Fast Exponentiation when only k digits are required - continued - c++

Where I need help...
What I want to do now is translate this solution, which calculates the mantissaof a number to c++:
n^m = exp10(m log10(n)) = exp(q (m log(n)/q)) where q = log(10)
Finding the first n digits from the result can be done like this:
"the first K digits of exp10(x) = the first K digits of exp10(frac(x))
where frac(x) = the fractional part of x = x - floor(x)."
My attempts (sparked by the math and this code) failed...:
u l l function getPrefix(long double pow /*exponent*/, long double length /*length of prefix*/)
{
long double dummy; //unused but necessary for modf
long double q = log(10);
u l l temp = floor(pow(10.0, exp(q * modf( (pow * log(2)/q), &dummy) + length - 1));
return temp;
}
If anyone out there can correctly implement this solution, I need your help!!
EDIT
Example output from my attempts:
n: 2
m: 0
n^m: 1
Calculated mantissa: 1.16334
n: 2
m: 1
n^m: 2
Calculated mantissa: 2.32667
n: 2
m: 2
n^m: 4
Calculated mantissa: 4.65335
n: 2
m: 98
n^m: 3.16913e+29
Calculated mantissa: 8.0022
n: 2
m: 99
n^m: 6.33825e+29
Calculated mantissa: 2.16596

I'd avoid pow for this. It's notoriously hard to implement correctly. There are lots of SO questions where people got burned by a bad pow implementation in their standard library.
You can also save yourself a good deal of pain by working in the natural base instead of base 10. You'll get code that looks like this:
long double foo = m * logl(n);
foo = fmodl(foo, logl(10.0)) + some_epsilon;
sprintf(some_string, "%.9Lf", expl(foo));
/* boring string parsing code here */
to compute the appropriate analogue of m log(n). Notice that the largest m * logl(n) that can arise is just a little bigger than 2e10. When you divide that by 264 and round up to the nearest power of two, you see that an ulp of foo is 2-29 at worst. This means, in particular, that you cannot get more than 8 digits out of this method using long doubles, even with a perfect implementation.
some_epsilon will be the smallest long double that makes expl(foo) always exceed the mathematically correct result; I haven't computed it exactly, but it should be on the order of 1e-9.
In light of the precision difficulties here, I might suggest using a library like MPFR instead of long doubles. You may also be able to get something to work using a double double trick and quad-precision exp, log, and fmod.

Related

distribute two integers according to ratio

Say, I have some integer n and would like to subdivide it into two other integers according to some ratio. I have some approach where I ask myself whether it does work or not.
For example: 20 with ratio 70% should be subdivided into 14,6.
The obvious solution would be:
int n = 20;
double ratio = .7;
int n1 = static_cast<int>(n * ratio);
int n2 = static_cast<int>(n * (1 - ratio));
Since the cast always floors, however, I usually underrate my result. If I use std::round, there are still cases that are not working. For example, if the first decimal place is a 5, then both numbers will be rounded up.
Some colleagues suggested: Ceil the first number and floor the second one. In most of my tests, this works, however:
1) Does it really always work, also taking into accounting possible rounding errors that naturally occur in multiplying numbers? What I think of: 20*.7 could be 14, while 20*.3 could be 5.999999. So, my sum might be 14 + 5 = 19. This is just my guess, however, I do not know whether these kind of results can or cannot occur (otherwise the answer would be simply that this kind of rounding proposition does not work)
2) Even if it does work... Why?
(I have in mind that I could just calculate number 1 by n * ratio and calculate number 2 by n - n * ratio, but I would still be interested in the answer to this question)
How about this?
int n = 20;
double ratio = .7;
int n1 = static_cast<int>(n * ratio);
int n2 = n - n1;
Here is example that confirms your suspicion and shows that the ceil+floor method doesn't always work. It is caused by the finite precision of floating point numbers on computer:
#include <iostream>
#include <cmath>
int main() {
int n = 10;
double ratio = 0.7;
int n1 = static_cast<int>(floor(n * ratio));
int n2 = static_cast<int>(ceil(n * (1.0 - ratio)));
std::cout << n1 << " " << n2 << std::endl;
}
Output:
7 4
7 + 4 is 11, so it's wrong.
Your solution doesn't always work, take a ratio of 77%, you'll get 15 and 4 (See on coliru).
Welcome to the domain of numerical analysis.
First, your computer can't always perfectly store a floating number. As you can see in the example, .77 is stored as 0.77000000000000001776 (it is an approach of the number by a sum of powers of 2).
When doing floating point calculation, you will always have a loss in precision. You can get this precision with std::numeric_limits<double>::epsilon().
Moreover, you'll still get more precision loss when converting from a floating number to an integer, and in your case the difference is big enough to give you an incoherent result.
The solution provided by #ToniBig and your last sentence has the advantage of "hiding" this loss and keep coherent data.

