feat: all statement batching

[libingye816]

Support batching of all statement types. Response messages are sent for each statement in the batch.
Cover most of the cases (not only the short-term solution to python driver) with the exception:

1. The remaining statements in a dml/ddl batch are skipped and throw exceptions if the dml/ddl batch fails in the middle. The preceding statements in dml/ddl batch may not take effect even though they succeed. The whole batch resumes after the dml/ddl batch.

[olavloite]

Thanks, this looks great! I've temporarily stopped reviewing at line 730 of JdbcSimpleModeMockServerTest.java, but I'll try to continue reviewing later today to give you a full review of everything.

Sorry for the confusion regarding error handling of implicit transactions. It seems that I got confused by the current implementation of batching when writing the first part of the design doc. The rule for what to do in case of an error in a batch is actually a lot simpler than I thought:

1. Always continue executing statements in the batch.
2. In case of implicit transaction: Execute each statement in the batch using auto-commit. This means that each statement that succeeds should be committed, regardless of whether other statements fail.

[libingye816]

@olavloite Do you have an idea why clirr and e2e-psql-xxx are failing?

[olavloite]

The e2e psql tests are failing because some of those tests check whether they get an expected error when executing a mixed batch. That is no longer the case, as this PR introduces support for that. The psql tests work with text files that contain the expected output. If you take a look at the original draft PR, you will see that I changed the expected output of some of those (see for example dml-ddl-batching.txt: https://github.com/GoogleCloudPlatform/pgadapter/pull/30/files)

Clirr is failing because this PR introduces breaking changes in the public API of PGAdapter. I think we should disable clirr for this repository. It does not really make very much sense, as this is not a library that is imported as a dependency into other applications.

EDIT: I've removed the clirr check as a required check for this repo. So it is still there and will still fail for this PR, but it won't block merging this PR.

[libingye816]

The e2e psql tests are failing because some of those tests check whether they get an expected error when executing a mixed batch. That is no longer the case, as this PR introduces support for that. The psql tests work with text files that contain the expected output. If you take a look at the original draft PR, you will see that I changed the expected output of some of those (see for example dml-ddl-batching.txt: https://github.com/GoogleCloudPlatform/pgadapter/pull/30/files)

I have already updated those expected output files. All tests were passed at last Friday, so I am surprised that they fail again. The psql receives error while expecting some outputs. Looks like the batch support doesn't take effect.

[libingye816]

I fixed the logic of statement split to ignore escaped quotes. It should work now. @olavloite

[olavloite]

Thanks! This looks good to me. I have some TODOs / nits in my comments, but feel free to defer those to a follow-up PR.

[libingye816]

Hi @olavloite Can you give a last round of quick review?