Custom upload endpoints

[amercader]

High-level spec: #65 (comment)

Update the ckanext-validation blueprint to include:

1. An endpoint that accepts an uploaded file or a URL and creates a resource with a file (POST /dataset/<dataset_id>/resource/file). This will be used when in step 1 of creating a new resource (uploading the file). This endpoint will:

   • Do some data processing and call resource_create with the package_id and the upload / url fields
   • Decide if this resource is a tabular one, and if so detect the schema calling resource_table_schema_infer (see Add a new resource_table_schema_infer action  #76)
   • Return the dataset id, the resource id and the inferred table schema description if present

2. An endpoint that accepts an uploaded file or a URL and updates the file of resource (POST /dataset/<dataset_id>/resource/<resource_id>/file). This will be used when updating an existing resource. This endpoint will:

   • Do some data processing and call resource_patch with the resource_id and the upload / url fields
   • Decide if this resource is a tabular one, and if so detect the schema calling resource_table_schema_infer (see Add a new resource_table_schema_infer action  #76)
   • Return the dataset id, the resource id and the inferred table schema description if present

Additional endpoints:

• Do we need a custom endpoint for step 2 of creating a new resource (clicking "Save")? The standard resource create endpoint (POST /dataset/<dataset_id>/new) seems to support updating an existing resource if a resource id is provided so that might work out of the box. We just need to make sure client-side that the resource_id is set after getting the response from step 1 (and that the file is not submitted again). That would cover the "Finish" and "Save and Add another" buttons
• What happens if a user hits "Previous" or leaves the page after uploading a file? CKAN does not have a concept of "draft resources" so I think we just need to delete the resource:
  • For the "Previous" rather than overriding the whole core /dataset/resource/new endpoint, I think it will be easier to have a custom endpoint (POST /dataset/<dataset_id>/resource/cancel) that accepts a resource_id (which we'll have if we have uploaded a file) and deletes the resource and the file. If there is no resource_id we just submit the default endpoint.
  • If the user leaves the page without having clicked on "Finish" we can maybe rely on beforeunload to fire a request to the previous endpoint?