A program that determines the nth positive root of the p

I want to count the nth positive root of p for example we have n=2 and p=16 the answer is 4 because
4^2 = 16. I want to do this for huge numbers (1 <= n <= 200, 1 <= p < 10^101). I don't know how should I do it as fast as possible.
Example:
n=2 p=16 Answer 4
n=7 p=4357186184021382204544 Answer 1234
There are arbitrary precision math packages out there, if you have to come up with your own algorithm.
But you might try this: Get p into a double any way you can (a double can handle 10^101.) Then use math.h::pow(p, 1.0/n), and that answer will be close to the right integer (round it?). But this will fail if p is more than 15 digits, and n is too small, e.g., p = 10^100, n=2 gives a 50 digit answer, which is too big an integer for double to represent exactly.
Get 101 digit p into double: cut the number (string) into 10 digit chunks, multiply each by 10 to the appropriate power, and add them up.
Try Newton's method as described here:
http://en.wikipedia.org/wiki/Nth_root_algorithm
Take log of p, divide by n, and take the anti-log:
nthRoot(p, n) := Math.Power(10, Math.Log(p) / n)
Not sure whether you're specifically dealing with integers or what but that is the psuedo-code for it.

to find power of a decimal number with exponent as floating point number without using math library [duplicate]

This question already has answers here:
How can I write a power function myself?
(14 answers)
Closed 9 years ago.
#include<iostream>
#include<cmath>
using namespace std;
int main()
{
double x,y,z;
cin>>x>>y;
z=exp(y*log(x));
cout<<z;
system("pause");
return 0;
}
this is code to find power of a numbers whose exponent is floating point number i.e 2.3^2.3 if we do using logs and antilogs we can get the answer easily but my interview question was to find power with out using any math library in c++. i googled it and did not able to understand some of the refere nces from google.
You can always implement exp() and log() yourself.
And it's easier to actually implement 2x and log2x for the purpose and use in the same way as exp() and log().
2x = 2integer_part(x)+fractional_part(x) = 2integer_part(x) * 2fractional_part(x)
2fractional_part(x) can be calculated for -1 <= x <= +1 using Taylor series expansion.
And then multiplying by 2integer_part(x) amounts to adjusting the exponent part of the floating point number by integer_part(x) or you can indeed raise 2 to the integer power of integer_part(x) and multiply by that.
Similarly, log2x = log2(x * 2N) - N
where N (an integer, a power of 2) is chosen such that 0.5 <= x * 2N <= 1 (or, alternatively, between 1 and 2).
After choosing N, again, we can use Taylor series expansion to calculate log2(x * 2N).
And that's all, just a little bit of math.
EDIT: It's also possible to use approximating polynomials instead of Taylor series, they are more efficient. Thanks Eric Postpischil for reminding. But you'd probably need a math reference to find or construct those.
You could use Taylor series expansions for ln(x) and e^x:
ln(x) = 2 * sum[ ((x-1)/(x+1))^(2n-1) / (2n-1), n=1..inf ]
= 2 [ (x-1)/(x+1) + (1/3)( (x-1)/(x+1) )^3 + (1/5)( (x-1)/(x+1) )^5 + (1/7) ( (x-1)/(x+1) )^7 + ... ]
e^x = sum( x^n / n!, n = 0 .. inf )
= 1/1 + x/1 + x^2 / 2 + x^3 / 6 + ...
Where you could implement the integral powers as a for-loop and continue the expansion for the desired approximation. Then plug in your values, and badda-bing, badda-boom. Note the convergence regions for the above are for x > 0 for ln(x) and for all values for e^x.

