Closed moeyensj closed 2 years ago
This breaks the data file format; old data files will be invalid in strange ways (probably breaking because string IDs will seem completely invalid). Now is not the time for this, but at some point it will be worthwhile to store a bit of metadata which identifies the version being used so we can detect this. Maybe that's a new metada.json file, or maybe it belongs inside index.db.
I'll file an issue so we remember to add this in the future. Tracking the version at indexing time is definitely something we are going to need to need as well when we index multiple datasets at the different times they may come available to us.
There really should be tests that cover this behavior. I think this test at least will break with this change, but more tests would be good.
Couldn't agree more. I haven't touched any unit tests but we definitely need to update them (some of the tests are using objects and modules that no longer exist). This may be a good sprint or hack-session goal. I'll merge this PR and make a note to create a new unit test PR. The good thing is that it shouldn't be too difficult to get the coverage to 100% while the code is still in this stage of the development.
This PR adds 1-sigma uncertainties to the indexed observations. I've tested this with the latest version of the observation HDF files.
Closes #9.