pg-dump-upsert 🐘💩

Simple tool to dump a Postgresql table as INSERT statements with ON CONFLICT clause (also known as "upsert" statements).

Pre-alpha software. Expect crashes, data loss, silent data corruption etc.

Rationale

The pg_dump command can dump tables as INSERT statements however you can't directly restore such dumps if the database has conflicting rows. Furthermore pg_dump is doing more work than simply querying the data and this sometimes causes seemingly unrelated failures.

Installation

$ go install github.com/tomyl/pg-dump-upsert@latest
$ pg-dump-upsert -h
Usage of pg-dump-upsert:
  -conflict-column string
        Append an ON CONFLICT clause for this column. All other columns will be included in a DO UPDATE SET list.
  -dsn string
        Connection string. Example: postgres://user:password@host:5432/db
  -from string
        Source table to dump.
  -insert-columns string
        Comma-separated list of columns to include in INSERT statement. Defaults to all columns.
  -noconflict
        Append ON CONFLICT DO NOTHING.
  -query string
        Use custom SELECT query. By default fetches all rows. Note that column order must match -insert-columns. It is also valid to just specify a WHERE clause. It will be appended to the default query.
  -to string
        Table name to use in INSERT statements. Defaults to the source table.
  -tx   
        Wrap INSERT statements in transaction.
  -verbose
        Log query statement to stderr.

Examples

Dump all rows in table employee:

$ pg-dump-upsert -dsn "postgres://user:password@host:5432/db" -from employee 
INSERT INTO employee (id, created_at, name, salary) VALUES (1, '2018-06-13 21:10:34.769555+08', 'Jane Doe', 123456);
...

Choose which columns to dump:

$ pg-dump-upsert -dsn "postgres://user:password@host:5432/db" -from employee -insert-columns id,name
INSERT INTO employee (id, name) VALUES (1, 'Jane Doe');
...

Ignore conflicts:

$ pg-dump-upsert -dsn "postgres://user:password@host:5432/db" -from employee -noconflict
INSERT INTO employee (id, created_at, name, salary) VALUES (1, '2018-06-13 21:10:34.769555+08', 'Jane Doe' 123456) ON CONFLICT DO NOTHING;
...

Update columns on conflict:

$ pg-dump-upsert -dsn "postgres://user:password@host:5432/db" -from employee -conflict-column id
INSERT INTO employee (id, created_at, name, salary) VALUES (1, '2018-06-13 21:10:34.769555+08', 'Jane Doe', 123456) ON CONFLICT (id) DO UPDATE SET created_at=EXCLUDED.created_at, name=EXCLUDED.name;
...

Fetch a subset of the rows:

$ pg-dump-upsert -dsn "postgres://user:password@host:5432/db" -from employee -query "WHERE salary > 12345"
INSERT INTO employee (id, created_at, name, salary) VALUES (1, '2018-06-13 21:10:34.769555+08', 'Jane Doe', 123456);
...

Use a different table name in the INSERT statements:

$ pg-dump-upsert -dsn "postgres://user:password@host:5432/db" -from employee -to minions
INSERT INTO minions (id, created_at, name, salary) VALUES (1, '2018-06-13 21:10:34.769555+08', 'Jane Doe', 123456);
...

To restore a dump, simply use the \i command in psql.

TODO

Implement support for all Postgres data types.
Skip generated columns when dumping views.
Allow which columns to update when specifying -conflict-column?
More unit tests would be nice...
Finish this TODO list.