Really Long Numbers

[qiemem]

I've noticed that quite a number of these problems use numbers of an extraordinary length (often 100 digits), far beyond the capabilities of a long double. Any tips for achieving these numbers? I've though about using arrays of either integers or characters, though I figured there has to be a much simpler way to go about it, as this would make multiplication in particular an extremely slow process. Is there something I'm missing? Or, if you consider this to be of a key problem to solve, any hints? Thanks!

ps- I'm using C++

[otinn]

You can't get numbers close to that in C++ without using arrays of ints/chars for the digits.
Why else would there be so few people who have solved MUL OR VFMUL?
Of course you can't use the multiplication algorithms used in school. Perhaps you should search for better algorithms.

[AntonioCS_sof]

Try using the Big Integer Library

[otinn]

Doing that would be about as useful as copy-pasting solutions.

[thechamp]

hi,
I am trying to print huge numbers using long double in C++. but this data type is not working properly.

long double x=1700000000.......000.0; (310 zeroes before decimal)
printf("%Lf",x);

the above piece of code gives 0.000000 as output even when we know that long double can support upto 4000+ digit number so it is not wrapped up for sure. than what else could be the reason? i am using dev-cpp 4.9.9.2 compiler

[leppyr64]

The same sort of thing happens with me and long long on my windows pc. Try it on ideone and see if you get a better result.

If you're trying to solve anything on SPOJ that requires integers with a double, you're likely going to be disappointed by precision errors.

[TripleM_sof]

There are two issues here:

a) Your code won't work properly on any compiler, since a constant like 1700...0.0; is interpreted as a double by default, just like 1700000000000000000000 is interpreted as an integer. You need to use 1700....0.0L; instead to show a long double, just like you use LL to show a long long.

b) Once you've done that, it will work on most compilers (eg ideone.com) but not on Dev-C++. There is a bug in Dev-C++ with long doubles where some parts think it is the same size as a double and some parts think it is longer.

Note that as Leppy says, there really isn't much point in using such numbers anyway. For example, if you do fix the problem in a) then run the code on something other than Dev-C++, it will print out something like this, since doubles and long doubles aren't very accurate:

17000000000000000000285816941724261703190.. (and a whole lot more junk)

[thechamp]

Hey guys! thank you both for quick reply. no i am not storing integer in it. while solving a problem in SPOJ i need to preprocess some data which wants me to divide two huge numbers (number of digits >400 in each) correct upto 1 decimal place.
@leppy i tried the same in ideone but it is very smart and outputs 'INF' instead of any real value
@TripleM yeah i think that appending L at the end of number should work, else i have to think a new algo

[TripleM_sof]

You won't get the quotient accurate to anywhere near 1dp due to the precision problems I mentioned above. (Divide 1700..00 by 1 and already you only get the first 20 digits right, let alone the next 290 or numbers after the decimal point.)

If you want to do that, you'll need a new algorithm.