Add DataFrame.plot.kde() (an alias of 'density')

[itholic]

This PR implements kde in DataFrame. It reuses Series' implementation. at #767

like Series's kde plot, Since DataFrame's also uses MLlib's KernelDensity API to calculate KDE , so slightly different from pandas but seems good enough either:

And also kde is an alias of 'density', you can get exactly same result when you use 'density' rather than 'kde' like below:

Multiple columns examples:

and for each row as Series.plot.kde looks same like below:

[HyukjinKwon]

@itholic, can you show an example with more than two columns?

[itholic]

@HyukjinKwon Sure, Here is

and for each row as Series.plot.kde looks same like below:

[HyukjinKwon]

@itholic, looks good. Can you fix the tests?

[itholic]

@HyukjinKwon i'm still working on it, but these tests are still failed, i'm trying to find out solution.

[ueshin]

@HyukjinKwon Let me share the question here, too. #767 (comment) which might relate to this PR.

[HyukjinKwon]

@itholic can you check ^? I'll check it tomorrow.

[itholic]

@HyukjinKwon Sure, i'll check this out. sorry to late reply (it was very busy today 😭 )

[HyukjinKwon]

There seems an issue about plot comparison. plot image has to be closed for each plot but seems not. So, overlapped plot images are being compared. For instance,

we should compare:

but now we're comparing:

[itholic]

@HyukjinKwon Thanks for review !! i just resolved all that you commented.

[softagram-bot]

Softagram Impact Report for pull/784 (head commit: d8fbcb8)

⭐ Change Overview

(Open in Softagram Desktop for full details)

💡 Insights

• Co-change Alert: You modified frame.py. Often test_dataframe.py (koalas/tests) is modified at the same time.

📄 Full report

• Permalink: Full report for pull/784

Impact Report explained. Give feedback on this report to [email protected]

[codecov-io]

Codecov Report

Merging #784 into master will decrease coverage by 0.03%.
The diff coverage is 66.66%.

@@            Coverage Diff             @@
##           master     #784      +/-   ##
==========================================
- Coverage   94.29%   94.26%   -0.04%     
==========================================
  Files          32       32              
  Lines        5911     5913       +2     
==========================================
  Hits         5574     5574              
- Misses        337      339       +2

Impacted Files	Coverage Δ
databricks/koalas/plot.py	94.04% <66.66%> (-0.26%)	⬇️
databricks/koalas/frame.py	96.76% <66.66%> (-0.07%)	⬇️

---

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 9186870...d8fbcb8. Read the comment docs.

[HyukjinKwon]

@ueshin, I am going to rewrite the plot tests soon. Let me merge this for now.

[ueshin]

Sounds good, thanks!

[HyukjinKwon]

I made a PR #838

BTW, @itholic, can you add some examples for plot implementations in their docstrings? Basically we finished the rough implementation of all plots (except 2 in DataFrame). So it should be better to have some examples.

Can you also check if we're able to post some images? e.g. https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.plot.box.html#pandas.DataFrame.plot.box

[itholic]

@HyukjinKwon sure, I'm going to add & check them 😃

[ueshin]

Seems like DataFrame.plot.hist is listed twice.
I'm not sure we should keep DataFrame.hist, though.

[HyukjinKwon]

@itholic can you fix this? DataFrame.hist seems in the pandas doc as an alias. Let's keep it for now.

[itholic]

sure, fixed in #851
Thanks for comment !!