An efficient way to compute mathematical constant e

The standard representation of constant e as the sum of the infinite series is very inefficient for computation, because of many division operations. So are there any alternative ways to compute the constant efficiently?
Since it's not possible to calculate every digit of 'e', you're going to have to pick a stopping point.
double precision: 16 decimal digits
For practical applications, "the 64-bit double precision floating point value that is as close as possible to the true value of 'e' -- approximately 16 decimal digits" is more than adequate.
As KennyTM said, that value has already been pre-calculated for you in the math library.
If you want to calculate it yourself, as Hans Passant pointed out, factorial already grows very fast.
The first 22 terms in the series is already overkill for calculating to that precision -- adding further terms from the series won't change the result if it's stored in a 64 bit double-precision floating point variable.
I think it will take you longer to blink than for your computer to do 22 divides. So I don't see any reason to optimize this further.
thousands, millions, or billions of decimal digits
As Matthieu M. pointed out, this value has already been calculated, and you can download it from Yee's web site.
If you want to calculate it yourself, that many digits won't fit in a standard double-precision floating-point number.
You need a "bignum" library.
As always, you can either use one of the many free bignum libraries already available, or re-invent the wheel by building your own yet another bignum library with its own special quirks.
The result -- a long file of digits -- is not terribly useful, but programs to calculate it are sometimes used as benchmarks to test the performance and accuracy of "bignum" library software, and as stress tests to check the stability and cooling capacity of new machine hardware.
One page very briefly describes the algorithms Yee uses to calculate mathematical constants.
The Wikipedia "binary splitting" article goes into much more detail.
I think the part you are looking for is the number representation:
instead of internally storing all numbers as a long series of digits before and after the decimal point (or a binary point),
Yee stores each term and each partial sum as a rational number -- as two integers, each of which is a long series of digits.
For example, say one of the worker CPUs was assigned the partial sum,
... 1/4! + 1/5! + 1/6! + ... .
Instead of doing the division first for each term, and then adding, and then returning a single million-digit fixed-point result to the manager CPU:
// extended to a million digits
1/24 + 1/120 + 1/720 => 0.0416666 + 0.0083333 + 0.00138888
that CPU can add all the terms in the series together first with rational arithmetic, and return the rational result to the manager CPU: two integers of perhaps a few hundred digits each:
// faster
1/24 + 1/120 + 1/720 => 1/24 + 840/86400 => 106560/2073600
After thousands of terms have been added together in this way, the manager CPU does the one and only division at the very end to get the decimal digits after the decimal point.
Remember to avoid PrematureOptimization, and
always ProfileBeforeOptimizing.
If you're using double or float, there is an M_E constant in math.h already.
#define M_E 2.71828182845904523536028747135266250 /* e */
There are other representions of e in http://en.wikipedia.org/wiki/Representations_of_e#As_an_infinite_series; all the them will involve division.
I'm not aware of any "faster" computation than the Taylor expansion of the series, i.e.:
e = 1/0! + 1/1! + 1/2! + ...
or
1/e = 1/0! - 1/1! + 1/2! - 1/3! + ...
Considering that these were used by A. Yee, who calculated the first 500 billion digits of e, I guess that there's not much optimising to do (or better, it could be optimised, but nobody yet found a way, AFAIK)
EDIT
A very rough implementation
#include <iostream>
#include <iomanip>
using namespace std;
double gete(int nsteps)
{
// Let's skip the first two terms
double res = 2.0;
double fact = 1;
for (int i=2; i<nsteps; i++)
{
fact *= i;
res += 1/fact;
}
return res;
}
int main()
{
cout << setprecision(50) << gete(10) << endl;
cout << setprecision(50) << gete(50) << endl;
}
Outputs
2.71828152557319224769116772222332656383514404296875
2.71828182845904553488480814849026501178741455078125
This page has a nice rundown of different calculation methods.
This is a tiny C program from Xavier Gourdon to compute 9000 decimal digits of e on your computer. A program of the same kind exists for π and for some other constants defined by mean of hypergeometric series.
