SQLite Reference

When to Load

Load when the project uses SQLite (signal: sqlite3, better-sqlite3, SQLite, better_sqlite3, .db / .sqlite file paths, DATABASE_URL with sqlite://).

CLI Connection

# Open a database file:
sqlite3 "$DB_PATH"

# Run a query directly:
sqlite3 "$DB_PATH" "SELECT * FROM users LIMIT 5;"

# Read-only mode (safe for production inspection):
sqlite3 "file:${DB_PATH}?mode=ro"

# In-memory database (testing):
sqlite3 ":memory:"

Useful dot-commands inside sqlite3:

.schema              -- show all CREATE statements
.tables              -- list all tables
.indexes <table>     -- list indexes for a table
.mode column         -- columnar output
.headers on          -- show column names
.timer on            -- show query execution time
EXPLAIN QUERY PLAN SELECT ...;   -- show index usage

Connection Modes

URI	Behaviour
file:path/to/db	Default read/write
file:path/to/db?mode=ro	Read-only — no write locks
file:path/to/db?mode=memory	Private in-memory DB
file:dbname?mode=memory&cache=shared	Shared in-memory DB (same process)
:memory:	Convenience alias for private in-memory

Multiple connections to the same file: always enable WAL mode for concurrent readers.

WAL Mode

Enable WAL (Write-Ahead Logging) for any concurrent workload:

PRAGMA journal_mode = WAL;
PRAGMA synchronous = NORMAL;  -- safe with WAL; 'FULL' for max durability

WAL benefits:

• Readers do not block writers; writers do not block readers
• Faster writes (sequential appends vs random page writes)
• Survives crashes without corruption (WAL is checkpointed periodically)

Checkpoint manually when needed:

PRAGMA wal_checkpoint(TRUNCATE);

Pragmas Worth Knowing

PRAGMA foreign_keys = ON;       -- OFF by default — always enable
PRAGMA journal_mode = WAL;
PRAGMA synchronous = NORMAL;
PRAGMA busy_timeout = 5000;     -- wait 5 s before returning SQLITE_BUSY
PRAGMA cache_size = -64000;     -- 64 MB page cache (negative = kibibytes)
PRAGMA temp_store = MEMORY;     -- store temp tables in RAM
PRAGMA mmap_size = 134217728;   -- 128 MB memory-mapped I/O

These pragmas are session-level unless set in the connection setup. Set them immediately after OPEN.

Indexes

-- Covering index (includes all queried columns):
CREATE INDEX idx_orders_user ON orders(user_id, created_at, status);

-- Partial index:
CREATE INDEX idx_active_users ON users(email) WHERE deleted_at IS NULL;

-- Expression index:
CREATE INDEX idx_lower_email ON users(LOWER(email));

SQLite uses the query planner (sqlite_stat1): run ANALYZE after bulk inserts to update statistics.

JSON1 Extension

Available in SQLite 3.9+; enabled by default in most distributions.

-- Extract a field:
SELECT json_extract(data, '$.name') FROM events;

-- Filter on JSON field:
SELECT * FROM events WHERE json_extract(data, '$.type') = 'purchase';

-- Index a JSON field via generated column (SQLite 3.31+):
ALTER TABLE events ADD COLUMN event_type TEXT
  GENERATED ALWAYS AS (json_extract(data, '$.type')) VIRTUAL;
CREATE INDEX idx_event_type ON events(event_type);

FTS5 (Full-Text Search)

CREATE VIRTUAL TABLE docs_fts USING fts5(title, body, content='docs', content_rowid='id');

-- Query:
SELECT rowid, rank FROM docs_fts WHERE docs_fts MATCH 'sqlite performance' ORDER BY rank;

-- Keep FTS table in sync via triggers:
CREATE TRIGGER docs_ai AFTER INSERT ON docs BEGIN
  INSERT INTO docs_fts(rowid, title, body) VALUES (new.id, new.title, new.body);
END;

Concurrency Limits

SQLite is not designed for high write concurrency:

• One writer at a time (WAL allows concurrent readers + one writer)
• For > 1 writer or network-facing applications, consider PostgreSQL or MySQL
• Acceptable for: embedded apps, mobile, local-first, low-concurrency servers, test fixtures, CLI tools

Common Gotchas

Gotcha	Fix
Foreign keys off by default	PRAGMA foreign_keys = ON on every connection
SQLITE_BUSY under write contention	PRAGMA busy_timeout = 5000; serialize writes in application
Type affinity surprises (1 = '1' is true)	Use strict typing (SQLite 3.37+ STRICT tables) or validate in application
WAL file grows unbounded if checkpoint never runs	PRAGMA wal_autocheckpoint = 1000 (default); or manual checkpoint
INTEGER PRIMARY KEY vs ROWID	INTEGER PRIMARY KEY is an alias for rowid; use it for efficient row lookup
No ALTER TABLE … DROP COLUMN before 3.35.0	Upgrade or recreate the table via the rename-insert-drop-rename pattern