CKAN中author字段分面显示分词值问题及相关技术问询

Answers to your CKAN Faceting Questions #

Let’s break down each of your questions with practical, CKAN-specific context:

1. Why does CKAN use the textgen type for the author field by default?

The textgen field type in CKAN’s default Solr schema is built for full-text search, not faceting. It runs text processing steps like tokenization (splitting names into individual words), lowercasing, and light stemming—ideal when users need to search for fragments of an author’s name (e.g., typing "statistik" to find "Amt für Statistik").

CKAN’s out-of-the-box setup prioritizes search functionality over faceting for fields like author, since most users first need to discover datasets via keyword searches rather than strict exact-value filters.

2. Will changing author to string break other CKAN features?

It depends on your instance’s needs:

• What won’t break: Displaying the author name on dataset pages (this pulls directly from the database metadata, not Solr’s indexed values).
• What might break: Full-text search for author names. If you switch to string, Solr will only return exact matches—so a user searching for "statistik" won’t find datasets where the author is "Amt für Statistik", since string treats the entire value as a single, case-sensitive token.
• Other edge cases: Autocomplete for author names (if you use that feature) will stop working as expected, since it relies on Solr’s tokenized text processing.

If your top priority is faceting and you don’t need partial-keyword search for authors, the switch is safe. But if you want both search and faceting, this isn’t the optimal fix.

3. Optimal approach: Use a copy field to keep both search and faceting

This is the recommended solution—it preserves the full-text search benefits of textgen while enabling precise, unmodified faceting. Here’s how to implement it:

1. Update your Solr schema (typically located at /etc/solr/schema.xml):

   • Add a new string field dedicated to faceting:

     <field name="author_facet" type="string" indexed="true" stored="true"/>
     

   • Add a copy rule to mirror the author field’s content to this new field:

     <copyField source="author" dest="author_facet"/>
     

2. Rebuild the CKAN search index to populate the new field:

   paster --plugin=ckan search-index rebuild -c /etc/ckan/default/production.ini
   

3. Adjust your dataset_facets method to use the new faceting field:

   def dataset_facets(self, facets_dict, package_type):
       if package_type == 'dataset':
           facets_dict['author_facet'] = toolkit._(u'Author')
       return facets_dict
   

This setup gives you the best of both worlds:

• Users can still search for author name fragments using the original textgen field
• Your faceting will display exact, unprocessed author names (like "Amt für Statistik") instead of tokenized lowercase values

内容的提问来源于stack exchange，提问作者Knud Müller