Value capture

[tzanko-matev]

• We snapshot all variables at each step of the script
• We don't step through internal Python code which is not associated with a real Python file, thus reducing significantly the size of the trace.

[Ro6afF]

It seems that failing to trace a frame/value is treated like a critical error that causes the execution of the traced program to fail. I think that this is not the desired behavior.

Logging the error and continuing seems like a better option to me.
Or alternatively recording some kind of error value in the trace is a good approach too.

[tzanko-matev]

I don't know what is the right behavior. During development I want the code to fail whenever something unexpected happens. This is the fastest way to notice that there is a bug. I try to split error conditions in two types:

• Getting to an invalid state due to wrong input or environment - those should be handled properly and a good error message should be sent to the user
• Getting to an unexpected state due to an error or omission in the program logic - During development and testing those should crash as close to the source of the bug as possible in order to make the bugs visible. Can we ignore them during production deployment?

[tzanko-matev]

I'm merging this. WIll implement better error handling in a separate PR