thisisnic commented on pull request #11534:
URL: https://github.com/apache/arrow/pull/11534#issuecomment-963937359


   > > Maybe worth noting expect_snapshot() works best with pure R (or 
R-captured errors). If the error is surfaced from C++ the snapshot ends up 
including a bunch of information specific to the user's machine.
   > 
   > What does this look like? Is there anyway we can construct the helper to 
avoid some of that? It would be unfortunate (though not at all the end of the 
world!) if we had to avoid these / have two patterns depending on where the 
error comes from
   
   How many examples of these are there? I know it's a bit controversial but 
I'd be in favour of adopting the practice of always adding our own messages (as 
well as the C++ error output) and just testing that bit.
   
   I think there is some objection to this on the basis of the idea that if we 
don't think the C++ error messages are informative enough, we should suggest 
updates to those, but I think that in most cases there is information which 
could be added to the error for the R user which doesn't make as much sense to 
the C++ user.
   
   I'm thinking of things like the statements containing the word "must" and 
suggestions for what the user could try doing.
   
   This might not fit every error though - but am unsure without thinking about 
concrete examples.
   


-- 
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