bugfix: cli export_datasources (fixes 7437)

[Aylr]

CATEGORY

Choose one

• Bug Fix
• Enhancement (new features, refinement)
• Refactor
• Add tests
• Build / Development Environment
• Documentation

SUMMARY

1. I needed a quick fix to move some datasources over to a new deployment.
2. The dict sorting wasn't working.
3. I implemented the quickest possible (read janky) fix by coercing to string.

BEFORE/AFTER SCREENSHOTS OR ANIMATED GIF

TEST PLAN

1. Checkout & install
2. superset db upgrade
3. superset init
4. superset load_examples
5. superset export_datasources and note the yml output.

ADDITIONAL INFORMATION

• Has associated issue: Fixes Exporting Datasource to YAML fails #7437
• Changes UI
• Requires DB Migration.
• Confirm DB Migration upgrade and downgrade tested.
• Introduces new feature or API
• Removes existing feature or API

REVIEWERS

[codecov-io]

Codecov Report

Merging #7728 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #7728   +/-   ##
=======================================
  Coverage   65.77%   65.77%           
=======================================
  Files         459      459           
  Lines       21911    21911           
  Branches     2410     2410           
=======================================
  Hits        14411    14411           
  Misses       7379     7379           
  Partials      121      121

Impacted Files	Coverage Δ
superset/models/helpers.py	85.54% <ø> (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 f278faa...c5f2b71. Read the comment docs.

[etr2460]

I'm not super familiar with this code, but I would've expected a change to the import function if the export format was changing. Does this still work when you try to import the new exported file?

[mistercrunch]

Actually I think the fix may be to not sort the list at all. I'm guessing the intention was to have a deterministic ordering so that aligns well in source control.

I think the ordering of the keys is stable/predictable given that it follow the export_fields ordering.

[Aylr]

@mistercrunch I guessed that the sorting was for the same thoughtful reason - to reduce VCS thrash, and that's why I though a hacky string coercion would be a suitable fix.

@etr2460 Since this affects only the ordering of the yml that is output by the function, there is no change required on the import code.

[Aylr]

What's the next step in the process to get this merged?

[mistercrunch]

Ok I resolved a conflict, let's move forward with this once CI passes

[mistercrunch]

Note that it fixes the bug but could also lead to random re-ordering when objects change. Maybe once #7472 lands, we can use the new uuid field as a deterministic way to sort lists of exported objects.

[smacker]

@mistercrunch we have identified more problems with export/import and now consider fixes that we would love to send upstream. The current proposed solution is also to use uuids. You can find the document here: https://docs.google.com/document/d/1vfiAw7yCzqvb3ALo_aiBX3XFPjLNZ1Mira-X7HCK1lI/edit?usp=sharing

[mistercrunch]

@smacker some work on uuid has been done here #7472 , but it should probably be refactored out into a smaller/simpler PR.

[mistercrunch]

Also related, found issues that I fixed here as bycatch #7773 in the export_dashboards method. Exporting dashboards would actually break the existing ones in the context of unit tests. Not sure if that was the same in production...

[Aylr]

Ping @mistercrunch what do we need to do to either get this merged or apply another fix? This 5 character PR has been open for 20 days now. What can I do to help?