Schema included in a data package not added to CKAN

[Stephen-Gates]

A data package can include schemas for each data resource. When I upload a datapackage.zip file containing schemas I expect these to be stored in CKAN.

If they were stored they could be

• made available for display
• provided as a tableschema.json resource
• included in the datapackage download (MIssing metadata fields when exporting data package #49)

[GeraldGrootRoessink]

Suppose we could import/export schema from/to datapackage.json to/from CKAN. Like that idea. I wonder though, how that relates to the field information provided bij de datastore-create-API and the datapusher-extension. For example, the test for the API provided in the CKAN documentation gives this result:

    "fields": [
        {
            "type": "int",
            "id": "_id"
        },
        {
            "type": "int4",
            "id": "a"
        },
        {
            "type": "text",
            "id": "b"
        }
    ],

Will this be compatible?

[Stephen-Gates]

@GeraldGrootRoessink this is related and worth a read ckan/ckanext-validation#1

[GeraldGrootRoessink]

Worth a read indeed. My understanding is that ckanext-validation introduces with ckanext-scheming a new field schema that accepts a json-object like (without the whitespace):
{
"fields" : [{
"name" : "regel",
"title" : "kolom 1",
"type" : "string",
"rdfType" : "http://lod.duo.nl/cdm/def/v0/regelnummer"
}, {
"name" : "brin",
"title" : "kolom 2",
"type" : "integer",
"rdfType" : "http://lod.duo.nl/cdm/def/v0/BRIN-V02"
}
]
}
So it should be possible to extract this part of het datapackage.json in
https://github.com/frictionlessdata/ckan-datapackage-tools/blob/master/ckan_datapackage_tools/converter.py at line 116.

Am I right?

[GeraldGrootRoessink]

Me again.
Added these lines after line 139

if resource.descriptor.get('schema'):
    resource_dict['schema'] = resource.descriptor['schema']

This works

[amercader]

@GeraldGrootRoessink That sounds like a good approach, can you submit a PR?

[GeraldGrootRoessink]

Submitted a PR. However the coverall/coverage checks fails at 96.465 because of a warning:

tests/test_converter.py::TestDataPackageToDatasetDict::test_datapackage_name_title_and_version
_/home/travis/build/frictionlessdata/ckan-datapackage-tools/.tox/py/lib/python2.7/site-packages/datapackage/package.py:420: UserWarning: Property "package.to_dict" is deprecated.
UserWarning)

I'm afraid I'm in over my head with this one.

[amercader]

@GeraldGrootRoessink this is coverage complaining that you didn't add a test for your change :)
Plus an unrelated warning which I fixed.

I did myself here, plus fixed the logic for the inverse path (dataset -> datapackage).

[GeraldGrootRoessink]

Super. For anyone who follows: It will only work nicely if there is a schema-field defined as a json-object. For example I managed this using https://github.com/ckan/ckanext-scheming.