rusackas commented on PR #39925:
URL: https://github.com/apache/superset/pull/39925#issuecomment-4724436006

   Bot findings of interest:
   
   Whining that it's too big:
   >  If any one piece regresses in production, you can't revert without losing 
all of it. At minimum, peel off: (a) the i18n catalog churn — it has no 
business in a URL refactor and will conflict-thrash; (b) the navigationUtils 
helper + call-site migration (the genuinely valuable, low-risk core); (c) the 
backend middleware/PWA/redirect pieces; (d) the response-normalizer.
   
   ^ I might agree with the translations, if that even CAN be taken out 
cleanly. If not, whatever.
   
   > Dangling references to PLAN.md and internal audit scaffolding in committed 
code. Comments reference PLAN.md (3×, and the file is not in the PR), plus ~125 
references to an internal methodology — AF-1 (26×), AF-2, RF-2, Slice 7, 
review-fix, C5 in PR, dated 2026-06-01 tags. Future maintainers can't decode 
any of it. Keep the actual rationale in the comments; strip the 
slice/audit-ID/PLAN.md vocabulary. This is a cleanliness blocker for a 
permanent codebase.
   
   Other random things that sound problematic, perhaps:
   
   >  The response normalizer is heavy machinery for one field. 
NORMALIZED_URL_FIELDS contains exactly explore_url. To strip that one field it 
deep-walks every JSON response (every chart-data payload, every large 
json_metadata) under subdir deployments, with a depth-100 cap and WeakSet cycle 
detection that is dead code for freshly-JSON.parsed data (parsed JSON can't 
contain cycles or shared refs). The shared-reference branch is also subtly 
inconsistent (first occurrence normalized, second returned raw) — harmless only 
because parsed JSON has no shared refs. For one field, a targeted strip at the 
few known call sites would be dramatically cheaper and simpler. I'd push back 
on landing the generic walker.
   
   > SupersetClient.getUrl semantic change is a public-API contract change. 
Third-party plugins consume this. Auto-deduping a leading appRoot is a 
reasonable safety net, but it changes published behavior and can mask genuine 
/superset/superset/... routes. Needs a deliberate call-out in UPDATING.md (the 
permalink flag is documented there; this isn't clearly).
   
   LMK what you think... we can take it from there.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to