I noticed that in the default skeleton file, there is a commented out
section that should be uncommented if the function throws errors. I
understand that the more documentation, the better, but in this case,
it's quite difficult to determine exactly what errors the function can
throw, because they are bubbled up from the underlying client library.
None of the errors are explicitly defined in the code-base.
What is the policy for handling these errors?
No policy yet but if we cannot decide if it emits errors then
document the function but do not include the error role. Not every
function creates E_* and I'm unsure what percentage do but this
should not hold up other aspects of the documentation. Adding the
role can be dealt with later.
Related possible TODOs:
- HOWTO figure out when/if a function returns E_*
- Script that checks what E_* level errors a function
may emit. It might be difficult to make this script
perfect but it could do a decent job and give us a
start.
Regards,
Philip
