.jpg&w=1536&q=75)
Free Lesson
SearchArray - let's rethink full-text search
60 min
Dec 19, 2025 12:30 PM
Virtual (Zoom)
In this video
What you'll learn
Why search engine DSLs continue to confuse users
Every search engine forces you to learn a DSL. Why we should be removing, not adding, to the cognitive load.
How to build your own full-text search
How I built an in-memory full text search system good enough to recreate core search functionality around dataframe
A different perspective on search APIs
Why representing search relevance in a dataframe creates maximum flexibility over a traditional API
Why this topic matters
What if we could treat a search corpus as a pandas/polars dataframe? Then we could write Python instead of complex query DSL.
We could add a title + body score with a a literal Python "+" operator. Or enforce "vespa" is in the title by a simple: scores[title.score('vespa') == 0] = 0.
Many of us are used to numpy + pandas + pals. I share how I've rethought search to be about pure Python.
You'll learn from
Doug Turnbull
Black belt search ninja space cowboy (and consultant)