A Singer postgres target, for use with Singer streams generated by Singer taps.

• Creates SQL tables for Singer streams
• Denests objects flattening them into the parent object's table
• Denests rows into separate tables
• Adds columns and sub-tables as new fields are added to the stream JSON Schema
• Full stream replication via record version and ACTIVATE_VERSION messages.

1. Add libpq dependency

# macos
brew install postgresql

# ubuntu
sudo apt install libpq-dev

1. install singer-target-postgres

pip install singer-target-postgres

1. Follow the Singer.io Best Practices for setting up separate tap and target virtualenvs to avoid version conflicts.

2. Create a config file at ~/singer.io/target_postgres_config.json with postgres connection information and target postgres schema.

   {
     "postgres_host": "localhost",
     "postgres_port": 5432,
     "postgres_database": "my_analytics",
     "postgres_username": "myuser",
     "postgres_password": "1234",
     "postgres_schema": "mytapname"
   }

3. Run target-postgres against a Singer tap.

   ~/.virtualenvs/tap-something/bin/tap-something \
     | ~/.virtualenvs/target-postgres/bin/target-postgres \
       --config ~/singer.io/target_postgres_config.json >> state.json

   If you are running windows, the following is equivalent:

   venvs\tap-exchangeratesapi\Scripts\tap-exchangeratesapi.exe | ^
   venvs\target-postgresql\Scripts\target-postgres.exe ^
   --config target_postgres_config.json
   

The fields available to be specified in the config file are specified here.

target-postgres only supports JSON Schema Draft4. While declaring a schema is optional, any input schema which declares a version other than 4 will be rejected.

target-postgres supports all versions of PostgreSQL which are presently supported by the PostgreSQL Global Development Group. Our CI config defines all versions we are currently supporting.

The above is copied from the current list of versions on Postgresql.org

• Requires a JSON Schema for every stream.
• Only string, string with date-time format, integer, number, boolean, object, and array types with or without null are supported. Arrays can have any of the other types listed, including objects as types within items.
  • Example of JSON Schema types that work
    • ['number']
    • ['string']
    • ['string', 'null']
  • Exmaple of JSON Schema types that DO NOT work
    • ['string', 'integer']
    • ['integer', 'number']
    • ['any']
    • ['null']
• JSON Schema combinations such as anyOf and oneOf are not supported.
• JSON Schema $ref is partially supported:
  • NOTE: The following limitations are known to NOT fail gracefully
  • Presently you cannot have any circular or recursive $refs
  • $refs must be present within the schema:
    • URI's do not work
    • if the $ref is broken, the behaviour is considered unexpected
• Any values which are the string NULL will be streamed to PostgreSQL as the literal null
• Table names are restricted to:
  • 63 characters in length
  • can only be composed of _, lowercase letters, numbers, $
  • cannot start with $
  • ASCII characters
• Field/Column names are restricted to:
  • 63 characters in length
  • ASCII characters

If the add_upsert_indexes config option is enabled, which it is by default, target-postgres adds indexes on the tables it creates for its own queries to be more performant. Specifically, target-postgres automatically adds indexes to the _sdc_sequence column and the _sdc_level_<n>_id columns which are used heavily when inserting and upserting.

target-postgres doesn't have any facilities for adding other indexes to the managed tables, so if there are more indexes required, they should be added by another downstream tool, or can just be added by an administrator when necessary. Note that these indexes incur performance overhead to maintain as data is inserted, These indexes can also prevent target-postgres from dropping columns in the future if the schema of the table changes, in which case an administrator should drop the index so target-postgres is able to drop the columns it needs to.

Note: Index adding is new as of version 0.2.1, and target-postgres does not retroactively create indexes for tables it created before that time. If you want to add indexes to older tables target-postgres is loading data into, they should be added manually.

Singer.io requires official taps and targets to collect anonymous usage data. This data is only used in aggregate to report on individual tap/targets, as well as the Singer community at-large. IP addresses are recorded to detect unique tap/targets users but not shared with third-parties.

To disable anonymous data collection set disable_collection to true in the configuration JSON file.

target-postgres utilizes poetry for package management, and PyTest for testing.

See also:

• DECISIONS: A document containing high level explanations of various decisions and decision making paradigms. A good place to request more explanation/clarification on confusing things found herein.
• TableMetadata: A document detailing some of the metadata necessary for TargetPostgres to function correctly on the Remote

If you have Docker and Docker Compose installed, you can easily run the following to get a local env setup quickly.

$ docker-compose up -d --build
$ docker logs -tf target-postgres_target-postgres_1 # You container names might differ

As soon as you see INFO: Dev environment ready. you can shell into the container and start running test commands:

$ docker-compose exec target-postgres bash
(target-postgres) root@...:/code# pytest

The environment inside the docker container has a virtualenv set up and activated, with an --editable install of target-postgres inside it and your local code mounted as a Docker volume. If you make changes on your host and re-run pytest any changes should be reflected inside the container.

See the PyTest commands below!

To run the tests, you will need a PostgreSQL server running.

NOTE: Testing assumes that you've exposed the traditional port 5432.

Make sure to set the following env vars for PyTest:

$ EXPORT POSTGRES_HOST='<your-host-name>' # Most likely 'localhost'
$ EXPORT POSTGRES_DB='<your-db-name>'     # We use 'target_postgres_test'
$ EXPORT POSTGRES_USER='<your-user-name'  # Probably just 'postgres', make sure this user has no auth

To run tests, try:

$ poetry run pytest

If you've bash shelled into the Docker Compose container (see above), you should be able to simply use:

$ pytest

Join the conversation over at the Singer.io Slack and on the #target-postgres channel.

Try to adhere to the following for contributing:

• File New Issue -> Fork -> New Branch(If needed) -> Pull Request -> Approval -> Merge

Users can file an issue without submitting a pull request but be aware not all issues can or will be addressed.

Target Postgres is sponsored by Data Mill (Data Mill Services, LLC) datamill.co.

Data Mill helps organizations utilize modern data infrastructure and data science to power analytics, products, and services.

Copyright Data Mill Services, LLC 2018