[degolfed version from https://codereview.stackexchange.com/a/33019 ]
#include <stdio.h>
int main() {
int N = 9009, a[9009], x;
for (int n = N - 1; n > 0; --n) {
a[n] = 1;
}
a[1] = 2;
while (N > 9) {
int n = N--;
while (--n) {
a[n] = x % n;
x = 10 * a[n-1] + x/n;
}
printf("%d", x);
}
return 0;
}
This program [when code-golfed] has 117 characters. It can be changed to compute more digits (change the value 9009 to more) and to be faster (change the constant 10 to another power of 10 and the printf command). A not so obvious question is to find the algorithm used.
I gave this answer at CodeReviews on the question regarding computing e by its definition via Taylor series (so, other methods were not an option). The cross-post here was suggested in the comments. I've removed my remarks relevant to that other topic; Those interested in further explanations migth want to check the original post.
The solution in C (should be easy enough to adapt to adapt to C++):
#include <stdio.h>
#include <math.h>
int main ()
{
long double n = 0, f = 1;
int i;
for (i = 28; i >= 1; i--) {
f *= i; // f = 28*27*...*i = 28! / (i-1)!
n += f; // n = 28 + 28*27 + ... + 28! / (i-1)!
} // n = 28! * (1/0! + 1/1! + ... + 1/28!), f = 28!
n /= f;
printf("%.64llf\n", n);
printf("%.64llf\n", expl(1));
printf("%llg\n", n - expl(1));
printf("%d\n", n == expl(1));
}
Output:
2.7182818284590452354281681079939403389289509505033493041992187500
2.7182818284590452354281681079939403389289509505033493041992187500
0
1
There are two important points:
This code doesn't compute 1, 1*2, 1*2*3,... which is O(n^2), but computes 1*2*3*... in one pass (which is O(n)).
It starts from smaller numbers. If we tried to compute
1/1 + 1/2 + 1/6 + ... + 1/20!
and tried to add it 1/21!, we'd be adding
1/21! = 1/51090942171709440000 = 2E-20,
to 2.something, which has no effect on the result (double holds about 16 significant digits). This effect is called underflow.
However, when we start with these numbers, i.e., if we compute 1/32!+1/31!+... they all have some impact.
This solution seems in accordance to what C computes with its expl function, on my 64bit machine, compiled with gcc 4.7.2 20120921.
You may be able to gain some efficiency. Since each term involves the next factorial, some efficiency may be obtained by remembering the last value of the factorial.
e = 1 + 1/1! + 1/2! + 1/3! ...
Expanding the equation:
e = 1 + 1/(1 * 1) + 1/(1 * 1 * 2) + 1/(1 * 2 * 3) ...
Instead of computing each factorial, the denominator is multiplied by the next increment. So keeping the denominator as a variable and multiplying it will produce some optimization.
If you're ok with an approximation up to seven digits, use
3-sqrt(5/63)
2.7182819
If you want the exact value:
e = (-1)^(1/(j*pi))
where j is the imaginary unit and pi the well-known mathematical constant (Euler's Identity)
There are several "spigot" algorithms which compute digits sequentially in an unbounded manner. This is useful because you can simply calculate the "next" digit through a constant number of basic arithmetic operations, without defining beforehand how many digits you wish to produce.
These apply a series of successive transformations such that the next digit comes to the 1's place, so that they are not affected by float rounding errors. The efficiency is high because these transformations can be formulated as matrix multiplications, which reduce to integer addition and multiplication.
In short, the taylor series expansion
e = 1/0! + 1/1! + 1/2! + 1/3! ... + 1/n!
Can be rewritten by factoring out fractional parts of the factorials (note that to make the series regular we've moved 1 to the left side):
(e - 1) = 1 + (1/2)*(1 + (1/3)*(1 + (1/4)...))
We can define a series of functions f1(x) ... fn(x) thus:
f1(x) = 1 + (1/2)x
f2(x) = 1 + (1/3)x
f3(x) = 1 + (1/4)x
...
The value of e is found from the composition of all of these functions:
(e-1) = f1(f2(f3(...fn(x))))
We can observe that the value of x in each function is determined by the next function, and that each of these values is bounded on the range [1,2] - that is, for any of these functions, the value of x will be 1 <= x <= 2
Since this is the case, we can set a lower and upper bound for e by using the values 1 and 2 for x respectively:
lower(e-1) = f1(1) = 1 + (1/2)*1 = 3/2 = 1.5
upper(e-1) = f1(2) = 1 + (1/2)*2 = 2
We can increase precision by composing the functions defined above, and when a digit matches in the lower and upper bound, we know that our computed value of e is precise to that digit:
