LuciferYang commented on PR #58045:
URL: https://github.com/apache/spark/pull/58045#issuecomment-5599486418

   Thanks for the approval. All four are right; fixed in 1d9ff29.
   
   **Acyclic sibling references.** Agreed, and my comment was wrong rather than 
merely vague: `withCommonExprs` registers every sibling slot before generating 
the child, so `d2 = Add(ref(d1), 1)` resolves and completes. What codegen 
refuses is the two id-keyed shapes plus a definition that reaches its own id, 
around a cycle or directly, through `filling`. `refsToBind` now says that, and 
`doGenCode`'s reason for no body enclosing another is restated accordingly: a 
definition reading a sibling is generated inside that sibling's guard, not 
inside its own.
   
   **Self-reference test.** Split. With distinct reference objects the 
interpreted failure was an unbound reference, since the scan over `child` binds 
only the object in `child`; the shared-object variant below is the one that 
re-entered a cell and overflowed.
   
   **Malformed sentence.** Rewritten. It was mangled by the comment trim in 
1133af6.
   
   **Tree-pattern bit.** Good catch, and cleaner: 
`containsPattern(WITH_EXPRESSION)` is the same root-or-descendant test in a bit 
check. Also re-ran `ColumnExpressionSuite` and `WholeStageCodegenSuite` for 
this one, since it decides `worthAMethod`.
   
   On the red `Build` check at the previous head: that is the 
`pull_request_target` waiter timing out, not a test failure -- it concluded at 
16:48Z while the fork's `sql - other tests`, `sql - extended tests` and Docker 
integration shards were still running, and the fork run had zero failed jobs.
   


-- 
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