System Design

Indexing & Fuzzy Matching Study Portal

B-trees and what they cannot do, trigram/GIN, blocking keys, Fellegi-Sunter, phonetics, Jaro-Winkler, IDF rarity, partitioning, row vs column stores, and why embeddings are wrong for names
Flash Cards Practice
System Design / Indexing & Fuzzy Matching / Indexing & Fuzzy Name Matching / References

Indexing & Fuzzy Name Matching · References

1 min read
Senior22 min read
Rapid overview

References

  • Splink — probabilistic record linkage implementing Fellegi-Sunter: (verified 2026-09-05)
  • OpenSearch Explain API — per-document score breakdown (boost, idf, tf, dl, avgdl): (verified 2026-09-05)
  • PostgreSQL GIN implementation notes — fastupdate, the unsorted pending list and gin_pending_list_limit:
  • PostgreSQL table partitioning — partition pruning, per-partition indexes, and detaching/dropping a partition for retention:
  • The trigram index this section is about, in our own AML code: PROOViD/AMLService/AMLService/Migrations/20260617062152_AddWatchlists.cs:83-88 (ix_watchlist_names_trgm ... USING gin ("NormalizedName" gin_trgm_ops)), read at the retrieval call site PROOViD/AMLService/AMLService/Infrastructure/Screening/LocalWatchlistProvider.cs:167 (verified 2026-09-05)

See also