Postgresql Development

When to Use

• Designing PostgreSQL schemas with proper constraints, types, and indexes.
• Writing advanced queries using CTEs, window functions, JSONB, or array operations.
• Analyzing and optimizing slow queries with EXPLAIN ANALYZE.
• Setting up indexing strategies (B-tree, GIN, GiST, partial, covering indexes).
• Configuring connection pooling with PgBouncer or built-in poolers.
• Managing migrations, transactions, isolation levels, and data integrity.
• Tuning PostgreSQL server parameters for workload characteristics.
• Backup, restore, and disaster recovery with pg_dump and pg_basebackup.

Critical Patterns

• Constraints at the Database Level: Enforce NOT NULL, CHECK, UNIQUE, and foreign keys in the

schema — never rely solely on application-level validation. The database is the last line of defense.

• Index Deliberately: Every index has a write cost. Create indexes based on actual query

patterns from EXPLAIN ANALYZE, not speculation. Monitor with pg_stat_user_indexes to find unused ones.

• Use Appropriate Types: Use uuid (not VARCHAR) for identifiers, timestamptz (not

timestamp) for times, text (not VARCHAR(n)) for variable strings, numeric for money, and jsonb (not json) for document data.

• CTEs Are Not Optimization Barriers (since PG 12): PostgreSQL can inline CTEs. Use them freely

for readability. Add MATERIALIZED only when you explicitly need to force a single evaluation.

• Connection Pooling Is Mandatory: PostgreSQL forks a process per connection. Always use

PgBouncer (or Supavisor, pgcat) in front of PostgreSQL in production. Target max 100-200 actual connections.

• Transactions Should Be Short: Long-running transactions hold locks and block autovacuum. Keep

transactions under a few seconds. For bulk operations, use batched commits.

• Always EXPLAIN ANALYZE: Never guess at query performance. Use

EXPLAIN (ANALYZE, BUFFERS, FORMAT TEXT) to see actual row counts, buffer hits, and execution time.

Code Examples

Schema Design with Proper Types and Constraints

-- Use uuid, timestamptz, text, and proper constraints
CREATE EXTENSION IF NOT EXISTS "uuid-ossp";
CREATE EXTENSION IF NOT EXISTS "pg_trgm";

CREATE TABLE users (
    id          uuid PRIMARY KEY DEFAULT uuid_generate_v4(),
    email       text NOT NULL UNIQUE,
    full_name   text NOT NULL CHECK (length(full_name) BETWEEN 1 AND 200),
    role        text NOT NULL DEFAULT 'member' CHECK (role IN ('admin', 'member', 'viewer')),
    metadata    jsonb NOT NULL DEFAULT '{}',
    created_at  timestamptz NOT NULL DEFAULT now(),
    updated_at  timestamptz NOT NULL DEFAULT now()
);

CREATE TABLE orders (
    id          uuid PRIMARY KEY DEFAULT uuid_generate_v4(),
    user_id     uuid NOT NULL REFERENCES users(id) ON DELETE CASCADE,
    status      text NOT NULL DEFAULT 'pending'
                     CHECK (status IN ('pending', 'confirmed', 'shipped', 'delivered', 'cancelled')),
    total       numeric(12, 2) NOT NULL CHECK (total >= 0),
    items       jsonb NOT NULL DEFAULT '[]',
    created_at  timestamptz NOT NULL DEFAULT now()
);

-- Partial index: only index rows that matter for active queries
CREATE INDEX idx_orders_pending ON orders (user_id, created_at)
    WHERE status = 'pending';

-- GIN index for JSONB containment queries
CREATE INDEX idx_orders_items_gin ON orders USING gin (items);

-- Trigram index for LIKE/ILIKE text search
CREATE INDEX idx_users_name_trgm ON users USING gin (full_name gin_trgm_ops);

-- Auto-update updated_at with a trigger
CREATE OR REPLACE FUNCTION update_updated_at()
RETURNS trigger AS $$
BEGIN
    NEW.updated_at = now();
    RETURN NEW;
END;
$$ LANGUAGE plpgsql;

