DocumentDB Extends RUM Index to Enable Ordered Scans on Flexible BSON Fields

DocumentDB has introduced Extended RUM, an enhancement to PostgreSQL's RUM access method designed to handle the flexible, schema-less nature of document database fields. Traditional B-tree indexes work well for fixed scalar columns, while GIN and RUM inverted indexes support flexible and repeated values but lack native support for document-style ORDER BY with LIMIT. Extended RUM addresses this by generating composite index terms from indexed paths and applying an ordering transform during the scan. The result is a multikey-style inverted index capable of filtering, sorting, and stopping at a LIMIT boundary within a single Index Scan operation. This approach preserves correct document semantics for arrays and missing fields, making it suitable for efficient querying of BSON documents stored in PostgreSQL.