corium-core

Core Corium data types: values, datoms, ids, schema, and the sortable byte encoding shared by every other crate.

What it does

corium-core is the foundation of the workspace. It defines the compact Rust representation of everything the engine stores and moves around:

• Value — the scalar value type (keywords, strings, numbers, refs, instants, …) plus TotalF64 for total-ordered floats.
• Datom [e a v tx added] and the four IndexOrders (EAVT, AEVT, AVET, VAET) that datoms are sorted into.
• Ids — EntityId, AttrId, TxId, KwId, and Partition/PartitionId for the partitioned entity-id space.
• Keyword and the KeywordInterner used to map :db/ident names to ids.
• Schema model — Attribute, ValueType, Cardinality, Unique.
• encoding — the order-preserving, tagged byte encoding (Encodable, encode_value) that makes index keys sort correctly and is reused by the segment and wire formats.

Dependencies

• thiserror for error types.

It has no internal corium-* dependencies — every other crate depends on it, directly or transitively. It is pure library code with no async, network, or storage dependencies.

Architecture

This crate is deliberately minimal and allocation-conscious: it holds only the types and encodings that must be identical everywhere in the system. The sortable encoding is the load-bearing piece — because index keys are produced here, the same bytes drive in-memory index ordering (corium-index), the persistent segment format (corium-store), and the value bytes carried over the wire (corium-protocol). Keeping it dependency-free keeps the whole engine testable in the deterministic simulator. See docs/design/data-model.md.