CREATE TRIGGER trg_users_updated_at
    BEFORE UPDATE ON users
    FOR EACH ROW EXECUTE FUNCTION update_updated_at();

Advanced Queries: CTEs and Window Functions

-- CTE + window function: rank users by order count per month
WITH monthly_orders AS (
    SELECT
        user_id,
        date_trunc('month', created_at) AS month,
        count(*) AS order_count,
        sum(total) AS month_total
    FROM orders
    WHERE status != 'cancelled'
    GROUP BY user_id, date_trunc('month', created_at)
)
SELECT
    u.email,
    mo.month,
    mo.order_count,
    mo.month_total,
    rank() OVER (PARTITION BY mo.month ORDER BY mo.order_count DESC) AS rank
FROM monthly_orders mo
JOIN users u ON u.id = mo.user_id
ORDER BY mo.month DESC, rank;

JSONB Queries

-- Query nested JSONB: find orders containing a specific product
SELECT id, total, items
FROM orders
WHERE items @> '[{"product_id": "abc-123"}]';

-- Extract and aggregate JSONB array elements
SELECT
    o.id,
    elem->>'product_id' AS product_id,
    (elem->>'quantity')::int AS quantity
FROM orders o,
     jsonb_array_elements(o.items) AS elem
WHERE (elem->>'quantity')::int > 5;

-- Update a JSONB field (merge)
UPDATE users
SET metadata = metadata || '{"verified": true, "tier": "premium"}'::jsonb
WHERE id = 'some-uuid';

EXPLAIN ANALYZE: Reading Query Plans

-- Always use ANALYZE + BUFFERS for real execution stats
EXPLAIN (ANALYZE, BUFFERS, FORMAT TEXT)
SELECT u.email, count(o.id) AS order_count
FROM users u
LEFT JOIN orders o ON o.user_id = u.id
WHERE u.created_at > now() - interval '30 days'
GROUP BY u.email
ORDER BY order_count DESC
LIMIT 20;

-- Key things to look for in output:
-- 1. Seq Scan on large tables → missing index
-- 2. actual rows vs. estimated rows differ wildly → stale statistics, run ANALYZE
-- 3. Buffers: shared hit vs. read → cache efficiency
-- 4. Sort Method: external merge → increase work_mem for this query

Indexing Strategies

-- B-tree (default): equality and range queries
CREATE INDEX idx_orders_created ON orders (created_at DESC);

-- Covering index: index-only scans, no table access needed
CREATE INDEX idx_orders_covering ON orders (user_id, created_at) INCLUDE (total, status);

-- Partial index: only index a useful subset
CREATE INDEX idx_orders_active ON orders (created_at)
    WHERE status NOT IN ('delivered', 'cancelled');

-- GiST: range types, geometric, full-text
CREATE INDEX idx_events_during ON events USING gist (tstzrange(start_at, end_at));

-- GIN: arrays, JSONB, trigrams, full-text search
CREATE INDEX idx_users_tags ON users USING gin (tags);

-- Identify unused indexes
SELECT schemaname, relname, indexrelname, idx_scan
FROM pg_stat_user_indexes
WHERE idx_scan = 0
ORDER BY pg_relation_size(indexrelid) DESC;

Migrations Pattern

-- Safe column addition (non-blocking)
ALTER TABLE users ADD COLUMN phone text;

-- UNSAFE: adding NOT NULL with default rewrites entire table on PG < 11
-- SAFE on PG 11+: metadata is stored, no rewrite
ALTER TABLE users ADD COLUMN is_active boolean NOT NULL DEFAULT true;

-- Safe index creation (non-blocking)
CREATE INDEX CONCURRENTLY idx_users_phone ON users (phone);

-- Safe enum value addition
ALTER TYPE order_status ADD VALUE IF NOT EXISTS 'refunded';

Connection Pooling (PgBouncer)

