Exception handling: How to handle errors when parsing data

What should we do with errors while parsing data? Report them to the user? Throw them to stdout as warning?

Examples:

• empty observation exception (empty nonsense columns resulting from quirky csv headers)
  • => skip and log to things journal
• Unique constraint on timestamp/datastream violated
  • => skip and log to things journal
• EmptyDataError when reading csv data with pandas. Currently we continue with an empty data frame.
  • => skip and log to things journal
• too unspecific error: "XyzParser" not (completely) implemented yet? We should consider more specific exceptions, especially when handling input data. #13 (closed)