Here is a first draft of the petab result format. There are some open questions to it:

• Which of the three versions should be keep? Personally I think all three have merit: the python description was used as a base and allows for changes later on. The json-schema is the actual resulting data schema and the yaml version is a much better readable version of the json schema.
• Where should this deposited?
• Should we add a description to the readthedocs in the documentation_data_format.rst file or make a separate file?
• Should be allow for each datatype to be extended arbitarily or require tools that want to save more open an issue on it?
• Certain things right now are non optional, there were discussion on whether to make anything non-optional. Opinion and at some point a vote are more then welcome.

Any feedback is welcome. As the PR was done mainly by pyPESTO users, we tried to extrapolate the essential things, but are aware that this might be biased. Therefore feedback from non-pypesto users is especially welcome.

The idea for the structure for the result follows this hierarchy:

• Result
  • Metadata like system, programming language and tool with versions
  • Problem snapshot: A snapshot of the petab problem used for any result stored. Implicitly this also means that results that (e.g. differ in the fixed parameters) should be saved separately
  • TaskResult: tasks like profiling, optimization, sampling. Is a list of singular results, each of which saves relevant data (e.g. optimization: start point, endpoint, fval) and additional metadata.