Incompatibility with Postgres Data Source

[rrowlands]

Hi,

I am having an issue with your library when I attempt to use it in conjunction with a Postgres data source.

The following very basic code:

final String url = DriverWrapper.POSTGIS_PROTOCOL + "//127.0.0.1:5432/postgres";
    
BaseDataSource ds = new PGSimpleDataSource();
ds.setUrl(url);
ds.setUser("postgres");
ds.setPassword("postgres");
Connection conn = ds.getConnection();
    
System.out.println(conn.toString());

Produces the error:

Exception in thread "main" java.lang.IllegalArgumentException: URL invalid jdbc:postgresql_postGIS://127.0.0.1:5443/postgres
	at org.postgresql.ds.common.BaseDataSource.setUrl(BaseDataSource.java:1367)
	at com.runwaysdk.Sandbox.main(Sandbox.java:76)

Due primarily to the fact that the Postgres BaseDataSource includes hardcoded assumptions that the jdbc protocol will always be 'postgresql'. See here and here.

Am I doing something wrong here? All of your test code appears to use the DriverManager.getConnection interface, which works because it bypasses that DataSource abstraction.

Additionally, I am unable to bypass your DriverWrapper, due to the fact that it is auto-loaded by the Java service loader paradigm. This causes issues because if I do not use the 'postgresql_postGIS' protocol, when DriverManager.getConnection is invoked via BaseDataSource.getConnection, the DriverWrapper throws a SQL exception Unknown protocol or subprotocol in url. This is already documented as an issue in ticket #113. As a result, I fear that my only path forward might be to repackage your jar without the META-INF service loader file, or to repackage the jar with a fix for #113 included.

Thanks

[phillipross]

Might this be related to the currently open PR (#115 ) ? I seem to recall there being a chain of related problems around this. But ultimately I think it was determined that the "driver wrapping" approach might be deficient for several reasons. It might very well be the case that using the datasource interface to produce wrapped/proxied connections might be better. This probably entails extracting the driver wrapping logic from the wrapper class(es) and using it in the datasource implementation class?

[rrowlands]

The changes contained within PR #115 would resolve ticket #113, which in turn would also resolve my blocker. In the PR, you mentioned that the tests were failing, however I can't quite tell why since it seems like there's some problem with logs. Can we please get that PR merged?

[phillipross]

right, i was hoping the original author of the PR could resolve the issues with the build but I'll have to take a look. it might just need a rebase or something simple 😉

[phillipross]

PR #115 has been debugged and merged, so I'll close this ticket out