[fix] handle null column_name in druid/sqla models

[graceguo-supercat]

when sqla table column is null, see following exception:

may related with #5445
@john-bodley @michellethomas @soboko @mistercrunch

[codecov-io]

Codecov Report

Merging #7063 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #7063      +/-   ##
==========================================
+ Coverage   64.57%   64.57%   +<.01%     
==========================================
  Files         422      422              
  Lines       20576    20577       +1     
  Branches     2251     2251              
==========================================
+ Hits        13286    13287       +1     
  Misses       7168     7168              
  Partials      122      122

Impacted Files	Coverage Δ
superset/connectors/base/models.py	89.89% <100%> (+0.05%)	⬆️
superset/connectors/druid/models.py	82.74% <100%> (ø)	⬆️

---

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 33d67da...9857d04. Read the comment docs.

[john-bodley]

@graceguo-supercat the issue here is we would be replacing None with an empty string in the sorted list. We want to keep the sorted data intact so the following should work:

sorted([c.colum_name for c in self.columns], key=lambda x: x or '')

[john-bodley]

I think your investigation surfaced that the are columns without a name which seems bad. I can look into why there are columns without names and produce a migration if needed.

I'm a little perplexed what order_by_choices means for empty strings and how this is used.

[graceguo-supercat]

@john-bodley I found an example, it is used by Ordering control in explore view:

datasource is part of bootstrap data which is sent to front-end at page load time:

[john-bodley]

@graceguo-supercat thanks for trouble shooting this. I plan on creating a migration tomorrow to remedy this issue which is related to ill-defined data as opposed to an underlying issue with the code.

[john-bodley]

@graceguo-supercat I think #7067 should resolve these issues.

[graceguo-supercat]

maybe useful for a quick solution in 0.31

[john-bodley]

Now that #7067 has been closed (which was slated for v0.31) in favor of #5452 and #5453 which will not make it into v0.31 I agree that this is the best interim solution moving forward. Once these PRs are merged we can revert this PR.