Change output of open_raw to an EchoData object

[leewujung]

This PR:

• changes output of open_raw from a Convert object to an EchoData object
• implements the behavior for open_raw to parse the raw data files and organize the parsed data into attributes of the output EchoData object. This requires user to use EchoData.to_zarr/to_netcdf to save the parsed data to file. This addresses separate nc4/xarray dataset group creation from file saving functionality #228 and is related to save vs set in class redesign #225.

Note:

• open_raw only works with parsing a single raw data file.
• open_raw and open_converted are now under echodata/open_file.py. I was debating if open_raw should remain in the convert module but eventually moved it to under echodata. Let me know if you think otherwise.

Todo for later PRs:

• still need to implement open_mfraw to parse multiple raw files and combine the parsed data into attributes of a single EchoData object.
• we should remove the Convert class since its function as an interface for using ParseXX and SetGroupsXX objects can now be substituted with Echodata

Updates

• Convert is now deprecated with all functionalities provided directly by Parse and SetGroup objects
• Next steps are to:
  • take care of specific functions under echopype/convert/api.py, such as _validate_path and to_file, to just provide with 1-to-1 (input/output) mapping
  • Other singular changes in Update open_raw leewujung/echopype#3

[leewujung]

@lsetiawan : Could you take a look at the use of storage_options and output_storage_options in EchoData this PR? I am pretty sure when trying to transplant _validate_path and associated functions I broke your test_convert, because the output of open_raw is now an EchoData object instead of a Convert object. Thanks!

[lsetiawan]

Looks like you integrated PR #281 into here is that correct?

[lsetiawan]

Hmm... looks like you integrated a lot of the stuff from Convert Object ... Not sure if this is the best way to go because I can see EchoData ending up being convoluted ... I think it might be better if you keep all of these functions in the Convert Object and use that within EchoData.

[leewujung]

Yeah, those transplanted functions are my biggest hesitation, as in whether to keep them in Convert or move them to EchoData. (I was thinking about getting rid of the Convert object!) But looking at this again maybe it'll be better that way.

[lsetiawan]

I see. I think it's okay to keep the Convert object. Since we know it already works and this would split up the functionalities, and we can just test individually.

[leewujung]

Oh now I remember why I did that: the main thing is that now the set groups functions return xr.Dataset and are linked directly to attributes of EchoData in open_raw. That's why I felt the Convert class can actually be removed and substituted by the same functionalities in EchoData. This does make EchoData more convoluted though I agree. Hmm... thoughts?

[lsetiawan]

For tests, it's better if you don't do relative paths since echodata should be installed before you run pytest. So you can just do from echodata import ...

[leewujung]

I see, cool, thanks for explaining this!

[leewujung]

Looks like you integrated PR #281 into here is that correct?

Yes, wanted to make sure I got your changes in before starting this.

[leewujung]

Noting conversations in leewujung#3 for more changes on this PR.

[leewujung]

@lsetiawan thanks for the changes! I just put updates the PR message to include next steps and will merge this now.