nealrichardson opened a new issue, #43664:
URL: https://github.com/apache/arrow/issues/43664

   ### Describe the enhancement requested
   
   See https://github.com/apache/arrow/issues/43627. We had two queries that 
were actually different, but the difference was in the SourceNode (dataset 
ScanNode?), particularly in the predicate pushdown. Unfortunately, both queries 
just printed `0:SourceNode{}` for the node, so they appeared to be the same. 
This led us to ship a significant performance without thinking it was real, and 
then made it hard to debug when I dug in more.
   
   Ideally, `SourceNode` would print what kind of source node it is (there are 
many subclasses), and for a dataset, something about the filter and projection 
(potentially just number of columns, if we're worried that it could be too 
big). I have less strong feelings about what other things should be printed for 
other subclasses, and whether the dataset scan node should print more (file 
format etc.), but something like this about the predicate pushdown would be 
important for showing what work the node is going to do.
   
   Looking through the source, I see that many *Nodes have a `ToStringExtra` 
method defined, so that seems to be the way.
   
   ### Component(s)
   
   C++


-- 
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]

Reply via email to