fix: Avoid 500 if end users write bad SQL

[Khrol]

SUMMARY

While end users deal with Superset+Trino we are observing quite some 500 errors on backend.

For example, a user can create/edit a virtual dataset and make mistakes in SQL. Trino client raises the exception and it goes uncaught to 500.

If SupersetDBAPIProgrammingError is raised it's better to return 400 to the end user because nothing bad happens with Superset instance.

I don't know what to do with SupersetDBAPIError in general. If underlying Database has problems, should it lead to Superset's 500 or it's ok to stay with 400 because Superset itself works fine.

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

All modified and coverable lines are covered by tests ✅

Comparison is base (7ca6d8c) 69.18% compared to head (d7443c4) 69.06%.
Report is 5 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #26638      +/-   ##
==========================================
- Coverage   69.18%   69.06%   -0.13%     
==========================================
  Files        1948     1938      -10     
  Lines       76036    75596     -440     
  Branches     8478     8478              
==========================================
- Hits        52604    52208     -396     
+ Misses      21265    21221      -44     
  Partials     2167     2167              

Flag	Coverage Δ
hive	53.58% <12.50%> (-0.11%)	⬇️
mysql	77.88% <12.50%> (-0.16%)	⬇️
postgres	77.98% <12.50%> (-0.19%)	⬇️
presto	53.53% <12.50%> (-0.11%)	⬇️
python	82.77% <100.00%> (-0.09%)	⬇️
sqlite	77.57% <12.50%> (-0.20%)	⬇️
unit	55.85% <100.00%> (-0.04%)	⬇️

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

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[villebro]

We have lots of these unnecessary 500s, which makes logs noisy, thanks for cleaning these up 👍 While the new Trino exception mappings are probably very uncontroversial, I'd like a few extra eyes on changing the status type of SupersetDBAPIProgrammingError. However, I feel 400 is probably a better default than 500 for that one.

[Khrol]

While the new Trino exception mappings are probably very uncontroversial

I've initially submitted Draft PR to see at least some feedback. The situation there is a bit more complex and I've pushed more code.

If underlying Database has problems, should it lead to Superset's 500 or it's ok to stay with 400 because Superset itself works fine?

I don't know how to interpret 500 errors with connection to underlying source Databases... 🤔