Rich-T-kid opened a new issue, #25757:
URL: https://github.com/apache/datafusion/issues/25757

   ## Summary
   
   When debugging the single-node → distributed planner pipeline in downstream 
projects like 
[datafusion-distributed](https://github.com/apache/datafusion-distributed), it 
is useful to inspect the original physical plan before any transformations are 
applied. Currently there is no built-in way to observe or log the plan at that 
stage without adding ad-hoc instrumentation downstream.
   
   ## Motivation
   
   In datafusion-distributed, adding a local log of the pre-transformation plan 
was proposed as a way to capture reproducers and understand what the planner 
receives. However, logging the full plan on every query may be expensive and is 
arguably not the right place to address this — the concern is really about 
observability of DataFusion's own plan-rewriting pipeline.
   
   Addressing this upstream would let any project that wraps DataFusion's 
physical planning observe the plan at well-defined checkpoints without 
duplicating logging logic or paying the cost unconditionally.
   
   ## Proposal
   
   Consider one of the following approaches:
   
   - **A structured tracing span** (e.g. using the `tracing` crate) that 
includes the serialized or display-formatted plan at the entry point of 
physical plan optimization/rewriting, gated behind a low-overhead level (e.g. 
`TRACE`).
   - **An observer / hook API** on `SessionContext` or `QueryPlanner` that 
downstream implementors can register to receive the plan before and/or after 
each rewrite pass.
   - **A `EXPLAIN`-style plan-snapshot facility** that can be enabled 
per-session for debugging without affecting production workloads.
   
   Any approach should ensure the logging/inspection is opt-in and has 
negligible overhead when disabled.
   
   ## Related
   
   - Raised during review of datafusion-distributed (internal discussion on 
pre-transformation plan logging for reproducer capture).


-- 
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