;Conversion and addition of [[wp:P-adic_number|p-adic Numbers]].


;Task.

Convert two rationals to p-adic numbers and add them up.
Rational reconstruction is needed to interpret the result.

p-Adic numbers were introduced around 1900 by Hensel. p-Adic expansions
(a series of digits 0 ≤ d < p times p-power weights)
are finite-tailed and tend to zero in the direction of higher positive
powers of p (to the left in the notation used here).
For example, the number 4 (100.0) has ''smaller'' 2-adic norm than 1/4 (0.01).

If we convert a natural number, the familiar p-ary expansion is obtained:
10 decimal is 1010 both binary and 2-adic. To convert a rational number a/b
we perform p-adic long division. If p is actually prime, this is always possible
if first the 'p-part' is removed from b (and the p-adic point shifted accordingly).
The inverse of b modulo p is then used in the conversion.

'''Recipe:''' at each step the most significant digit of the partial remainder
(initially a) is zeroed by subtracting a proper multiple of the divisor b.
Shift out the zero digit (divide by p) and repeat until the remainder is zero
or the precision limit is reached. Because p-adic division starts from the right,
the 'proper multiplier' is simply
d = partial remainder * 1/b (mod p).
The d's are the successive p-adic digits to find.

Addition proceeds as usual, with carry from the right to the leftmost term,
where it has least magnitude and just drops off. We can work with approximate rationals
and obtain exact results. The routine for rational reconstruction demonstrates this:
repeatedly add a p-adic to itself (keeping count to determine the denominator),
until an integer is reached (the numerator then equals the weighted digit sum).
But even p-adic arithmetic fails if the precision is too low. The examples mostly
set the shortest prime-exponent combinations that allow valid reconstruction.


;Related task.

[[p-Adic square roots]]


;Reference.

[https://www.cut-the-knot.org/blue/p-adicExpansion.shtml] p-Adic expansions


__TOC__