Implemented GroupBy.tail

[itholic]

This PR proposes GroupBy.tail() for DataFrameGroupBy and SeriesGroupBy.

>>> df = ks.DataFrame({'a': [1, 1, 1, 1, 2, 2, 2, 3, 3, 3],
...                    'b': [2, 3, 1, 4, 6, 9, 8, 10, 7, 5],
...                    'c': [3, 5, 2, 5, 1, 2, 6, 4, 3, 6]},
...                   columns=['a', 'b', 'c'],
...                   index=[7, 2, 4, 1, 3, 4, 9, 10, 5, 6])
>>> df
    a   b  c
7   1   2  3
2   1   3  5
4   1   1  2
1   1   4  5
3   2   6  1
4   2   9  2
9   2   8  6
10  3  10  4
5   3   7  3
6   3   5  6

>>> df.groupby('a').tail(2).sort_index()
   a  b  c
1  1  4  5
4  1  1  2
4  2  9  2
5  3  7  3
6  3  5  6
9  2  8  6

>>> df.groupby('a')['b'].tail(2).sort_index()
1    4
4    1
4    9
5    7
6    5
9    8
Name: b, dtype: int64

[itholic]

This implementation basically same as GroupBy.head() except this line - used descending order -.

[ueshin]

Then, shall we combine those two? Like:

def _limit(n, asc: bool):
   ...
   window = ... orderBy(F.col(NATURAL_ORDER_COLUMN_NAME).asc() if asc else F.col(NATURAL_ORDER_COLUMN_NAME).desc())
   ...

def head(self, n):
   return self._limit(n, asc=True)

def tail(self, n):
   return self._limit(n, asc=False)

[itholic]

Cool! let me address it. Thanks for the suggestion :)

[codecov-io]

Codecov Report

Merging #1949 (22f5e74) into master (138c7b8) will decrease coverage by 0.90%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #1949      +/-   ##
==========================================
- Coverage   94.64%   93.74%   -0.91%     
==========================================
  Files          49       49              
  Lines       10818    10839      +21     
==========================================
- Hits        10239    10161      -78     
- Misses        579      678      +99     

Impacted Files	Coverage Δ
databricks/koalas/missing/groupby.py	100.00% <ø> (ø)
databricks/koalas/groupby.py	91.50% <100.00%> (+0.08%)	⬆️
databricks/koalas/usage_logging/__init__.py	24.78% <0.00%> (-67.53%)	⬇️
databricks/koalas/usage_logging/usage_logger.py	47.82% <0.00%> (-52.18%)	⬇️
databricks/koalas/__init__.py	85.93% <0.00%> (-4.69%)	⬇️
databricks/conftest.py	97.10% <0.00%> (-2.90%)	⬇️
databricks/koalas/series.py	96.85% <0.00%> (-0.19%)	⬇️
databricks/koalas/namespace.py	84.19% <0.00%> (-0.04%)	⬇️
databricks/koalas/frame.py	96.72% <0.00%> (-0.03%)	⬇️
databricks/koalas/missing/frame.py	100.00% <0.00%> (ø)
... and 4 more

---

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 138c7b8...22f5e74. Read the comment docs.

[ueshin]

Otherwise, LGTM.

[ueshin]

Then, shall we combine those two? Like:

def _limit(n, asc: bool):
   ...
   window = ... orderBy(F.col(NATURAL_ORDER_COLUMN_NAME).asc() if asc else F.col(NATURAL_ORDER_COLUMN_NAME).desc())
   ...

def head(self, n):
   return self._limit(n, asc=True)

def tail(self, n):
   return self._limit(n, asc=False)

[xinrong-meng]

ref #1929

[xinrong-meng]

Great 👍 !

[itholic]

Thanks @ueshin @xinrong-databricks , I'd merge this now.