calculate_angle(0,45) = 112 (right answer is 111.5)

calculate_angle(10,3) = 77 (right answer is 76.5)

calculate_angle(7,19) = 105 (right answer is 105.5)

calculate_angle(2,23) = 67 (right answer is 66.5)

calculate_angle(10,47) = 41 (right answer is 41.5)

Where is rounding logic?

Sorry if my variables' names mislead you (and me also) because there is interchanging x and u.

Thx! It is nice solution but it has problem with random tests (number rounding).

=================
This is wall for author's head. Each brick is for each edge case!

Ruby tests have bad visibility.

Testing for random group of 19 individuals
You should find all the duplicates in random tests -  Expected: [#, #, #], instead got: [#, #, #]

There are several solutions for this problem and 4kyu author must know one of them.

This practice is tipical for polymorphic functions. Usualy polymorphic functions differ argument tuple but this fucntion is polymorphic because has two diffirent definitions for n > 0 and n < 0.

Why is expand(1,5) = [109601, 40320] although 109601 (numerator) contains 6 digits?
exp(1) has continued fraction (see Wolfram Alpha) [2;1,2,1,1,4,1,1,6,1,1,8,1,1,10....]
[2;1,2,1,1,4,1,1,6,1,1,8,1,1] = 49171/18089
[2;1,2,1,1,4,1,1,6,1,1,8,1,1,10] = 517656/190435
IMHO, [49171,18089] is more appropriable for expand(1,5)
If you want irreducible fraction only Taylor series then this moment must be designated in the description.
Also exp(1.85) = 6.359819523 but expand(1.85,60) = [
1255640015507986459344754396106984611112931890102125595005801691,
205688069665150755269371147819668813122841983204197482918576128].
125...1 / 205...8 = 6.1045
Nearest irreducible fraction of Taylor series is [
17212490183856113080811676174541242934582010358766327687597249
2706443181710666438004021657600000000000000000000000000000000].

In the description author doesn't write that direct simulation of queue is pointless.

Guys, you overuse extreme conditions!