Fix mod & rmod for matching with pandas.

[itholic]

Resolves #1398

modulo calculation for negative numbers in Koalas is different from pandas', so fixed it.

For example.

When calculate the 10 % (-3).

1. the Koalas'(and the PySpark') way: (only wanna get the result as a positive number)

• just switch the both numbers to positive and calculate.

10 % 3  # the result will be 1

2. the pandas' way:(the result can be a negative number)

• first, switch the both numbers to positive and calculate just same as Koalas.

10 % 3  # the result will be 1

• and then, plus the original negative number to the result

1 + (-3)  # the final result will be -2

You can refer the https://stackoverflow.com/questions/1082917/mod-of-negative-number-is-melting-my-brain for more detail about the difference of such a modulo.

[itholic]

I have no idea how can I handling the 'negative zero' like the below.

>>> pser
0    100.0
1      NaN
2   -300.0
3      NaN
4    500.0
5   -700.0
Name: Koalas, dtype: float64
>>> kser
0    100.0
1      NaN
2   -300.0
3      NaN
4    500.0
5   -700.0
Name: Koalas, dtype: float64

>>> pser.mod(150)
0    100.0
1      NaN
2      0.0
3      NaN
4     50.0
5     50.0
Name: Koalas, dtype: float64
>>> kser.mod(150)
0    100.0
1      NaN
2     -0.0  # << Here is the matter. how can we handle this negative zero?
3      NaN
4     50.0
5     50.0
Name: Koalas, dtype: float64

I tried to check the negative zero in the when condition and manage it, but It seems that PySpark doesn't distinguish 0.0 and -0.0.

Any idea?

[codecov-io]

Codecov Report

Merging #1399 into master will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #1399   +/-   ##
=======================================
  Coverage   95.24%   95.24%           
=======================================
  Files          34       34           
  Lines        7826     7830    +4     
=======================================
+ Hits         7454     7458    +4     
  Misses        372      372           

Impacted Files	Coverage Δ
databricks/koalas/base.py	97.55% <100.00%> (+0.04%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2579312...da68b8b. Read the comment docs.

[HyukjinKwon]

There have been ongoing issues about 0.0 vs -0.0 in Spark itself. I think you can avoid to address it here.

[itholic]

There have been ongoing issues about 0.0 vs -0.0 in Spark itself. I think you can avoid to address it here.

I made it works properly for every cases by fixing the existing calculation properly.

[ueshin]

LGTM.

[ueshin]

Thanks. merging.
Since I have a concern, I'll submit a follow-up PR soon.