cli: add filter option for the get-metadata command

[joudmas]

cli: add --filter option in get-metadata command
modified get-metadata to support multiple fields

[codecov-commenter]

Codecov Report

Merging #130 (a95f805) into master (eead01e) will decrease coverage by 0.14%.
The diff coverage is 80.48%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #130      +/-   ##
==========================================
- Coverage   79.84%   79.71%   -0.14%     
==========================================
  Files          11       11              
  Lines         665      700      +35     
==========================================
+ Hits          531      558      +27     
- Misses        134      142       +8     

Files	Coverage Δ
cernopendata_client/cli.py	91.37% <80.48%> (-3.08%)	⬇️

[tiborsimko]

The "workflow" here is a remainder from the REANA client?

The help should rather say here something like:

• Filter only certain output values matching filtering criteria. Use --filter somefield=somevalue.

[tiborsimko]

You can also add a full example to the Examples section of the docstring:

$ cernopendata-client get-metadata --recid 329 --output-value authors.orcid --filter name="Rousseau, David"

[tiborsimko]

Moreover, you can enrich the user guide usage.rst the section "Getting metadata" so that the David Rousseau example will also appear here: https://cernopendata-client.readthedocs.io/en/latest/usage.html#getting-metadata

[tiborsimko]

You may want to quote "Rousseau, David" as an argument.

BTW this works nicely:

$ cernopendata-client get-metadata --recid 329 --output-value authors.orcid --filter authors.name="Rousseau, David"
0000-0001-7613-8063
$ cernopendata-client get-metadata --recid 329 --output-value authors.orcid --filter name="Rousseau, David"
0000-0001-7613-8063

But the following ones do not fully work:

$ cernopendata-client get-metadata --recid 329 --output-value authors --filter name="Rousseau, David"

(All output values are printed, even though user might expect only the values matching the filtering criteria.)

Can you please have a look?

[tiborsimko]

This works well with the PR update 👍 You may want to also add a test case for this, i.e. get all authors, then with the filter, and then compare the two lists whether the second one has one less item as expected

[tiborsimko]

I guess we can do with multiple=False for now, but in the future we may want to extend it to match more than one. E.g. in our canonical example:

$ cernopendata-client get-metadata --recid 329 --output-value authors.orcid --filter name="Rousseau, David"
0000-0001-7613-8063

There could be more people with the same name, so one would like to write:

$ cernopendata-client get-metadata --recid 329 --output-value authors.orcid --filter name="Smith, John" --filter affiliation="Harvard"

to get John Smith from Harvard, an not from Cornell. (Just an example.)

If it it easy to implement, you can have a look at it. Otherwise we can open an issue for later.

[joudmas]

I will create a new issue for this and link it back to this discussion for reference.

[tiborsimko]

You can rather say:

• If the output field produces a list of values, you may want to filter only certain field values of interest. For example, you may want to ask not to output all authors of a dataset record, but only the authors matching certain particular author name. You can use the --filter command-line option to achieve this:

[tiborsimko]

This one comes "in the middle" of the ORCID example. I think we have two options:

1. First show example getting all authors, and filtering on one. Then show another example how to get only ORCID identifiers.
2. Alternatively, we don't show full author JSON output, only the ORCID filtering.

[tiborsimko]

If the first argument is named output_field, then we should also update the docstring of the function, see the lines:

:param field:

...

:type field:

[tiborsimko]

field_name=value could be misunderstood: is value supposed to be the field name? Or the value of the field? Therefore it would be better to keep somefield=somevalue speaking abstractly, or fieldname=fieldvalue or somefieldname=somefieldvalue or some_field_name=some_field_value speaking more precisely.

This in all places:

$ rg 'field_name='
• tests/test_cli_get_metadata.py
144:    assert "Invalid filter format. Use --filter field_name=value" in test_result.output

• cernopendata_client/cli.py
76:    :param filters: Argument of the --filter option in the format field_name=value
89:            msg="Invalid filter format. Use --filter field_name=value",
145:    help="Filter only certain output values matching filtering criteria. [Use --filter field_name=value]",

[tiborsimko]

Thanks, works nicely 👍

We can merge this initial version, and please create that additional issue for multiple filters.