post: 02-03-2002, 06:38 PM