Bulk-load data into PostgreSQL fast with COPY and psql \copy

Documented steps

1. Server-side load (file readable by the server process): COPY table_name (col1, col2) FROM '/absolute/path/file.csv' WITH (FORMAT csv, HEADER true);
2. Client-side load from your own machine, in psql: \copy table_name (col1, col2) FROM 'local/file.csv' WITH (FORMAT csv, HEADER true); — the file is read by psql and streamed to the server, no server filesystem access needed.
3. Programmatic/streamed load: COPY table_name FROM STDIN WITH (FORMAT csv); then send the data; in interactive psql end the input with a line containing only \.
4. Control NULL handling explicitly, e.g. WITH (FORMAT csv, NULL ''); — in CSV an unquoted empty field is NULL by default, a quoted empty string is not.
5. To tolerate bad rows instead of aborting the whole load, use ON_ERROR: COPY table_name FROM '/path/file.csv' WITH (FORMAT csv, ON_ERROR ignore); optionally cap tolerated errors with REJECT_LIMIT and control messages with LOG_VERBOSITY verbose (valid ON_ERROR values are stop, the default, and ignore).
6. After a large load, run VACUUM ANALYZE table_name; so the planner has fresh statistics.
7. Official docs: https://www.postgresql.org/docs/current/sql-copy.html