Database Compatibility Matrix

This page shows which FraiseQL features are supported on each database backend. Check this before starting a project to avoid discovering limitations mid-development.

Tip

PostgreSQL is the recommended database for production deployments. It is the only backend with full feature support. All other databases have one or more limitations.

---

Core Features

Feature	PostgreSQL	MySQL 8+	SQL Server	SQLite
Queries (SELECT from views)	✅	✅	✅	✅
Mutations	✅	✅	✅	✅ DirectSql
Relay cursor pagination	✅	✅	✅	✅
Automatic WHERE from args	✅	✅	✅	✅
Rich filters (operators, ILIKE, IS NULL)	✅	✅	✅	✅
Full-text search	✅	⚠️ Limited	⚠️ Limited	❌
JSONB field traversal	✅	⚠️ JSON_EXTRACT	⚠️ JSON_VALUE	⚠️ json_extract
Row-Level Security (RLS)	✅ Native	⚠️ View-based	⚠️ View-based	❌

---

Advanced Query Features

Feature	PostgreSQL	MySQL 8+	SQL Server	SQLite
Window functions	✅ Full	⚠️ Partial	⚠️ Partial	⚠️ Partial
Fact table analytics (tf_*)	✅	❌	❌	❌
Automatic Persisted Queries (APQ)	✅	✅	✅	✅
Query result caching	✅	✅	✅	✅
Subscriptions (CDC)	✅ pg_notify	❌	❌	❌
Multi-tenancy / RLS injection	✅	⚠️ Manual	⚠️ Manual	❌

---

Window Function Support Detail

Function	PostgreSQL	MySQL 8+	SQL Server	SQLite
ROW_NUMBER(), RANK(), DENSE_RANK()	✅	✅	✅	✅
NTILE(n)	✅	✅	✅	✅
PERCENT_RANK(), CUME_DIST()	✅	❌	✅	❌
LAG(), LEAD(), FIRST_VALUE(), LAST_VALUE()	✅	✅	✅	✅
SUM() OVER, COUNT() OVER	✅	✅	✅	✅
STDDEV() OVER, VARIANCE() OVER	✅	❌	✅	❌
FILTER (WHERE ...) on aggregates	✅	❌	❌	❌
GROUPS frame type	✅	❌	❌	❌

---

Operational Features

Feature	PostgreSQL	MySQL 8+	SQL Server	SQLite
GraphQL transport	✅	✅	✅	✅
REST transport	✅	✅	✅	✅
gRPC transport	✅	✅	✅	✅
Wire protocol (binary streaming)	✅	❌	❌	❌
Apollo Federation subgraph	✅	✅	✅	❌
Federation circuit breaker	✅	✅	✅	❌
TLS / mTLS connections	✅	✅	✅	N/A
Connection pooling (PgBouncer-compatible)	✅	✅	✅	N/A
Health check endpoint	✅	✅	✅	✅

---

Legend

Symbol	Meaning
✅	Fully supported
⚠️	Partial support — see notes below
❌	Not supported
N/A	Not applicable

---

Notes by Database

MySQL 8+

• Mutations — fully supported. All mutation functions use the same mutation_response return type as PostgreSQL.
• Window functions — ranking and value functions work. PERCENT_RANK, CUME_DIST, STDDEV, VARIANCE aggregate-as-window, and the FILTER clause are not supported. The planner raises an error at compile time for unsupported functions.
• Fact table analytics — not supported. Fact table introspection requires JSONB operators (PostgreSQL-native). MySQL uses JSON_EXTRACT which lacks the binary JSONB aggregation operators.
• RLS — FraiseQL implements view-based security (inject a WHERE tenant_id = $user filter). MySQL does not have native RLS like PostgreSQL.
• Full-text search — MySQL FULLTEXT indexes are supported via the MATCH ... AGAINST operator but require explicit index declarations. Not auto-detected.

SQL Server

• Mutations — fully supported, using SQL Server’s SELECT * FROM fn_name(args) calling convention.
• Window functions — ROWS/RANGE frames work. GROUPS frame type is not supported. PERCENT_RANK is available on SQL Server 2012+.
• Fact table analytics — not supported (no JSONB, limited JSON aggregation).
• JSONB traversal — available via JSON_VALUE(col, '$.path') but requires explicit path declarations; JSONB operators (->, ->>) are PostgreSQL-only.
• UUID columns — use UNIQUEIDENTIFIER; cursor pagination uses CONVERT(UNIQUEIDENTIFIER, @p) comparisons.

SQLite

• Mutations — supported via MutationStrategy::DirectSql. SQLite does not have stored procedures, so FraiseQL executes INSERT, UPDATE, and DELETE statements directly with a RETURNING clause to produce the mutation response. This enables full read-write development and testing against SQLite without requiring PostgreSQL.
• Subscriptions — not supported. SQLite has no LISTEN/NOTIFY equivalent.
• Wire protocol — not supported. SQLite is in-process only.
• Window functions — basic ranking and SUM/COUNT OVER work. PERCENT_RANK, CUME_DIST, STDDEV, VARIANCE, and frame exclusion are not supported.

---

Choosing a Database

Tip

Start with PostgreSQL. All FraiseQL features work, and you can always switch to a different backend later — the GraphQL API surface does not change.

Scenario	Recommendation
Production SaaS / multi-tenant	PostgreSQL (native RLS, subscriptions, JSONB)
Existing MySQL codebase	MySQL 8+ (mutations, relay pagination, most query features)
Enterprise .NET / Windows	SQL Server (full mutation support, most query features)
Local development / tests	SQLite (zero config, fast, no server needed, full read-write support)
Analytics / data warehouse	PostgreSQL (fact tables, window functions, JSONB dimensions)
Real-time features (subscriptions)	PostgreSQL only

---

See Also

• PostgreSQL guide — Production configuration, connection pooling, RLS
• MySQL guide — MySQL-specific idioms and limitations
• SQL Server guide — Enterprise configuration and Windows auth
• SQLite guide — Local development and testing setup