Hi, A colleague of mine investigated that in conjunction with DocValues some time ago:
*Intro* enableLazyFieldLoading is a query time configuration that regulates how field values are retrieved from the index data structures and returned to the user via search results. Currently Apache Solr fetches field contents from the stored data structures or the docValues : <http://goog_698189083/> https://sease.io/2020/03/docvalues-vs-stored-fields-apache-solr-features-and-performance-smackdown.html If this parameter is set to true, then fields that are not directly requested will be loaded lazily as needed. This can boost performance if the most common queries only need a small subset of fields, especially if infrequently accessed fields are large in size. Hope it helps -------------------------- Alessandro Benedetti Apache Lucene/Solr Committer Director, R&D Software Engineer, Search Consultant www.sease.io On Thu, 3 Jun 2021 at 15:32, 기준 <[email protected]> wrote: > Hello experts! > > In ES, even a client specified only id fields to retrieve, it tries to > lookup stored fields after acquired document ids for it's own good. > > > https://github.com/elastic/elasticsearch/blob/master/server/src/main/java/org/elasticsearch/search/fetch/FetchPhase.java#L109 > > I just wonder the case is same in Solr. > > Does it involve any kind of reading stored fields even a client asked it > to retrieve only id field of documents that match the given query? > > The reason for asking this is that, reading a stored field basically is a > decompression which requires a lot of computation costs. > > It would be appreciated if anyone could tell this, or could hint me a name > of class to digging into. > > Thank you in advance. >
