add to_frame method to DataFrame

[letalvoj]

Add convenience method which makes sure that we can safely call to_frame on any method returning both Series and DataFrame instances, such as groupby etc.

• closes Export to excel for multiindex columns #11292 xxx (there is no issue, should I open one?)
• tests added / passed
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

[pep8speaks]

Hello @letalvoj! Thanks for submitting the PR.

• There are no PEP8 issues in the file pandas/core/frame.py !

• There are no PEP8 issues in the file pandas/tests/sparse/frame/test_frame.py !

[jreback]

-0 on this. just cluttering the API. semantically this is alway wrong, it should copy.

[letalvoj]

Sometimes I find people struggling to use groupby + apply / loc. As these methods can easily return both DataFrame / Series given not only the params but the instance on which you call it as well. IMO this is more elegant then isinstanceing.

ad. api
I get the argument regarding the API. yet in a sense all the other classes have the method

• Series.to_frame
• Panel.to_frame

even some indices

• Index.to_frame
• MultiIndex.to_frame
• DatetimeIndex.to_frame
• TimedeltaIndex.to_frame

Therefore I feel like it's not just some other random method. Or is it?

ad. copying
Oh, right. The implementation is copy-on-write, right? Is it going to be a great overhead then?
What is the idiomatic way to copy the frame in this case?

[codecov]

Codecov Report

Merging #24336 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master   #24336      +/-   ##
==========================================
+ Coverage   92.28%   92.28%   +<.01%     
==========================================
  Files         162      162              
  Lines       51833    51835       +2     
==========================================
+ Hits        47833    47835       +2     
  Misses       4000     4000

Flag	Coverage Δ
#multiple	90.68% <100%> (ø)	⬆️
#single	42.99% <50%> (-0.01%)	⬇️

Impacted Files	Coverage Δ
pandas/core/frame.py	96.91% <100%> (ø)	⬆️
pandas/core/indexes/period.py	93.06% <0%> (-0.04%)	⬇️
pandas/core/reshape/tile.py	94.82% <0%> (+0.06%)	⬆️

---

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 6c8aabc...7038a79. Read the comment docs.

[codecov]

Codecov Report

Merging #24336 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master   #24336      +/-   ##
==========================================
+ Coverage   92.28%   92.28%   +<.01%     
==========================================
  Files         162      162              
  Lines       51833    51835       +2     
==========================================
+ Hits        47833    47835       +2     
  Misses       4000     4000

Flag	Coverage Δ
#multiple	90.68% <100%> (ø)	⬆️
#single	42.99% <50%> (-0.01%)	⬇️

Impacted Files	Coverage Δ
pandas/core/frame.py	96.91% <100%> (ø)	⬆️
pandas/core/indexes/period.py	93.06% <0%> (-0.04%)	⬇️
pandas/core/reshape/tile.py	94.82% <0%> (+0.06%)	⬆️

---

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 6c8aabc...7038a79. Read the comment docs.

[jreback]

needs some discussion

[TomAugspurger]

I don't have a strong preference here. Probably +0.5 since it makes writing generic code for Series / Frame easier (although, you won't be able to use the name keyword).

FWIW, Series.to_frame doesn't copy.

In [6]: s = pd.Series([1, 2])

In [7]: r = s.to_frame()

In [8]: s[0] = 10

In [9]: r
Out[9]:
    0
0  10
1   2

though it arguably should (as an option).

[datapythonista]

@letalvoj do you mind writing a real case example on when this would be useful. I understand your point, but would be useful for the discussion to see how a real example would look with and without DataFrame.to_frame()

[jreback]

@letalvoj I think this needs an issue for discussion. If you'd like to open one with a realistic usecase we can see if there is appetite. closing this for now.