We currently use Sphinx 2.2 and have recently found out about Manticore.
We currently index data from the oldest to newest(id is increased) in Sphinx.
We use ngram_len =1 for CJK indexing but have poor performance for common letters(indexing about 45GB of sphinx CJK data).
On out tests using cutoff and putting the data in newest to oldest in Sphinx, we would be able to solve the very slow queries by having Sphinx not search the whole index.
However, Sphinx does not support negative values in sql_range_step so we could not index backwards from newest to oldest so that we could use cutoff.
Also we have already have negative index values for the id as negative(id) in MySQL. This allows us to use asc indexes in mysql which are faster than desc. However, when we tried this method since negative sql_range_step are not supported in Sphinx, we noticed that Sphinx converts all the negative values to positive values for $start and $end.
Therefore we have four questions regarding Manticore search.
Does Manticore sql_range_step support negative values
Does Manticore change negative sql return values to positive values like Sphinx?
Would Manticore have superior performance regarding ngram_len = 1 compared to Sphinx?
Does stopwords = ko for Manticorehave an effect on ngram_len = 1?
Are there any way to use an Analyzer for Korean language in Manticore?