Build an application on PLOMID

Any PostgreSQL client works — PLOMID speaks PG Frontend/Backend v3 natively. No plugins.

flowchart TB
    subgraph App["your application"]
        DRV["PG driver\n(psycopg / pg / JDBC)"]
        POOL["your connection pool"]
        DRV --- POOL
    end
    App -->|"postgresql://user:pass@host:5432/db"| SRV["plomid-server\nport 5432"]
    SRV --> ENG["engine: SQL → MVCC → storage"]
    ENG --> DISK["./data\nWAL + checkpoints"]

flowchart TB
    subgraph App["your application"]
        DRV["PG driver\n(psycopg / pg / JDBC)"]
        POOL["your connection pool"]
        DRV --- POOL
    end
    App -->|"postgresql://user:pass@host:5432/db"| SRV["plomid-server\nport 5432"]
    SRV --> ENG["engine: SQL → MVCC → storage"]
    ENG --> DISK["./data\nWAL + checkpoints"]

Python (psycopg)#

import psycopg
conn = psycopg.connect("postgresql://plomid:[email protected]:5432/plomid", autocommit=True)
with conn.cursor() as cur:
    cur.execute("SELECT id, name FROM users WHERE id = %s", (1,))
    print(cur.fetchall())
# transaction:
with conn.transaction():
    with conn.cursor() as cur:
        cur.execute("INSERT INTO orders VALUES (%s, %s, %s, now())", (201, 1, 19.99))

Node (pg)#

import { Client } from "pg";
const c = new Client({ connectionString: "postgresql://plomid:[email protected]:5432/plomid" });
await c.connect();
await c.query("BEGIN");
await c.query("INSERT INTO users VALUES ($1, $2)", [10, "grace"]);
await c.query("COMMIT");
await c.end();

Extended protocol (Parse/Bind/Describe/Execute/Sync, $1 params, WITH ORDINALITY table funcs, generate_series) is implemented in crates/network/src/server/handler.rs. Batchable INSERT … VALUES without RETURNING/ON CONFLICT is staged across Executes and flushed as one implicit transaction at Sync.

Sessions and errors#

• Session params: server_version 17.0, client_encoding UTF8, DateStyle ISO MDY, TimeZone UTC, search_path "$user",public. SQLSTATEs map to PG codes (28P01 auth, 08P01 protocol).
• On error in extended mode the connection enters InFailedTransaction until Sync.
• pg_catalog + information_schema expose live catalog rows for tool introspection (see pg_catalog).

Checklist#

• Use parameterized queries; never interpolate.
• Keep transactions short; handle Conflict by retrying the whole transaction.
• Handle empty results (zero rows) distinctly from NULL.
• Pool at the app layer; server limit is max_connections=100000 with graceful reject.

Next: Transactions · Errors