lower(e-1) = f1(f2(f3(1))) = 1 + (1/2)*(1 + (1/3)*(1 + (1/4)*1)) = 41/24 = 1.708333
upper(e-1) = f1(f2(f3(2))) = 1 + (1/2)*(1 + (1/3)*(1 + (1/4)*2)) = 7/4 = 1.75
Since the 1s and 10ths digits match, we can say that an approximation of (e-1) with precision of 10ths is 1.7. When the first digit matches between the upper and lower bounds, we subtract it off and then multiply by 10 - this way the digit in question is always in the 1's place where floating-point precision is high.
The real optimization comes from the technique in linear algebra of describing a linear function as a transformation matrix. Composing functions maps to matrix multiplication, so all of those nested functions can be reduced to simple integer multiplication and addition. The procedure of subtracting the digit and multiplying by 10 also constitutes a linear transformation, and therefore can also be accomplished by matrix multiplication.
Another explanation of the method:
http://www.hulver.com/scoop/story/2004/7/22/153549/352
The paper that describes the algorithm:
http://www.cs.ox.ac.uk/people/jeremy.gibbons/publications/spigot.pdf
A quick intro to performing linear transformations via matrix arithmetic:
https://people.math.gatech.edu/~cain/notes/cal6.pdf
NB this algorithm makes use of Mobius Transformations which are a type of linear transformation described briefly in the Gibbons paper.
From my point of view, the most efficient way to compute e up to a desired precision is to use the following representation:
e := lim (n -> inf): (1 + (1/n))^n
Especially if you choose n = 2^x, you can compute the potency with just x multiplications, since:
a^n = (a^2)^(n/2), if n % 2 = 0
The binary splitting method lends itself nicely to a template metaprogram which produces a type which represents a rational corresponding to an approximation of e. 13 iterations seems to be the maximum - any higher will produce a "integral constant overflow" error.
#include <iostream>
#include <iomanip>
template<int NUMER = 0, int DENOM = 1>
struct Rational
{
enum {NUMERATOR = NUMER};
enum {DENOMINATOR = DENOM};
static double value;
};
template<int NUMER, int DENOM>
double Rational<NUMER, DENOM>::value = static_cast<double> (NUMER) / DENOM;
template<int ITERS, class APPROX = Rational<2, 1>, int I = 2>
struct CalcE
{
typedef Rational<APPROX::NUMERATOR * I + 1, APPROX::DENOMINATOR * I> NewApprox;
typedef typename CalcE<ITERS, NewApprox, I + 1>::Result Result;
};
template<int ITERS, class APPROX>
struct CalcE<ITERS, APPROX, ITERS>
{
typedef APPROX Result;
};
int test (int argc, char* argv[])
{
std::cout << std::setprecision (9);
// ExpType is the type containing our approximation to e.
typedef CalcE<13>::Result ExpType;
// Call result() to produce the double value.
std::cout << "e ~ " << ExpType::value << std::endl;
return 0;
}
Another (non-metaprogram) templated variation will, at compile-time, calculate a double approximating e. This one doesn't have the limit on the number of iterations.
#include <iostream>
#include <iomanip>
template<int ITERS, long long NUMERATOR = 2, long long DENOMINATOR = 1, int I = 2>
struct CalcE
{
static double result ()
{
return CalcE<ITERS, NUMERATOR * I + 1, DENOMINATOR * I, I + 1>::result ();
}
};
template<int ITERS, long long NUMERATOR, long long DENOMINATOR>
struct CalcE<ITERS, NUMERATOR, DENOMINATOR, ITERS>
{
static double result ()
{
return (double)NUMERATOR / DENOMINATOR;
}
};
int main (int argc, char* argv[])
{
std::cout << std::setprecision (16);
std::cout << "e ~ " << CalcE<16>::result () << std::endl;
return 0;
}
In an optimised build the expression CalcE<16>::result () will be replaced by the actual double value.
Both are arguably quite efficient since they calculate e at compile time :-)
#nico Re:
..."faster" computation than the Taylor expansion of the series, i.e.:
e = 1/0! + 1/1! + 1/2! + ...
or
1/e = 1/0! - 1/1! + 1/2! - 1/3! + ...
Here are ways to algebraically improve the convergence of Newton’s method:
https://www.researchgate.net/publication/52005980_Improving_the_Convergence_of_Newton's_Series_Approximation_for_e
It appears to be an open question as to whether they can be used in conjunction with binary splitting to computationally speed things up. Nonetheless, here is an example from Damian Conway using Perl that illustrates the improvement in direct computational efficiency for this new approach. It’s in the section titled “𝑒 is for estimation”:
http://blogs.perl.org/users/damian_conway/2019/09/to-compute-a-constant-of-calculusa-treatise-on-multiple-ways.html
(This comment is too long to post as a reply for answer on Jun 12 '10 at 10:28)
From wikipedia replace x with 1

