lgbo-ustc opened a new issue, #13607:
URL: https://github.com/apache/gluten/issues/13607

   ## Problem
   
   The ClickHouse backend executes Spark/Java regular-expression functions 
using RE2, but the pattern translation added for `regexp_extract` is local to 
that parser. `regexp_extract_all`, `rlike`, `regexp_replace`, and regex `split` 
can interpret the same pattern differently. The existing bracket flattening is 
heuristic and does not generally preserve Java semantics for intersections, 
scoped flags, or other unsupported constructs.
   
   ## Proposed improvement
   
   * Introduce a shared Spark/Java-to-RE2 pattern translator in the CH backend, 
used by the regex functions above for literal patterns.
   * Keep replacement-string conversion separate from pattern translation.
   * Translate supported constructs while preserving capture-group numbering 
and DOTALL scope; explicitly reject or fall back for unsupported cases instead 
of silently changing semantics.
   * Add unit coverage for character classes, escaped brackets, anchors, 
DOTALL, and capture groups, plus Spark/CH comparison tests.
   
   ## Scope and follow-ups
   
   Some Java patterns cannot be faithfully implemented by RE2 (e.g. lookaround 
and backreferences). Other constructs need additional lowering (e.g. 
intersections and multiline-anchor behavior). The initial implementation should 
document and test its supported subset, and ensure unsupported cases fall back 
to Spark. CH native build/CI is needed to verify integration.
   


-- 
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