Ridge Compatibility and Limits

Ridge uses familiar relational SQL and a PostgreSQL-wire transport while retaining its own storage, catalog, type, planner, operational, and security contracts.

Compatibility matrix

Capability Ridge v0.3
Embedded database Yes
Native loopback server Bounded pgwire
Relational SQL Bounded documented surface
MVCC Persistent read committed and repeatable read
Concurrent writes Disjoint staging/publication with conflict validation
PostgreSQL drivers Frozen client and query subset
PostgreSQL catalogs No broad compatibility
MySQL-like introspection Ridge-native SHOW and DESCRIBE aliases
Replication or sharding Not implemented

Unsupported or intentionally deferred

Bounded means exact

Limits are part of the API, not suggestions. Queries, rows, messages, plans, work memory, spill, tabs, histories, profiles, exports, and maintenance operations reject overflow explicitly. Consult the current capability and release documents for numeric limits.

Performance claims

Published benchmarks identify the host, engine versions, durability settings, warmup, sample count, workload, correctness checks, and raw artifacts. They are regression baselines only. Do not generalize the one-million-operation or focused concurrency numbers to another schema, machine, durability policy, or fleet.

Porting guidance

  1. Inventory application SQL and driver startup behavior.
  2. Replace PostgreSQL catalog introspection with Ridge SHOW commands.
  3. Test schema migration and every prepared query against a real ridged.
  4. Verify transaction and retry behavior, especially uncertain commits.
  5. Keep unsupported features behind explicit application alternatives.
  6. Run differential fixtures for business-critical semantics.