Fixed reraising of plotting exceptions

[philippjfr]

This has been plaguing me for a long time. In python2 I always get these exceptions when there's an issue with the options. I think this is due to the nested try/excepts. Reraising an exception in this way differs slightly between py2 and py3 hence extra conditional.

TypeError: exceptions must be old-style classes or derived from BaseException, not NoneType

[philippjfr]

Appears there's some persistent py2 failure now, will have to look into it.

[jlstevens]

I've always found reports of this error very mysterious - not only does the error message make no sense (all our classes derive from object, we don't have any old style classes!) but I've never been able to reproduce it myself.

Using raise to re-raise an exception is official, documented Python behavior so I don't know why it would cause issues. I am happy to merge a fix as long as the right traceback is displayed - ideally, it would just be a re-raised exception but that is what it is doing now and for some reason it has issues!

[philippjfr]

Using raise to re-raise an exception is official, documented Python behavior so I don't know why it would cause issues

It is if you're simply reraising, if you have another try/except in the same block you have to do what I did in the PR.

[jlstevens]

I am happy to merge if you have checked the correct traceback is shown (which was a previous issue with this code that was fixed) and once the tests pass.

[jbednar]

This problem has also bedeviled me, and I'm very glad there is a fix!

[philippjfr]

I've tested it on py2 and py3. The only thing that would be nice is to maybe shorten the traceback a bit, or rather show just the exception message for ValueErrors coming from param validation.

[jlstevens]

The only thing that would be nice is to maybe shorten the traceback a bit, or rather show just the exception message for ValueErrors coming from param validation.

That can be implemented in a separate PR.Merging.