Skip to main content

Fory v1.7.7 Released

· 3 min read
Shawn Yang
Shawn Yang
Apache Fory PMC Chair

Apache Fory 1.7.7 is now available. This release adds read-side JSON null handling and improves serialization performance, object mapping, and generated code. See Getting Started for installation instructions.

Highlights​

  • Configure how JSON readers handle null properties, collection elements, and map values.
  • Reduce JSON codec generation cost, improve string writing, and reduce C++ serialization overhead.
  • Fix JSON object field semantics, compatible field-type validation, and schema compiler output.
  • Make Java and Kotlin release archives reproducible and improve license and notice packaging.

JSON Null Handling​

Use onNullRead to control explicit null property values and onContentNullRead to control null array or collection elements and map values. Both default to SET, which retains the existing decoding behavior. SKIP omits that input occurrence; FAIL rejects it.

import org.apache.fory.json.ForyJson;
import org.apache.fory.json.annotation.JsonProperty;
import org.apache.fory.json.annotation.JsonProperty.NullHandling;

public final class Response {
  @JsonProperty(onNullRead = NullHandling.SKIP, onContentNullRead = NullHandling.SKIP)
  public java.util.List<String> seats = java.util.List.of("default");
}

ForyJson json = ForyJson.builder().build();
Response first = json.fromJson("{\"seats\":null}", Response.class);
Response second = json.fromJson("{\"seats\":[null,\"a\"]}", Response.class);
// first.seats is ["default"]; second.seats is ["a"].

The same settings are available on the builder. Property annotations override the corresponding global setting. Content overrides apply to immediate elements or values; nested containers use their applicable settings. These options affect explicit JSON nulls and do not change writing or a root value's null behavior.

Kotlin and Scala support these settings through their JSON builders. Skipping a null property can retain constructor defaults; required-property checks still apply. See read-side null handling, Kotlin, and Scala for details.

Correctness and Performance​

JSON object mapping now restores eligible ordinary final fields and consistently selects a subclass field when it hides a superclass field. Writing a subclass through a concrete base type includes the subclass's mapped properties; restoring the subtype still requires declared subtype mapping or a custom codec. See object mapping.

Schema compiler fixes cover Java, Kotlin, Scala, Python, and TypeScript output. Generated union carriers use an alternate name when a schema case is named Unknown or Swift's unknown, preserving the schema case name. See generated code.

Performance work reduces JSON codec generation cost, avoids unnecessary NON_EMPTY checks for final property types, improves escaped-string writing, and avoids a temporary vector when serializing C++ std::forward_list.

The complete release notes are available on GitHub.

Full Changelog: v1.7.6...v1.7.7