fix(sqla): copy temporal range logic to helper

[villebro]

SUMMARY

When exploring a query containing a temporal column I noticed that the query would fail unless persisted as a virtual dataset when the GENERIC_CHART_AXES feature flag is enabled. Digging around I noticed that the ExploreMixin class in the superset.models.helpers module (introduced in #20281) duplicates essentially all logic from the
SqlaTable class in the superset.connectors.sqla.models module. From a maintenance perspective this is highly problematic, as the SQLA connector model is by far one of the biggest pieces of tech debt in the backend with poor test coverage, while simultaneously being very fragile and frequently receiving minor tweaks/fixes. Duplicating this code without having tests in place to ensure that they're in sync makes it difficult for contributors to know that changes need to be applied in both modules.

This is just a hotfix to make the ExploreMixin class from the helpers module work with GENERIC_CHART_AXES. However, IMO we need to deduplicate this code ASAP, as they've already diverged quite a bit, with many fixes to the SQLA model not having made their way into the helpers module (#22280 is an example of such). I haven't reviewed the original PR, but a simple solution that comes to mind is removing the original logic and adding ExploreMixin to SqlaTable, which ensures that the legacy model (which is still the main one in use) uses the exact same logic as the new models.

AFTER

Now a query that hasn't been saved as a virtual dataset works with temporal filters:

BEFORE

Previously the same chart would fail.

BEFORE/AFTER SCREENSHOTS OR ANIMATED GIF

TESTING INSTRUCTIONS

ADDITIONAL INFORMATION

• Has associated issue:
• Required feature flags:
• Changes UI
• Includes DB Migration (follow approval process in SIP-59)
  • Migration is atomic, supports rollback & is backwards-compatible
  • Confirm DB migration upgrade and downgrade tested
  • Runtime estimates and downtime expectations provided
• Introduces new feature or API
• Removes existing feature or API

[codecov]

Codecov Report

Merging #22405 (afa119b) into master (a235078) will decrease coverage by 0.00%.
The diff coverage is 25.00%.

@@            Coverage Diff             @@
##           master   #22405      +/-   ##
==========================================
- Coverage   66.87%   66.86%   -0.01%     
==========================================
  Files        1847     1847              
  Lines       70574    70578       +4     
  Branches     7748     7748              
==========================================
- Hits        47195    47192       -3     
- Misses      21378    21385       +7     
  Partials     2001     2001              

Flag	Coverage Δ
hive	52.52% <25.00%> (-0.01%)	⬇️
mysql	77.97% <25.00%> (-0.01%)	⬇️
postgres	78.02% <25.00%> (-0.02%)	⬇️
presto	52.42% <25.00%> (-0.01%)	⬇️
python	81.22% <25.00%> (-0.02%)	⬇️
sqlite	76.50% <25.00%> (-0.01%)	⬇️
unit	50.91% <25.00%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
superset/models/helpers.py	38.13% <25.00%> (-0.06%)	⬇️
superset/reports/commands/log_prune.py	85.71% <0.00%> (-3.58%)	⬇️
superset/commands/importers/v1/utils.py	92.20% <0.00%> (-1.30%)	⬇️
superset/common/query_object.py	93.84% <0.00%> (-0.52%)	⬇️
superset/reports/commands/execute.py	91.55% <0.00%> (-0.34%)	⬇️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[zhaoyongjie]

ExploreMixin has made the already confusing superset.connectors even more difficult to maintain. I agree with this PR change and hope the community can reduce the superset.connectors complexity.

[villebro]

Tagging @hughhhh and @AAfghahi for a review as you authored and approved #20281 to see what your thoughts are on DRYing up ExploreMixin and SqlaTable.

[eschutho]

HI @villebro. Thanks for flagging this. Yes, @hughhhh will be working on removing the code duplication in the next few weeks. The original plan was to incorporate SIP68 work here and remove the sqlaTable and migrate the tables and columns over to the new SIP68 models and at the same time use the new mixin, but we decided not to allow charts to be created by tables and columns and the new dataset models and now we need to dig right into sqlaTable instead. But the plan moving forward is to use the exploreMixin and clean up the sqlaTable code during the process of moving the code over.

[villebro]

HI @villebro. Thanks for flagging this. Yes, @hughhhh will be working on removing the code duplication in the next few weeks. The original plan was to incorporate SIP68 work here and remove the sqlaTable and migrate the tables and columns over to the new SIP68 models and at the same time use the new mixin, but we decided not to allow charts to be created by tables and columns and the new dataset models and now we need to dig right into sqlaTable instead. But the plan moving forward is to use the exploreMixin and clean up the sqlaTable code during the process of moving the code over.

Thanks @eschutho for the context. Please feel free to tag me in the cleanup PRs to help push this cleanup effort forward.