How many digits in this base?

The problem is to derive a formula for determining number of digits a given decimal number could have in a given base.
For example: The decimal number 100006 can be represented by 17,11,9,8,7,6,8 digits in bases 2,3,4,5,6,7,8 respectively.
Well the formula I derived so far is like this : (log10(num) /log10(base)) + 1.
in C/C++ I used this formula to compute the above given results.
long long int size = ((double)log10(num) / (double)log10(base)) + 1.0;
But sadly the formula is not giving correct answer is some cases,like these :
Number 8 in base 2 : 1,0,0,0
Number of digits: 4
Formula returned: 3
Number 64 in base 2 : 1,0,0,0,0,0,0
Number of digits: 7
Formula returned: 6
Number 64 in base 4 : 1,0,0,0
Number of digits: 4
Formula returned: 3
Number 125 in base 5 : 1,0,0,0
Number of digits: 4
Formula returned: 3
Number 128 in base 2 : 1,0,0,0,0,0,0,0
Number of digits: 8
Formula returned: 7
Number 216 in base 6 : 1,0,0,0
Number of digits: 4
Formula returned: 3
Number 243 in base 3 : 1,0,0,0,0,0
Number of digits: 6
Formula returned: 5
Number 343 in base 7 : 1,0,0,0
Number of digits: 4
Formula returned: 3
So the error is by 1 digit.I just want somebody to help me to correct the formula so that it work for every possible cases.
Edit : As per the input specification I have to deal with cases like 10000000000, i.e 10^10,I don't think log10() in either C/C++ can handle such cases ? So any other procedure/formula for this problem will be highly appreciated.
There are fast floating operations in your compiler settings. You need precise floation operations. The thing is that log10(8)/log10(2) is always 3 in math. But may be your result is 2.99999, for expample. It is bad. You must add small additive, but not 0.5. It should be about .00001 or something like that.
Almost true formula:
int size = static_cast<int>((log10((double)num) / log10((double)base)) + 1.00000001);
Really true solution
You should check the result of your formula. Compexity is O(log log n) or O(log result)!
int fast_power(int base, int s)
{
int res = 1;
while (s) {
if (s%2) {
res*=base;
s--;
} else {
s/=2;
base*=base;
}
}
return res;
}
int digits_size(int n, int base)
{
int s = int(log10(1.0*n)/log10(1.0*base)) + 1;
return fast_power(base, s) > n ? s : s+1;
}
This check is better than Brute-force test with base multiplications.
Either of the following will work:
>>> from math import *
>>> def digits(n, b=10):
... return int(1 + floor(log(n, b))) if n else 1
...
>>> def digits(n, b=10):
... return int(ceil(log(n + 1, b))) if n else 1
...
The first version is explained at mathpath.org. In the second version the + 1 is necessary to yield the correct answer for any number n that is the smallest number with d digits in base b. That is, those numbers which are written 10...0 in base b. Observe that input 0 must be treated as a special case.
Decimal examples:
>>> digits(1)
1
>>> digits(9)
1
>>> digits(10)
2
>>> digits(99)
2
>>> digits(100)
3
Binary:
>>> digits(1, 2)
1
>>> digits(2, 2)
2
>>> digits(3, 2)
2
>>> digits(4, 2)
3
>>> digits(1027, 2)
11
Edit: The OP states that the log solution may not work for large inputs. I don't know about that, but if so, the following code should not break down, because it uses integer arithmetic only (this time in C):
unsigned int
digits(unsigned long long n, unsigned long long b)
{
unsigned int d = 0;
while (d++, n /= b);
return d;
}
This code will probably be less efficient. And yes, it was written for maximum obscurity points. It simply uses the observation that every number has at least one digit, and that every divison by b which does not yield 0 implies the existence of an additional digit. A more readable version is the following:
