We currently have a fairly comprehensive suite of tests that validate the behavior of the various $scoreFusion features/options (like combination, normalization, and weights). These tests use simple documents to create precise expectations about the documents and their values that result from the $scoreFusion.

Additionally, we have a rentals dataset (simulating an AirBnB/Zillow-like corpus) that have more realistic documents our users would run a hybrid search on. We currently have $rankFusion tests that query this dataset, and we should add analogous $scoreFusion ones:

• score_fusion_multipipeline_test.js (model off of ranked_fusion_multipipeline_test.js)
• score_fusion_geo_near.js (model off of ranked_fusion_geo_near.js)
  • This is an example $scoreFusion input from the TDD that uses $geoNear in its pipelines
• score_fusion_custom_sort_order.js (model off of ranked_fusion_custom_sort_order.js)
  • This one already exists to test $text, but it looks like it can be expanded to add more cases

Feel free to add any other interesting tests, and potentially retroactively add the analogous one to rank fusion (if it applies).