In Solr 7 document boosting aka index time boosting support has been dropped entirely.
This feature was used for the TypeBoost & HtmlFilter processors. In Search API Solr a similar feature has been implemented that gives the same result and uses lucene payloads.
When setting the boost to 0 for a bundle in the TypeBoost
SearchApiProcessor
the score for all documents of that bundle will be 0!
For example:
[details] => Array
(
[0] => Array
(
[match] => 1
[value] => 8.615736
[description] => sum of:
[details] => Array
(
[0] => Array(4)
)
)
[1] => Array
(
[match] => 1
[value] => 0
[description] => float(boost_document)=0.0
)
)
The boost factor for the entire document is 0. 8.615736*0=0
We should update the description in the UI to make this clear to the end user.
For HtmlFilter it now reads:
"Assign a boost of 0 to ignore the text content of that HTML element."
For TypeBoost there's no documentation yet, we could set this:
"Assign a boost of 0 to not score the item at all."
Comment | File | Size | Author |
---|---|---|---|
#6 | 3054875-6--type_boost_description.patch | 714 bytes | drunken monkey |
|
Comments
Comment #2
mpp CreditAttribution: mpp at AmeXio for District09 commentedComment #3
mpp CreditAttribution: mpp at AmeXio for District09 commentedAdded a description:
Comment #4
mpp CreditAttribution: mpp at AmeXio for District09 commentedComment #5
mkalkbrennerIf the DB backend behaves the same way, we should add that description!
Thomas, would do you think about it?
Comment #6
drunken monkeyYes, this makes a lot of sense. Thanks for the suggestion!
This does work the same in the DB backend, I think: results with item boost 0 will just always be ranked last when sorting by relevance.
Maybe we should also note that 1 is the default, though? Or would you say that’s clear enough?
Comment #7
borisson_I think this is clear enough.
Comment #9
drunken monkeyAlright, good to hear.
Committed. Thanks again, everyone!