Dynamic summary not working with query on input field

[eostis]

Describe the bug
Dynamic summary not working with query on input field

To Reproduce

1. Define a schema field with a summary attribute "wpsolr_title"

field wpsolr_title type string {
     indexing: summary | index
     index: enable-bm25
     summary: dynamic
   }

3. Define a schema input field "wpsolr_input_text"

field wpsolr_input_text type string {
    indexing: (input wpsolr_text || "") . " " . (input wpsolr__id || "") . " " . (input wpsolr_id || "") . " " . (input wpsolr_title || "") . " " . (input wpsolr_post_status_s || "") . " " . (input wpsolr_content || "") . " " . (input wpsolr_snippet_s || "") . " " . (input wpsolr_author || "") . " " . (input wpsolr_author_s || "") . " " . (input wpsolr_post_author_s || "") . " " . (input wpsolr_type || "") . " " . (input wpsolr_meta_type_s || "") . " " . (input wpsolr_permalink || "") . " " . (input wpsolr_comments || "") . " " . (input wpsolr_categories || "") . " " . (input wpsolr_categories_str || "") . " " . (input wpsolr_category_str || "") . " " . (input wpsolr_tags || "") . " " . (input wpsolr_tags_str || "") . " " . (input wpsolr_post_tag_str || "") . " " . (input wpsolr_blog_name_str || "") . " " . (input wpsolr_post_thumbnail_href_str || "") . " " . (input wpsolr_post_href_str || "") . " " . (input wpsolr_title_s || "") . " " . (input wpsolr_post_sticky_s || "") . " " . (input wpsolr_is_excluded_s || "") . " " . (input wpsolr_flat_hierarchy_categories_str || "") . " " . (input wpsolr_non_flat_hierarchy_categories_str || "") . " " . (input wpsolr_product_visibility_str || "") . " " . (input wpsolr_product_stock_str || "") . " " . (input wpsolr_product_cat_str || "") . " " . (input wpsolr_flat_hierarchy_product_cat_str || "") . " " . (input wpsolr_non_flat_hierarchy_product_cat_str || "") | index
    index: enable-bm25
  }

Expected behavior

1. Querying wpsolr_input_text should but does not produce a dynamic summary for wpsolr_title

curl -X POST -H "Content-Type: application/json" --data '
{"query":"page", "yql":"select wpsolr_title from index1 where ([{\"defaultIndex\":\"wpsolr_input_text\"}]userInput(@query)) "}
' \
http://localhost:8080/search/

{"root":{"id":"toplevel","relevance":1.0,"fields":{"totalCount":5},"coverage":{"coverage":100,"documents":16,"full":true,"nodes":1,"results":1,"resultsFull":1},"children":[{"id":"index:wpsolr/0/c81e728db67e6fc24b6be56d","relevance":0.31803513173241105,"source":"wpsolr","fields":{"wpsolr_title":"Sample Page"}},...]}}

2. Querying wpsolr_title should and does produce a dynamic summary for wpsolr_title

curl -X POST -H "Content-Type: application/json" --data '
{"query":"page","yql":"select wpsolr_title from index1 where ([{\"defaultIndex\":\"wpsolr_title\"}]userInput(@query)) "}
' \
http://localhost:8080/search/

{"root":{"id":"toplevel","relevance":1.0,"fields":{"totalCount":1},"coverage":{"coverage":100,"documents":16,"full":true,"nodes":1,"results":1,"resultsFull":1},"children":[{"id":"index:wpsolr/0/c81e728db67e6fc24b6be56d","relevance":0.16343879032006284,"source":"wpsolr","fields":{"wpsolr_title":"Sample <hi>Page</hi>"}}]}}

Environment (please complete the following information):

• Docker Vespa 8.253.3 self hosted on MacOS

[jobergum]

There must be a misunderstanding here.

Expected behavior
Querying wpsolr_input_text should but does not produce a dynamic summary for wpsolr_title

This might be your expectation, but this is not how synthetic fields works.

document doc {
   field foo type string {..}
}
  field foo_derived type string {
    indexing: input foo | something else
  }

Creates a new field foo_derived with the contents from foo and then some transformation. This is the context of indexing. At query time, there is no reference to where or how the field contents was created during indexing.

[eostis]

I asked because it is a standard feature of Lucene engines that I already use.

Example with OpenSearch:

• "text" field defined as an input (copied from other fields, like title and description):

• Query on "text" returns highlighting on "title" field:

[jobergum]

That you get with Vespa fieldsets which don’t require copying anything. Synthetic fields outside of document using Input is just when you need additional processing, like embedding or splitting.

[jobergum]

Closing this out - see fieldset if you want to search multiple fields. This also avoids bloating the index by copying data to many field indexes.

[eostis]

Fieldsets have limitations that can only be overcome with input fields.

> I bet you get a deployment warning about the mismatch in match settings between the fields you include in the fieldsets. Fields bundled in a fieldset must have the same match settings, implicit (defaults) or explicit. This is because the query can only be transformed with one setting.

[jobergum]

It’s a convenience feature for searching string fields that share match settings. For other cases one can simply expand the query to search multiple fieldsets or fields using for example OR operator.

[eostis]

Thanks.