Skip to content

Maintenance release 1.3
Compare
Choose a tag to compare
@romainruaud romainruaud released this 17 Feb 16:56
· 8 commits to master since this release
1.3
138e073

New features :

  • New sort order on facets : "Admin sort" that displays facets options sorted by the attribute's options sort order defined in Back-Office.
  • Various performances improvements
    • Replication is now stopped during full reindex
    • Better performances on clusters
  • Revamp of categories indexation :
    • Performance gain during indexation
    • Ability to choose if category names should be use for fulltext search on products or not (for each category in the back office).

Relevance improvements :

Stopwords and cutoff frequency :

Stopwords are now indexed. Fulltext queries now uses Cutoff Frequency (https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl-match-query.html#query-dsl-match-query-cutoff).

Based on this Cutoff Frequency, the engine is now able to determine which words are stopwords and autodetect them. They will trigger a boost on the results if they are found, but won't impact filtering. Eg. the following query :

"The little horse" : Conjunction of "little" and "horse" is the mandatory part of the query for matching. "the" is optional part of the query, but if the word is found, a boost will be applied on matched results.

Cutoff frequency can be managed on the Back-Office via System > Configuration > Smile Searchandising Suite > Advanced Search Settings > Cutoff Frequency.

Phrase matching :

Phrase matching has also been revamped. It allows to apply a boost on phrase "chunks" if they are found in documents. Eg for the query "white horse", if "white" and "horse" appears in the document, and also the position of the term "horse" is 1 more than the term "white", the document will have a boost into results.

Phrase matching boost can be managed on the Back-Office via System > Configuration > Smile Searchandising Suite > Advanced Search Settings > Phrase match configuration.

Spellchecking :

Spellchecking has also been improved to benefits both precedents points. It's relevance is now far better than before and additionally, the engine is now able to deal with pure stopwords queries such as "to be or not to be", or "he and she".

Boolean attributes :

Boolean attributes indexation has been reworked too : attribute label is now indexed for products having attribute's value to "Yes", this means a product (let's say, a watch) having a "Adjustable" boolean attribute set to "Yes" will now match a fulltext query like "adjustable watch".

Fixed :

  • Autocomplete blocks now supports generic blocks cache tags
  • Bug where multiselect attributes were not properly merged between parent and children products
  • Bug where rating facet was computed on non-approved ratings
Assets 2
Loading