Good morning 👋 To introduce *autocompletion* for *Elasticsearch* queries, we: • use `Completion

UK5GL0DEF

Good morning 👋
To introduce autocompletion for Elasticsearch queries, we:
• use CompletionQueryExpanderPlugin and CompletionResultFormatterPlugin in the CatalogDependencyProvider 
• use edge-ngram as filter in page.json to provide the starting tokens of our words from the index —> this works fine 🎉
• Spryker docs say that the completion words will be picked from the completion_terms field in page.json: but we don’t have such a field, instead the completion terms are being picked from the index as expected (because we have edge-ngram) ✅
To introduce spelling suggestion, we:
• use SpellingSuggestionQueryExpanderPlugin and SpellingSuggestionResultFormatterPlugin in the CatalogDependencyProvider
• Spryker docs say that all suggestions will be picked from the suggestion_terms field in page.json -> can we avoid this and pick from the index instead?
• can we not specify sth (e.g. fuzzy) in page.json to get spelling suggestions from the index ❓ 🙇

U01T075RRHD

I'm not sure what you mean by from the index but I guess you are referring to the full_text and full_text_boosted fields.
You should be careful with completions though. The idea is that these are extremely fast to provide almost instant feedback to the user. This is why Elasticsearch provides a special strategy for this:
https://www.elastic.co/guide/en/elasticsearch/reference/7.12/search-suggesters.html#completion-suggester
Spryker is making use of that by providing the completion_terms field (as far as I know)

The approach is very similar for providing spelling suggestions:
https://www.elastic.co/guide/en/elasticsearch/reference/current/search-suggesters.html

UK5GL0DEF

I’m also not sure what i mean with from the index 😄
(i think it’s the magic Elasticsearch index that maps words from documents to the related document-id (and the word’s frequency)).
I thought there was a possibility to get spelling suggestions directly from the index and not from a field in the page.json because maybe the suggestion_terms are not complete while the index is complete.
The Elasticsearch docs refer to it here https://www.elastic.co/guide/en/elasticsearch/reference/current/search-suggesters.html#fuzzy, but I think I understand now: for fuzzy queries, they don’t suggest corrected terms but just return corrected results.
So, I guess I have to add the suggestion_terms list if I want to have spelling suggestions.

UK5GL0DEF

I guess, suggestion_terms should be added like this in page.json ?

"mappings": {
    "page": {
      "properties": {
        "full-text": {
          (...)
        },
        "full-text-boosted": {
          (...)
        },
        "search_data": {
          "suggestion_terms": {
            "type": "string",
            "index_analyzer": "term_suggestion_analyzer",
            "search_analyzer": "lowercase_analyzer"
          }
        }
      }
    }

I don’t need to provide the actual spelling suggestions, right?

"suggestion_terms": {
   "type": "string",
   "suggestion_terms": ["hammer", "radio", "sofa"]

U01T075RRHD

Correct, you don't. Albeit the results can sometimes be a bit weird since Elasticsearch determines the suggestions from the terms it knows and they are not always suitable.
I think Spryker provides all the necessary plugins out of the box so you don't need to manually add a new field to your page document.

UK5GL0DEF

Yes, I see that now, it’s all there 🎉
The query already returns suggestions in case of a typo: I’ll just call CatalogController again in this case now, so the results for the suggested term are being displayed immediately.

U01T075RRHD

Nice, I'm happy you got it to work