ENH: Add pandas_compat.table_from_frame(df)

[kernc]

Issue

Various third parties work with pandas.DataFrames. Missing was a convenient method of transforming result DataFrame into Orange Table.

Description of changes

Add table_from_frame utility function. No hard pandas dependencies. Please advise on the best way to run the tests.

Includes

• Code changes
• Tests
• Documentation

[astaric]

AFAIAC pandas can be installed on travis (and appveyor) to run the tests while not being added to the install_requires listed in setup.py.

I assume this is how pandas tests HDFS support as there are tests in their repository and tables listed in optional-requirements, while it does not get installed when you run pip install pandas.

[codecov-io]

Codecov Report

Merging #2836 into master will increase coverage by 0.01%.
The diff coverage is 95.94%.

@@            Coverage Diff             @@
##           master    #2836      +/-   ##
==========================================
+ Coverage   81.96%   81.98%   +0.01%     
==========================================
  Files         321      323       +2     
  Lines       54841    54915      +74     
==========================================
+ Hits        44952    45023      +71     
- Misses       9889     9892       +3

[astaric]

Why don't you move the body of the test here?

[astaric]

I guess you do not want to use is_discrete_values here as it is too complicated, but you could at least use the same constant :)

[kernc]

👍

It's the exact same [universal] constant, just computed with more precision. 😜

[astaric]

Is this ever going to be different than df.shape[0] (or equivalent)?

[kernc]

Constructing Table objects, one sometimes gets so invested with details that they tend to lose grasp of the whole picture.

Thanks.

[astaric]

The code look ok to me. Squash if you intended to and I'll merge.

Since you ticked the documentation checkbox, should this be mentioned in the docs somewhere? I do not have a specific place in mind so the aswer to this could be "no", or "not yet", but I think it is a useful helper more people should know about :)

[kernc]

Was referring to reasonable docstring as provided. Guessing reviewers should be the ones to catch it until DataFrames become a more frequently integrated pattern.