`Converter<String, …>` is applied to derived query

[genuss]

Hi, maintainers!
I started my upgrade to spring data jdbc 3.2.0 and found an issue with my setup. I've got a postgresql database and some tables with jsonb type which are converted to simple strings in my entities. To achieve this, I created a pair of Converters. The code is pretty straitforward:

@ReadingConverter
class StringJsonbReadingConverter implements Converter<PGobject, String> {

  @Override
  public String convert(@NonNull PGobject source) {
    if (source.getValue() == null) {
      return null;
    }
    return source.getValue();
  }
}

@WritingConverter
class StringJsonbWritingConverter implements Converter<String, PGobject> {

  @Override
  public PGobject convert(@NonNull String source) {
    var pGobject = new PGobject();
    pGobject.setType("jsonb");

    try {
      pGobject.setValue(source);
    } catch (SQLException e) {
      throw new IllegalStateException(e);
    }

    return pGobject;
  }
}

With them, I could use Strings in my entities both as varchar or jsonb up until 3.1.6 but it broke with 3.2.0. To better illustrate it I created a reproducer here.

I'm not sure whether it's a regression, or I just misused an existing API. Could you help me out here?

[mp911de]

This is expected behavior and should have had never worked otherwise. By specifying a Converter<String, PGobject>, all String properties are using the converter regardless.

Derived queries have an association to the property that is being queried. You can bypass the converter on the query side by providing a custom query, such as @Query("SELECT * FROM test_entity where text_column = :textColumn").

[schauder]

While this converter setup is indeed "interesting", I think this is actually a duplicate of #1681

[genuss]

Indeed my issue is related to #1681, and your fix solves my issue as well.
However, I suppose it's better to stick with a wrapper and and a pair of Converters in order to avoid surprises in future.

record JsonbValue(String value) {}

@ReadingConverter
class JsonbValueReadingConverter implements Converter<PGobject, JsonbValue> {
  @Override
  public JsonbValue convert(@NonNull PGobject source) {
    if (source.getValue() == null) {
      return null;
    }
    return new JsonbValue(source.getValue());
  }
}

@WritingConverter
class JsonbValueWritingConverter implements Converter<JsonbValue, PGobject> {
  @Override
  public PGobject convert(@NonNull JsonbValue source) {
    var pGobject = new PGobject();
    pGobject.setType("jsonb");
    try {
      pGobject.setValue(source.value());
    } catch (SQLException e) {
      throw new IllegalStateException(e);
    }
    return pGobject;
  }
}

This should be more robust and it doesn't require adding stringtype=unspecified to JDBC driver parameters.

Thank you for the clarification!