Make groupby.apply in pandas<0.25 run the function only once per group.

[ueshin]

groupby.apply in pandas<0.25 runs the functions twice for the first group, so if the function has a side effect, it could cause an unexpected behavior.

>>> import pandas as pd
>>> pd.__version__
'0.24.2'

>>> kdf = ks.DataFrame({"d": [1.0, 1.0, 1.0, 2.0, 2.0, 2.0], "v": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0]})
>>> acc = ks.utils.default_session().sparkContext.accumulator(0)
>>> def sum_with_acc_frame(x) -> ks.DataFrame[np.float64, np.float64]:
...   global acc
...   acc += 1
...   return np.sum(x)
...
>>> kdf.groupby("d").apply(sum_with_acc_frame)
    c0    c1
0  3.0   6.0
1  6.0  15.0
>>> acc.value
4

whereas:

>>> import pandas as pd
>>> pd.__version__
'0.25.3'

...

>>> acc.value
2

The acc.value should be 2 since there are only 2 groups and the function should be called twice.

[HyukjinKwon]

LGTM

[codecov-io]

Codecov Report

Merging #1462 into master will decrease coverage by 0.13%.
The diff coverage is 6.66%.

@@            Coverage Diff             @@
##           master    #1462      +/-   ##
==========================================
- Coverage   93.88%   93.75%   -0.14%     
==========================================
  Files          36       36              
  Lines        8373     8386      +13     
==========================================
+ Hits         7861     7862       +1     
- Misses        512      524      +12     

Impacted Files	Coverage Δ
databricks/koalas/groupby.py	89.74% <6.66%> (-2.16%)	⬇️

---

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 a9b4147...1d8b706. Read the comment docs.

[itholic]

LGTM. Nice fixing.