unsigned int
digits(unsigned long long n, unsigned long long b)
{
unsigned int d = 1;
while (n /= b) {
d++;
}
return d;
}
Number of digits of a numeral in a given base
Since your formula is correct (I just tried it), I would think that it's a rounding error in your division, causing the number to be just slightly less than the integer value it should be. So when you truncate to an integer, you lose 1. Try adding an additional 0.5 to your final value (so that truncating is actually a round operation).
What you want is ceiling ( = smallest integer not greater than) logb (n+1), rather than what you're calculating right now, floor(1+logb(n)).
You might try:
int digits = (int) ceil( log((double)(n+1)) / log((double)base) );
As others have pointed out, you have rounding error, but the proposed solutions simply move the danger zone or make it smaller, they don't eliminate it. If your numbers are integers then you can verify -- using integer arithmetic -- that one power of the base is less than or equal to your number, and the next is above it (the first power is the number of digits). But if you use floating point arithmetic anywhere in the chain then you will be vulnerable to error (unless your base is a power of two, and maybe even then).
EDIT:
Here is crude but effective solution in integer arithmetic. If your integer classes can hold numbers as big as base*number, this will give the correct answer.
size = 0, k = 1;
while(k&lt=num)
{
k *= base;
size += 1;
}
Using your formula,
log(8)/log(2) + 1 = 4
the problem is in the precision of the logarithm calculation. Using
ceil(log(n+1)/log(b))
ought to resolve that problem. This isn't quite the same as
ceil(log(n)/log(b))
because this gives the answer 3 for n=8 b=2, nor is it the same as
log(n+1)/log(b) + 1
because this gives the answer 4 for n=7 b=2 (when calculated to full precision).
I actually get some curious resulting implementing and compiling the first form with g++:
double n = double(atoi(argv[1]));
double b = double(atoi(argv[2]));
int i = int(std::log(n)/std::log(b) + 1.0);
fails (IE gives the answer 3), while,
double v = std::log(n)/std::log(b) + 1.0;
int i = int(v);
succeeds (gives the answer 4). Looking at it some more I think a third form
ceil(log(n+0.5)/log(b))
would be more stable, because it avoids the "critical" case when n (or n+1 for the second form) is an integer power of b (for integer values of n).
It may be beneficial to wrap a rounding function (e.g. + 0.5) into your code somewhere: it's quite likely that the division is producing (e.g.) 2.99989787, to which 1.0 is added, giving 3.99989787 and when that's converted to an int, it gives 3.
Looks like the formula is right to me:
Number 8 in base 2 : 1,0,0,0
Number of digits: 4
Formula returned: 3
log10(8) = 0.903089
log10(2) = 0.301029
Division => 3
+1 => 4
So it's definitely just a rounding error.
Floating point rounding issues.
log10(216) / log10(6) = 2.9999999999999996
But you cannot add 0.5 as suggested, because it would not work for the following
log10(1295) = log10(6) = 3.9995691928566091 // 5, 5, 5, 5
log10(1296) = log10(6) = 4.0 // 1, 0, 0, 0, 0
Maybe using the log(value, base) function would avoid these rounding errors.
I think that the only way to get the rounding error eliminated without producing other errors is to use or implement integer logarithms.
Here is a solution in bash:
% digits() { echo $1 $2 opq | dc | sed 's/ .//g;s/.//' | wc -c; }
% digits 10000000000 42
7
static int numInBase(int num, int theBase)
{
if(num == 0) return 0;
if (num == theBase) return 1;
return 1 + numInBase(num/theBase,theBase);
}