Why Jsonnet Is the Ultimate Tool for Flexible JSON Generation and Transformation

Background

In recent projects we needed a flexible tool for generating and transforming JSON with high performance and flexibility. Existing custom JSONPath solutions were insufficient, leading us to evaluate jsonpath, jsonpatch, jsonata, and ultimately choose Jsonnet.

Jsonnet Overview

Jsonnet is a Google‑open‑source, Turing‑complete configuration language. It is a superset of JSON, can generate JSON, YAML, INI, and other formats, and is widely used in Kubernetes, Grafana, and similar systems. Implementations exist for C++, Rust, Go, and Java.

Implementation Analysis

The sjsonnet implementation consists of four main components: Parser, StaticOptimizer, Evaluator, and Materializer. The Parser converts Jsonnet source text into an abstract syntax tree (AST). The StaticOptimizer performs static optimizations such as branch pruning, constant folding, and variable substitution, caching the optimized AST for reuse. The Evaluator combines input parameters (extVar and tlaVar) with the optimized AST to produce a value. The Materializer converts the resulting value into the desired output format (JSON, YAML, markdown, etc.).

Key characteristics include interpreter mode, native output support, library usage, and multi‑language implementations (CPP, Go, Rust, Scala). The project is maintained by Google and community contributors.

Usage Guide

Basic Usage

Examples demonstrate common scenarios: extracting specific fields, removing unwanted keys, replacing keys, performing field calculations, combining Jsonnet with JsonPath for selective extraction, and generating markdown output. Java code snippets show how to read Jsonnet files, provide external variables, and invoke the interpreter.

{
  person1: {
    name: "Alice",
    welcome: "Hello " + self.name + "!"
  },
  person2: self.person1 { name: "Bob" }
}

Advanced Usage

Advanced examples cover outputting markdown directly, defining custom functions (e.g., a UUID generator) in Java/Scala, and invoking them from Jsonnet scripts.

local uuid = tb.random.uuid;
uuid()

Performance Optimization

Recommendations for further speed improvements include using ExternalVariable with pre‑parsed Expr to avoid repeated parsing, leveraging the parser cache, implementing custom extensions in Java for better performance, pre‑warming frequently used scripts at application startup, and employing optimizer techniques such as tableSwitch, lazy evaluation, and static caching. A benchmark chart compares jsonnet, sjsonnet, and jrsonnet performance.

Conclusion

Jsonnet provides a powerful solution for complex JSON generation and transformation needs, offering good performance, extensibility, and broad ecosystem support. Integrating it into JVM projects simplifies configuration handling, and contributions back to the upstream project have improved performance and fixed issues.