Added buffer size for used BufferedOutputStream

[kowalczm]

Upgraded postgrsql jdbc driver to 42.2.2

[bytefish]

Thanks for the Pull Request! 🙏 I will need some time to review it. Do you have any figures on performance improvements?

[kowalczm]

I have played with 'IntegrationTest.java' and see that having configrable buffer size we can control how many calls are done for Postgres Db and see significant performance improvment when sending data. So for larger buffer size than default in PGCopyOutputStream (65536) we are deacresing calls to db. Additionality byte array copying from BufferedOutputStream to PGCopyOutputStream is removed when BufferedOutputStream buffer size is less that PGCopyOutputStream buffer size.

[bytefish]

Sorry for my very late reply. I am now reviewing it.

[bytefish]

@kowalczm Could you explain, why you are setting the Buffer Size to 1 for the PGCopyOutputStream? Maybe I am not understanding Java good enough.

new DataOutputStream(new BufferedOutputStream(new PGCopyOutputStream(copyIn, 1), bufferSize));

[bytefish]

@kowalczm Ah I get it. The PgCopyOutputStream has some range checks and if the buffer size is 1, then it will always write as soon as the BufferedOutputStream flushes. Nice one. 👍

[bytefish]

Yes, I have read through the Postgres JDBC Driver code. I see two options here...

1. Remove the BufferedOutputStream, because the PGCopyOutputStream is already buffered.
2. Set the PGCopyOutputStream buffer size to 1 and use a BufferedOutputStream.

If one uses the 2. approach, then at the end of the day the PGCopyOutputStream simply dispatches writes through to the CopyIn. This will indeed stop any Memory Copying Operations introduced by me due to a small Buffer Size in the BufferedOutputStream.

I have opted to go for the 1. option, that you have used as well. If it mysteriously breaks, then we will use the 2. approach. The PgBulkInsert class now accepts a IConfiguration to adjust the Buffer Size.

[bytefish]

Released Version 2.1 with a Configurable Buffer Size. Thanks! 👍