No traceback for ValueErrors or TypeErrors when importing Things

[julianstirling]

Describe the bug
As of LabThings 0.0.12. Parsing the server config and importing Things is done within a Pydantic model. However Pydantic only expects it is serialising/deserialising data and hence captures every ValueError and TypeError and raises a ValidationError without the traceback of the original error.

This means that if there is ever a TypeError or ValueError when importing a Thing the traceback is removed and an unhelpful Pydantic error is shown instead,

To Reproduce

• Create a Value error in a Thing file that will be called at import time
• Run labthings with this thing in the configuration

Expected behaviour
A traceback that points at the ValueError itself.

Suggested fix

While it could be possible to catch the ValueErrors or TypeErrors and raise them as a different type that Pydantic will not capture, however this is fragile to pydantic capturing changing overtime and capturing different things. It would be better to parse the configuration and serialise it with the strings. And then import the modules outside pydantic validation.

System:

• OS: Kubuntu 24.01
• Python version: 3.11
• Version: 0.0.12

[rwb27]

Having LabThings die immediately if a module fails to import may or may not be desirable, but it's hard to do without rolling back to my old import code. That would make the server config model rather more complicated: in particular, it would make it very hard to support either Python classes or import strings (which makes test and example code differ from production code).

I think a single error that makes it clear where the problem is, should also be a reasonable resolution here, and I've implemented that in #226 .

[julianstirling]

This was closed by #226 and #227