fvh adaptor to highlight the top boost phrase only #799
Conversation
The <em>margarita pizza</em> and the <em>marinara pizza</em> in this pizzeria are yummy and inexpensive.
delicious: 4
margarita pizza: 3
marinara pizza: 3
yummy: 2
|
26c0db4 to
575baa1
Compare
...a/com/yelp/nrtsearch/server/luceneserver/highlights/TopBoostOnlyFragmentsBuilderAdaptor.java
Outdated
Show resolved
Hide resolved
| buffer, index, values, s, fragInfo.getEndOffset(), modifiedStartOffset); | ||
| int srcIndex = 0; | ||
| double topBoostValue = | ||
| fragInfo.getSubInfos().stream().map(SubInfo::getBoost).max(Float::compare).orElse(0f); |
There was a problem hiding this comment.
Please use for loop instead of stream to avoid extra object creation. This code is called per document, so it can be called 1000s of times per second and we don't want it to be slow.
You can actually also store the subInfos with the highest boost along with the topBoostValue, and then you won't need to iterate over all of the subInfos again.
There was a problem hiding this comment.
This is called once for each fragment-to-be-created. We have to calculated the topBoostValue for each fragment as not all fragments contains all the desired phrases.
| if (subInfo.getBoost() < topBoostValue) { | ||
| continue; | ||
| } |
There was a problem hiding this comment.
Does the modifiedStartOffset still track the correct offset even if some of the subInfos are skipped?
There was a problem hiding this comment.
Yes.
This method uses pointer to concatenate the "highlighted" and the other parts. Skipping some of the phrases will still create a complete fragment.
RP-12281
Add
top_boost_onlytoggle to highlight the found highest boosted phrase. This is achieved by adding an adaptor to the FragmentBuilder.The test case is self-explanatory. Only the multiple occurances of the best match will be highlighted.