; pgbouncer.ini
[databases]
myapp = host=127.0.0.1 port=5432 dbname=myapp

[pgbouncer]
listen_addr = 0.0.0.0
listen_port = 6432
auth_type = scram-sha-256
auth_file = /etc/pgbouncer/userlist.txt

; Transaction pooling: connections returned after each transaction
pool_mode = transaction
default_pool_size = 20
max_client_conn = 1000
max_db_connections = 50

; Timeouts
server_idle_timeout = 600
client_idle_timeout = 0
query_timeout = 30

Performance Tuning (postgresql.conf)

# Memory (adjust based on available RAM)
shared_buffers = '4GB'          # 25% of total RAM
effective_cache_size = '12GB'   # 75% of total RAM
work_mem = '64MB'               # Per-operation sort/hash memory
maintenance_work_mem = '1GB'    # For VACUUM, CREATE INDEX

# WAL
wal_buffers = '64MB'
checkpoint_completion_target = 0.9
max_wal_size = '4GB'

# Query planner
random_page_cost = 1.1          # SSD storage (default 4.0 is for HDD)
effective_io_concurrency = 200  # SSD

# Connections
max_connections = 200           # Use with PgBouncer in front

# Autovacuum (don't disable it, tune it)
autovacuum_max_workers = 4
autovacuum_naptime = '30s'
autovacuum_vacuum_cost_limit = 1000

Commands

# Backup (custom format, compressed, parallel)
pg_dump -Fc -j 4 -f backup.dump mydb

# Restore
pg_restore -d mydb -j 4 --clean --if-exists backup.dump

# Plain SQL dump (for version control or small DBs)
pg_dump --schema-only -f schema.sql mydb

# Check bloat and table sizes
psql -c "SELECT relname, pg_size_pretty(pg_total_relation_size(oid))
         FROM pg_class WHERE relkind = 'r' ORDER BY pg_total_relation_size(oid) DESC LIMIT 20;"

# Check active queries and locks
psql -c "SELECT pid, state, query_start, query FROM pg_stat_activity WHERE state != 'idle';"

# Run ANALYZE to update statistics after bulk load
psql -c "ANALYZE;"

# Check replication lag (if applicable)
psql -c "SELECT client_addr, state, sent_lsn, write_lsn, replay_lsn,
         pg_wal_lsn_diff(sent_lsn, replay_lsn) AS byte_lag
         FROM pg_stat_replication;"

Best Practices

DO

• Use timestamptz everywhere — timestamp silently drops timezone info and causes bugs across

regions.

• Use text over varchar(n) — PostgreSQL stores them identically, and varchar(n) just adds a

CHECK constraint that's painful to change later.

• Run EXPLAIN (ANALYZE, BUFFERS) on every slow query before adding indexes.
• Use CREATE INDEX CONCURRENTLY in production — regular CREATE INDEX locks the table for writes.
• Use transactions for multi-statement operations: BEGIN; ... COMMIT;.
• Run ANALYZE after bulk inserts or major data changes to keep the query planner accurate.
• Monitor with pg_stat_user_tables, pg_stat_user_indexes, and pg_stat_activity.

DON'T

• Don't use SERIAL — use GENERATED ALWAYS AS IDENTITY or uuid for primary keys.
• Don't create indexes on every column "just in case" — each index slows writes and consumes disk.
• Don't use SELECT * in application queries — fetch only the columns you need.
• Don't run VACUUM FULL routinely — it locks the table exclusively. Regular VACUUM (via

autovacuum) is sufficient.

• Don't store large blobs in PostgreSQL — use object storage (S3) and store the URL/reference.
• Don't use timestamp without timezone — timezone-unaware timestamps are a common source of

production bugs.

• Don't disable autovacuum — tune it instead. Disabled autovacuum leads to transaction ID wraparound

and table bloat.

• Don't connect directly from application servers without a connection pooler in production.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: dallay
• Source: dallay/agents-skills
• License: MIT

Install and usage instructions live in the source repository linked above.