Catching exceptions
leebenson opened this issue · 1 comments
Using Graphene with graphql-core 2.0, I was able to write middleware that would catch errors:
class ErrorHandlerMiddleware():
def resolve(self, next, root, info, **args):
try:
return next(root, info, **args)
except Exception as ex:
# log error with Sentry
info.context.sentry.captureException()
# Return a generic rejection
return GraphQLError("System error. Please try again later.")
It wasn't perfect. Both the original exception and GraphQLError would wind up appearing as Sentry issues, but it'd at least get reported and send back a generic error to the user.
I've since bumped graphql-core to 2.1 to take advantage of the logging improvements, but I'm not clear on how the above could be achieved in 2.1? With the latest version, the original exception winds up in the message
field of the response... potentially exposing sensitive info, like SQL errors, etc.
Could you please post an example in the README of how one might write middleware or use the new logging set-up to capture errors and replace the response with a generic message? I think that would be a really common use-case that many would find useful.
FWIW, this is what I wound up doing. Can anyone see an issue with this approach?
def error_handler(ex):
print("original exception", ex) # replace this with third-party logging
raise Exception("System error") # some error message to show to users
class ErrorMiddleware:
def resolve(next, root, info, **args):
res = next(root, info, **args)
return res.then(None, error_handler)
With the print line in the error handler being replaced with Sentry logging, and the new exception being raised to mask the original (hiding sensitive details like SQL queries, etc)
It still logs out graphql.error.located_error.GraphQLLocatedError: System error
to stderr but setting the main logger to silence those should take care of it.
I'll leave this open to get consensus on this approach.