Create default artificial test objects/datasets

[leewujung]

Increasingly we have the need to unit test functionalities that are higher on the processing chain and many of the current tests right now need to go through the convert-->calibrate process before the actual functionality can be tested. This has a few drawbacks:

• hard to keep track of which test data files are used in what tests
• "real" data files do not allow the flexibility of testing what may be missing or in peculiar forms in the parsed data, corresponding EchoData object, and calibrated Sv
• data files tend to be large, especially those containing complex samples from EK80

Proposal

We can create artificial test objects and datasets the same way we have a default test path. A few thoughts on what we may want to have for these

• can create irregularity according to what we are aware of in the parsed data (parser objects): use these in tests for set_group_X
  • examples include and are not limited to: irregular range_sample across channels or across ping_time, missing water_depth, missing lat/lon info
• can create irregularity in EchoData object: use these in tests for the subpackages calibrate and consolidate
• can create irregularity in calibrated Sv datasets: use these in tests for the subpackages for higher levels

I also propose that we put together a couple hackday for this since our tests are pretty adhoc at the moment and would benefit from a group discussion with a top-down approach. Once that's setup, we can identify specific tasks, and then divide the tasks to each tackle a subset.

[leewujung]

Get together ~Sept 1 Thu~ Sept 8 Thursday: focused on EK data for now:

• @lsetiawan : hackday chief, pytest fixture setup/modules
• @b-reyes : parsed output modules/functions
• @leewujung : irregularity in files