Crate mongreldb_core 

MongrelDB core — a log-structured columnar store with sub-ms writes, learned indexes over a shared row-id space, page-level native encryption, an MVCC-tagged content-addressed cache, and an AI-native access layer.

The crate owns the WAL + memtable + Bε-tree write path, sorted-run container formats, MVCC snapshots, page cache, encryption, compaction, and indexes.

Re-exports

pub use auth::hash_password;

pub use auth::verify_password;

pub use auth::Permission;

pub use auth::Principal;

pub use auth::RoleEntry;

pub use auth::UserEntry;

pub use be_tree::BeTree;

pub use cache::PageCache;

pub use columnar::decode_column;

pub use columnar::encode_column;

pub use cursor::drain_cursor_to_columns;

pub use cursor::Cursor;

pub use cursor::MultiRunCursor;

pub use cursor::NativePageCursor;

pub use database::CheckIssue;

pub use database::Database;

pub use database::ExternalTriggerBaseWrite;

pub use database::ExternalTriggerBridge;

pub use database::ExternalTriggerWrite;

pub use database::ExternalTriggerWriteResult;

pub use encryption::Cipher;

pub use encryption::PlaintextCipher;

pub use engine::AggState;

pub use engine::ApproxAgg;

pub use engine::ApproxResult;

pub use engine::CachedAgg;

pub use engine::ColumnStat;

pub use engine::IncrementalAggResult;

pub use engine::IndexBuildPolicy;

pub use engine::NativeAgg;

pub use engine::NativeAggResult;

pub use engine::Table;

pub use epoch::Epoch;

pub use epoch::EpochAuthority;

pub use epoch::EpochClock;

pub use epoch::Snapshot;

pub use error::MongrelError;

pub use error::Result;

pub use external_table::ExternalTableDefinition;

pub use external_table::ExternalTableEntry;

pub use external_table::ModuleArg;

pub use external_table::ModuleCapabilities;

pub use gc::CheckReport;

pub use gc::DoctorReport;

pub use gc::GcReport;

pub use index::AnnIndex;

pub use index::BitmapIndex;

pub use index::ColumnLearnedRange;

pub use index::FmIndex;

pub use index::HotIndex;

pub use index::LearnedIndex;

pub use index::SparseIndex;

pub use memtable::Memtable;

pub use memtable::Row;

pub use memtable::Value;

pub use mutable_run::MutableRun;

pub use page::CachedPage;

pub use page::Encoding;

pub use page::PageStat;

pub use procedure::ProcedureBody;

pub use procedure::ProcedureCallOutput;

pub use procedure::ProcedureCallResult;

pub use procedure::ProcedureCallRow;

pub use procedure::ProcedureCondition;

pub use procedure::ProcedureEntry;

pub use procedure::ProcedureMode;

pub use procedure::ProcedureParam;

pub use procedure::ProcedureStep;

pub use procedure::ProcedureValue;

pub use procedure::StoredProcedure;

pub use query::Condition;

pub use query::Query;

pub use reservoir::Reservoir;

pub use retention::OwnedSnapshotGuard;

pub use retention::SnapshotGuard;

pub use retention::SnapshotRegistry;

pub use rowid::RowId;

pub use rowid::RowIdAllocator;

pub use schema::AlterColumn;

pub use schema::ColumnDef;

pub use schema::ColumnFlags;

pub use schema::IndexDef;

pub use schema::IndexKind;

pub use schema::Schema;

pub use schema::TypeId;

pub use sorted_run::read_column_dir;

pub use sorted_run::read_header;

pub use sorted_run::write_run;

pub use sorted_run::write_run_with;

pub use sorted_run::ColumnPayload;

pub use sorted_run::RunHeader;

pub use sorted_run::RunReader;

pub use sorted_run::RunSpec;

pub use sorted_run::RunWriter;

pub use trace::IndexRebuild;

pub use trace::QueryTrace;

pub use trace::ScanMode;

pub use trigger::StoredTrigger;

pub use trigger::TriggerCell;

pub use trigger::TriggerCondition;

pub use trigger::TriggerConfig;

pub use trigger::TriggerDefinition;

pub use trigger::TriggerEntry;

pub use trigger::TriggerEvent;

pub use trigger::TriggerExpr;

pub use trigger::TriggerProgram;

pub use trigger::TriggerRaiseAction;

pub use trigger::TriggerStep;

pub use trigger::TriggerTarget;

pub use trigger::TriggerTiming;

pub use trigger::TriggerValue;

pub use txn::IsolationLevel;

pub use txn::OwnedRow;

pub use txn::PutResult;

pub use txn::UpsertAction;

pub use txn::UpsertActionKind;

pub use txn::UpsertResult;

pub use wal::AddedRun;

pub use wal::DdlOp;

pub use wal::Op;

pub use wal::Record;

pub use wal::SharedWal;

pub use wal::Wal;

pub use wal::WalReader;

pub use wal::SYSTEM_TXN_ID;

Modules

auth

User, role, and credential management for MongrelDB’s catalog-level auth.

be_tree

Buffered Bε-tree (B-epsilon-tree) over composite (RowId, Epoch) keys — the Phase 1 memtable target.

cache

MVCC-tagged, content-addressed page cache.

catalog

DB-wide catalog checkpoint (spec §5.1).

columnar

Plain columnar encode/decode for the prototype type set.

compaction

Tiered compaction (Phase 5) with snapshot retention.

constraint

Engine-side declarative constraints (unique, foreign key, check) enforced authoritatively inside the core transaction path. Opt-in per-table via crate::schema::Schema::constraints; tables with an empty constraint set (the default — including every legacy table and every Kit-managed table) behave exactly as before. This subsystem is independent of the Kit’s own guard-table mechanism: the Kit continues to enforce its constraints exactly as before, and these engine constraints only fire for tables whose schema carries a non-empty TableConstraints.

cursor

Lazy, page-aware native-column cursor for streaming scans (Phase 6.2).

database

Multi-table Database container (spec §5, §6, §10).

encryption

Native page-level encryption (optional, behind the encryption feature).

engine

The engine tying the write and read paths together.

epoch

error

external_table

gc

Operational tools: garbage collection and integrity check.

global_idx

Global index file — _idx/global.idx.

index

Indexes. All share the crate::rowid::RowId space so multi-condition queries intersect with SIMD bitmap ops.

manifest

Manifest — the atomic pointer to the current set of sorted runs.

memtable

In-memory write buffer (the “memtable”).

mutable_run

Mutable run tier — the LSM layer between the skip-list memtable and the immutable .sr sorted runs (Phase 11.1).

page

pma

Packed Memory Array — a cache-oblivious sorted array with amortized O(log² n) insert and no full rewrite per insert.

procedure

query

Tool-call-native query surface.

reservoir

Reservoir sampling for approximate analytics (Phase 8.2).

retention

Global snapshot-retention registry for a multi-table Database.

rowid

schema

sorted_run

Sorted Run — the immutable columnar unit (.sr).

trace

Query path instrumentation (OPTIMIZATIONS.md Priority 0 / 16).

trigger

tsv

TSV import/export (MongrelDB-compatible format).

txn

Cross-table transactions on the shared WAL (spec §8.2, single-applier subset — parallelism arrives in P3).

wal

Append-only, group-commit, torn-write-safe WAL.