Decapitalization of property names in repository queries breaks for properties with single lowercase letter followed by an uppercase letter [DATACMNS-1589]

[spring-projects-issues]

Andre Kupka opened DATACMNS-1589 and commented

The property resolution in custom repository queries fails if the property name starts with a single lowercase letter followed by an uppercase letter. The following entity and repository demonstrate the problem:

Entity

 

@Entity
@Table(name="customer")
public class Customer {

    @Id
    @Column(name = "id")
    private String id;

    @Column(name = "c_login_id", nullable = false)
    private String cLoginId;

    public String getId() {
        return id;
    }

    public void setId(String id) {
        this.id = id;
    }

    public String getCLoginId() {
        return cLoginId;
    }

    public void setCLoginId(String cLoginId) {
        this.cLoginId = cLoginId;
    }
}

Repository

 

@Repository
public interface CustomerRepository extends JpaRepository<Customer, String> {

    Optional<Customer> findByCLoginId(String cLoginId);
}

 

 

In Spring Boot version 2.1.8.RELEASE (includes spring-data-commons 2.1.10.RELEASE) the code above was working but in version 2.1.9.RELEASE (includes spring-data-commons 2.1.11.RELEASE) the following exception was thrown:

 

Caused by: java.lang.IllegalArgumentException: Unable to locate Attribute  with the the given name [CLoginId] on this ManagedType [de.andrekupka.demo.propertyresolution.Customer]
	at org.hibernate.metamodel.internal.AbstractManagedType.checkNotNull(AbstractManagedType.java:128) ~[hibernate-core-5.3.12.Final.jar:5.3.12.Final]
	at org.hibernate.metamodel.internal.AbstractManagedType.getAttribute(AbstractManagedType.java:113) ~[hibernate-core-5.3.12.Final.jar:5.3.12.Final]
	at org.springframework.data.jpa.repository.query.QueryUtils.toExpressionRecursively(QueryUtils.java:633) ~[spring-data-jpa-2.1.11.RELEASE.jar:2.1.11.RELEASE]

 

 

I have implemented a minimal working example in https://github.com/andrekupka/jpa-property-name-resolution to reproduce the behavior and tagged a working, a broken and two workaround versions:

Working version: https://github.com/andrekupka/jpa-property-name-resolution/tree/working/2.1.8

Broken version: https://github.com/andrekupka/jpa-property-name-resolution/tree/not-working/2.1.9

Workaround with a custom Query annotation: https://github.com/andrekupka/jpa-property-name-resolution/tree/fix/query

Workaround by renaming the query method: https://github.com/andrekupka/jpa-property-name-resolution/tree/fix/property-renaming

The Testcase CustomerRepositoryTest can be used to reproduce the behavior.

 

The regression was introduced with PR #402 which uses https://docs.oracle.com/javase/7/docs/api/java/beans/Introspector.html#decapitalize(java.lang.String) for property decapitalization instead of Spring's own StringUtils class.

Tthe documentation at https://docs.spring.io/spring-data/data-commons/docs/current/reference/html/#repositories.query-methods.query-property-expressions does not mention the changed behavior as well. The documentation states that the first character of the property is decapitalized therefore the former example should still work.

Solution Proposal

I think using the decapitalization from the Java Bean Introspector is the correct behavior, thus the documentation should be updated and explicitly mention this special case.

 

 

---

Affects: 2.1.11 (Lovelace SR11)

Issue Links:

• DATACMNS-1570 Parsing of query method names is inconsistent for attributes beginning with a single lower case character followed by an upper case character

[spring-projects-issues]

Oliver Drotbohm commented

Looks like you're running into a side-effect of the fix for DATACMNS-1570. It might be surprising – it definitely is to me – but according to the Java Beans specification the expected getter method name is getcLoginId.

The documentation states that the first character of the property is decapitalized therefore the former example should still work.

Actually, I can't find that in the section you linked. We don't state any details except "uncapitalized" which I guess we can improve on by replacing it with "converted ac cording to the Java Beans specification"

[spring-projects-issues]

Andre Kupka commented

I was referring to the statement "... checks the domain class for a property with that name (uncapitalized).". I have interpreted the word "uncapitalized" as "the first character of the property is decapitalized" which was the original behavior as well.

getcLoginId feels some kind of weird although adhering to the Java Beans specification is the correct way IMO.

I would appreciate if you would explicitly mention this detail in the linked documentation, as this would have helped a lot in the first place

[spring-projects-issues]

Jens Schauder commented

As described by Oliver the current behavior is intentionally consistent with the Java Beans Specification. We are actually using java.beans.Introspector

I'm therefore closing this issue