I want to combine full text search with autocomplete in a search field. I use a hunspell token filter in the main analyzer for searching. But I want to add ngram in addition.
When I type queries in a search field letter by letter I can see results from time to time (only in those cases when my tokenized query matched at least with one stem) otherwise results will be empty (in a moment when word has been written partly for example). In shuch situations I want to add ngram working results in order to avoid emptiness.
Should I do 2 indexes for this reason. In the first index I will keep data which have been stored by hunspell tokenizer. In the second with Ngram. This separation helps to avoid mixing saved tokens (stems) of both analyzers(tokenizers). As consequence I have to send second request in case when the first response (to 1 index) is empty.
If I keep all data in one index which processed by both analyzers a tokens intersection of different analyzers will interfere in relevance scoring.
Is there elegant way to solve this problem using only one index??