LocoPostgres is a JDBC driver for PostgreSQL (and also compatible with amazon Redshift).

The only dependency that LocoPostgres has while running is Log4j, it uses other dependencies but just for testing.

Should you use LocoPostgres in production? No, period. I wrote this driver mainly out of curiosity, during my job i became quite familiar with the PostgreSQL protocol and decided that it would be interesting to write a JDBC driver just for fun. As there are no external dependencies, the driver is a good way to understand the PostgreSQL protocol. Also the driver logs every single packet that is sent and received from the server.

To use loco postgres you will first have to register the driver. To do so do

/tmp/locopostgres-debug.log

LocoPostgres is a JDBC driver for PosgreSQL database. It also works with AWS Redshift, as Redshift uses the same protocol.

You should not! If you want to connect to PostgreSQL or Redshift you should either use the official PostgreSQL driver or the RedshiftJDBC driver (in the case of Redshift). LocoPostgres is, by no means, a replacement for those.

Mainly due to curiosity, but also literally because I can :).

As part of my job I ended up learning the PostgreSQL protocol and I wonder whether I could write a JDBC driver just for fun.

I have decided to use NO dependencies in this code except Log4j. This means that I am not even using some of the basic Apache Commons libraries. The reason for this is that I want to simply write everything from scratch. Again, this is just for fun, it is not practical.

I have allowed myself to use libraries for the unit tests, but not for the main driver.

The other interesting part of this driver is that I have tried to make the code easy to read, I have added plenty of comments around. Also the driver spits out the protocol packets into the logs, this makes this driver very convenient if you want to simply study the protocol.

To establish a connection you must use a url that looks like jdbc:loco:postgresql instead of jdbc:postgresql, this is actually not compliant with the actual PostgreSQL protocol, and it is done on purpose. Essentially I am trying to prevent anybody from using LocoPostgres in a production environment by mistake.

• LocoDriver: The class implementing Driver, it is in charge of establishing the connection to the server.
• Packet: The main abstraction representing network packets being sent and received from the server.
• LocoNetwork: The object in charge of opening the actual TCP Connection to the server to send and receive network packets. This is done with a Java Socket.
• LocoStatement: The object implementing the Statement interface.
• LocoResultSet: The object implementing the ResultSet interface.
• LocoConnection: The object implementing the Connection interface.

Next steps:

• [DONE] Implement unit tests using the actual PostgreSQL JDBC driver as a reference or valid way to get data from the database and compare the results with LocoPostgres

• [DONE] Improve documentation (specially javadocs)

• [DONE] Implement the read of the result set

• [DONE] Implement batch methods for statement

• [DONE] Implement query execution cancel

• [DONE] Pass all the unit tests

  • Regression, as more tests are added
  • Provided a base method so we can have tests with actual data in a reusable manner

• [WIP] read operations.

  • Identified issues when reading ts tz ranges
  • [DONE] Identified issue while reading arrays of bytes

• Implement authentication methods other than md5

  • [DONE] Implement clear text password
  • Implement SCRAM SHA 256

• Implement prepared statements, and the parse-bind-execute packets in the postgres protocol

  • [DONE] Prepared statement, no parameters
  • Prepared statement parameters
    • int, float, double, short
    • strings
    • boolean

• Fully implement the Statement interface