Repeating Subfields and Multiple Text with ckanext-scheming

  • Ian Ward
  • 20 Apr 2021
ckanext-scheming 2.1 now support Datasets with repeating subfields and repeating text fields. Repeating subfieds support custom snippets and validation just like regular fields. This work is inspired by the excellent ckanext-composite extension and replaces ckanext-repeating repeating_text fields. You must be using CKAN 2.8 or later and a custom IPackageController plugin to index datasets with repeating subfields.

Repeating Subfields

Repeating subfields let you define a group of fields to repeat within dataset and resource forms. All of the features available for normal ckanext-scheming fields can be used in subfields. Repeating Subfields repeating_label may be used to provide a singular label for each group of subfields. repeating_subfields contains a list of field definitions to repeat.
- field_name: submission
  label: Submissions
  repeating_label: Submission

  - field_name: date
    label: Date
    preset: date
    required: true

  - field_name: text
    label: Text
    preset: markdown
    required: true

  - field_name: flags
    label: Flags
    preset: multiple_checkbox
    - label: Draft
      value: D
    - label: Approved
      value: A
    - label: Response Required
      value: R
Data stored in subfields is represented as lists of JSON objects in the API.
"submission": [
    "date": "2021-02-01",
    "text": "an example submission",
    "flags": [
    "date": "2021-02-05",
    "text": "another one",
    "flags": [
Normal and custom validation rules apply and are displayed in the form by referencing the subfield group and the field with the error, e.g. “Submission 2: Text: Missing value”

Indexing Repeating Subfields

Repeating subfields can’t be indexed in Solr with CKAN’s default schema and indexing code since they expect extra fields to contain simple string values. Create a new IPackageController before_index plugin and Solr schema to handle indexing repeating subfields the best way for your own site.
class SubmissionsIndexPlugin(p.SingletonPlugin):
    Map submission dataset fields to Solr fields
    p.implements(p.IPackageController, inherit=True)

    def before_index(self, data_dict):
        flags = set()
        text = []
        for sub in data_dict.get('submission', []):
            flags |= set(sub['flags'])

        # replace list of dicts with plain text to prevent Solr errors
        data_dict['submission'] = '\n'.join(text)
        # index flags present in any submission
        data_dict['submission_flags'] = sorted(flags)

        return data_dict
For submission_flags to accept multiple values we must add a multivalued field to our Solr schema <fields> configuration:
<field name="submission_flags" type="string" indexed="true" stored="true" multiValued="true"/>
These new fields will now be available for use with CKAN advanced search e.g. submission:example or submission_flags:D. If you don’t need advanced search or faceting based on repeating subfields you may use the included scheming_nerf_index plugin. This plugin passes repeating fields to Solr as JSON strings to prevent indexing errors instead and doesn’t require a customized Solr schema. Future CKAN support for dynamic fields in Solr will simplify this required configuration.

Future Work

Some features not yet supported:
  • validating the number of subfield groups (e.g. requiring at least one)
  • nesting repeating subfields
If you need these features consider discussing your work on a new issue and working a pull request.

Multiple Text

Multiple text fields improve on the repeating_text fields from ckanext-repeating with a dynamic form with add and remove buttons. Multiple Text
- field_name: contributors
  label: Contributors
  preset: multiple_text
Data stored in repeating text fields is represented as lists of strings in the API.
"contributors": [
  "Person A",
  "Person B",
  "Person C"
required: true may be used to require at least one entry. Per-field and other types of validation are not yet implemented. Add a comment to the multiple text validation issue if you would like to work on this feature. original post