Making thrift, pyhive and tableschema as extra_requires

[mistercrunch]

Looking at the dependency tree for license related questions, I noticed
that tableschema had a huge tree, and only people running Hive really
need it. Making this as well as pyhive and thrift optional.

[codecov-io]

Codecov Report

Merging #6696 into master will decrease coverage by 0.01%.
The diff coverage is 56.25%.

@@            Coverage Diff             @@
##           master    #6696      +/-   ##
==========================================
- Coverage   56.14%   56.12%   -0.02%     
==========================================
  Files         521      521              
  Lines       23179    23173       -6     
  Branches     2769     2769              
==========================================
- Hits        13014    13006       -8     
- Misses       9756     9758       +2     
  Partials      409      409

Impacted Files	Coverage Δ
superset/db_engines/hive.py	0% <0%> (ø)	⬆️
superset/db_engine_specs.py	54.08% <0%> (-0.21%)	⬇️
superset/connectors/sqla/views.py	63.06% <0%> (-0.33%)	⬇️
superset/utils/core.py	88.4% <100%> (-0.05%)	⬇️
superset/dataframe.py	94.64% <100%> (-0.05%)	⬇️
superset/viz.py	71.66% <100%> (-0.02%)	⬇️
superset/connectors/base/models.py	90.37% <100%> (-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 057c43c...f74b6d0. Read the comment docs.

[john-bodley]

My only comment is whether these optional dependencies should reside in the requirements.txt file. I can see the merit in having them there as one could argue it represents the frozen state of the union of all viable configurations.

I speculate we may need to add a requirements.in file to ensure that these are capture if someone wanted to regenerate the requirements.txt file in the future (see this for context).

[mistercrunch]

Good call. I just didn't think about running pip-compile. I ran it just now and it removes the deps from requirements.txt which I think is what we want.

Now given that the extra_requires is optional, we could just be more prescriptive and pin things in that section of setup.py if needed/appropriate. Being prescriptive inside something optional is very different than being prescriptive in something that is forced/required.

[john-bodley]

@mistercrunch I think the following would work if the desire is to pin the optional packages (and their dependencies) especially for testing purposes:

> cat requirements.in
-e .[gsheets,hive,presto]
> pip-compile --output-file requirements.txt requirements.in
...

[michellethomas]

@mistercrunch this PR got into 0.31 so this should be in the 0.31 section. Also I think we should call out the specific dependencies that were removed, we were impacted by removing flower would be good to list that out.

[mistercrunch]

Oh I assumed this would hit 0.32... let me change that then...

[mistercrunch]

Here: #7117