apognu / motiva / 27405010049

90%

push

github

Always scope queries to their scope's index pattern, skip combining names if enough names already exist.

This is a performance commit:

 - A query's scope was only used to determine the dataset filter, but it still only targeted the overall
   alias, meaning performance was left on the table while other member indices were scanned, even though
   they would never yield any result.
   Now, we add an `_index` prefix filter to the query, which Elasticsearch is able to optimize to skip
   non-matching indices entirely.
 - We used to always create new names by combining name parts (first, second, middle, mother, father, last),
   which is useful for search entities that don't provide a lot of information.
   Yet, when full entities are used for search, the combinatorial explosion can get unwieldy (we found
   a case where 105k synthetic names were built, adding significantly to the latency, before even hitting
   Elasticsearch.
   This PR skips the name combination function if we already have more than 20 name-like properties.

12 of 15 new or added lines in 2 files covered. (80.0%)

4020 of 4435 relevant lines covered (90.64%)

1171.71 hits per line