Fix panic on invalid number
I added a small check for a case that would previous panic. This was found through fuzzing. I've found another issue where it is possible to overflow the stack using a query with a long series of ['s. The change for this might be intrusive though. Before I fix it I wanted to send this PR up and check to see if you are interested in the more intrusive change.
Excellent work. I would be open to including some fuzzing infrastructure in this repo if you feel like raising a PR.
For the more "intrusive" change, feel free to open a PR and we can work through it. Definitely keen to prevent panics & stack overflows.
I'll open a separate PR when time allows for the others. Where can I view the results of a travisci run? I can seem to find that and it will be useful when doing the fuzzing PR.
I was wondering if we could get this merged and a new version released. We hit this again today.