[Elasticsearch] Consider multi-field for blob.content
When working on https://gitlab.com/gitlab-org/gitlab-ee/issues/5585 an issue with our usage of an edgeNGram
filter for our code analyzer came to light: The filter has no way of keeping the original term, which means that searches for long terms above the max_gram
setting (example: Gitlab::CurrentSettings.current_application_settings
) won't bring back results.
We should consider removing the edgeNGram
filter entirely from our code analyzer and instead use a multi field on any fields that need the edgeNgram as a separate filter, so that we can keep the original. This also should have the added benefit of generating less ngrams.
We should also test how this change affects our index size.