JDBC: Update to pgjdbc 42.7.5

[amotl]

About

pgjdbc 42.7.5 has been released yesterday. When running the stock jdbc test cases, errors can be observed on Jenkins after bumping manually per this PR.

References

Validating an automatic Dependabot update failed over here on GHA, apparently tripping test_getFunctions, test_getColumns, test_getPrimaryKeys, and test_getTables in JdbcMetaDataTest.java.

• Bump org.postgresql:postgresql from 42.7.4 to 42.7.5 in /by-language/java-qa cratedb-examples#804

/cc @kneth, @seut

[kneth]

Issues encountered:

• In CrateDB indnkeyatts in pg_catalog.pg_index is always zero. This messes up the query (it has result.KEY_SEQ <= KEY_COUNT where KEY_COUNT is indnkeyatts in the WHERE clause). There will require a fix in the CrateDB server.
• In CrateDB SELECT current_database() returns a column with the name of the current database, while in PostgreSQL the name is current_database. It means that https://github.com/pgjdbc/pgjdbc/blob/master/pgjdbc/src/main/java/org/postgresql/jdbc/PgDatabaseMetaData.java#L1273 will fail as the column name is not specified in https://github.com/pgjdbc/pgjdbc/blob/master/pgjdbc/src/main/java/org/postgresql/jdbc/PgDatabaseMetaData.java#L1253 . This can be fixed by in the CrateDB server or in PG JDBC.

[amotl]

Thank you very much for your evaluations. Can you converge relevant flaws into corresponding tickets to improve CrateDB?

[hlcianfagna]

Regarding indnkeyatts, if crate/crate#17377 were addressed, perhaps we could submit a patch to
https://github.com/pgjdbc/pgjdbc/blob/b7bd9ffa25647ca0df76f7fa5b0883324ce79e45/pgjdbc/src/main/java/org/postgresql/jdbc/PgDatabaseMetaData.java#L2345 using

(
SELECT count(*)
FROM information_schema.key_column_usage
JOIN pg_catalog.pg_class ci2 ON ci2.relname = key_column_usage.constraint_name
WHERE ci2.oid = i.indexrelid
)

?

[amotl]

Hi. Because this patch seems to possibly converge well (thank you, @kneth)...

• Set column name explicitely when using current_database() in queries pgjdbc/pgjdbc#3526

and that issue has apparently been resolved (thank you, @hlcianfagna and @matriv)...

• Inconsistent naming of PK contraints between key_column_usage , pg_class and SHOW CREATE TABLE crate#17377

is there another patch to be staged to pgjdbc/pgjdbc, about the problems around indnkeyatts?

[hlcianfagna]

is there another patch to be staged to pgjdbc/pgjdbc, about the problems around indnkeyatts?

One option could be pgjdbc/pgjdbc#3533 , another option could be to raise a ticket in crate/crate to have indnkeyatts populated.

[kneth]

Dave Cramer says: "Please make sure this does not introduce a performance regression"

My question to @hlcianfagna, do you want me to try to evaluate the performance?

[hlcianfagna]

Dave Cramer says: "Please make sure this does not introduce a performance regression"

My question to @hlcianfagna, do you want me to try to evaluate the performance?

Let's maybe see what the timeline would be for crate/crate#17486 instead?

[kneth]

I have opened crate/crate#17616 as a fix to the issue with indnkeyatts