Open
Description
There are numerous instances of panic in the codebase, which are not ideal for production use cases. This often results in crashes of the processes on our production machines because we accept dynamic documents from users, which might contain malformed data. I believe that returning errors instead of panicking would be a better approach.
For instance, the following line has been causing panics in our production instance:
Line 239 in 567f83d
We cannot always verify whether the incoming data from the client is correct.
I'm ready to work on this if you give me some hints about your preferred approach.