Querying indexes

Start with Index design and the operator reference.

Query semantics

Ordered index queries

Ordered indexes are used through normal property filters and order clauses.

Examples:

Equals(
    Person.ref { surname } with "garcia lopez"
)

Prefix(
    Person.ref { surname } with "gar"
)

Orders(
    Person.ref { surname }.ascending(),
    Person.ref { firstName }.ascending()
)

Those map naturally onto:

• Reference
• Normalize
• Reversed
• ReferenceToMax
• Multiple

Search index queries

Named search indexes are queried explicitly through search filters.

Examples:

Matches(
    "name" with "garcia lopez"
)

MatchesPrefix(
    "name" with "gar"
)

MatchesRegEx(
    "name" with Regex("^gar.*$")
)

This separation is intentional.

Property filters mean: “match this property using property semantics”

Named search filters mean: “search this named search surface using search semantics”

Geospatial candidate indexes

GeoHash(location.ref(), precisionBits = 32u) narrows GeoWithinBox and GeoWithinRadius scans. It stores one stable geohash cell per record. The planner covers the query bounds with at most 256 byte-aligned prefixes, reducing precision when needed, and exact geometry removes false positives. Geohash ordering is spatial cell order; it does not provide distance sorting or nearest-neighbour ranking.

That keeps ordered equality/prefix/range logic separate from token search logic.