.. | ||
src | ||
Cargo.lock | ||
Cargo.toml | ||
README.md |
Solution Day 7
Calculating expression and testing whether we can make them equal on both sides. Not too hard.
Parsing the input is rather easy with a bit of .split()
and .parse()
. Algorithm for testing the
expressions is of the brute-force variety: generate all possible combinations of operators, calculate
the expression for each combination and see if it matches the result.
Task 1
This one was easy at first: all possible combinations of operators (since there are only 2) can be generated
by counting between 0
and pow(2, size)
(where size
is the number of operators we need to fill the equation)
and treating each number as a bit mask, where the bit in position n
represents an operator (Add
if the bit
is 0
, Multiply
if it is 1
). This was easy enough to implement and I even got it right on the first try.
Task 2
I kind of had the inkling, that my simple "bitmask" solution for task 1 would get me into trouble with task 2. And
wouldn't you know? It did. Three operators cannot be easily expressed with a binary number (more on that in a bit).
I thought about implementing symbolic calculations (i.e. representing the number as Vec<usize>
and doing the
counting by hand, including overflow and everything), but that seemed too much effort. The crate itertools
contains
a .combinations()
method, which can be made to work like this: [Add, Add, Multiply, Multiply, Concat, Concat].combinations(2)
(not the actual syntax, but you get the idea). This will generate all possible length 2 combination of the operators.
Works, but is exceedingly slow, so I abandoned that approach.
My actual solution goes back to my teaching days at uni: calculating the digits of a number in arbitrary base (in
this case: base 3). To get the next digit at the back, calculate number % base
. Then continue with number /= base
to
get the number without the last digit and continue until you run out of digits (or, in my case: until you've generate enough
digits corresponding to the operators needed).
Another tricky bit was the implementation of the ||
operator. Since the whole implementation is done numerically, it
should be implemented as left * pow(10, round(log10(right))) + right
(basically: multiply with 10 ^ number of right digits to
get enough 0s at the end and add the number on top). While this works in theory, there is probably a precision or something,
since ceil
and log10
only work in f64
instead of u64
. I consistently got the wrong results. In the end I gave up and
used a combination of format!()
and .parse()
to concatenate in string-space.