JsonGUI

A fast, native desktop app for exploring and searching large JSON files. Built with Tauri 2 + React 19 + TypeScript.

---

Features

• Open and explore JSON files of any size (lazy tree loading)
• Full-text search across keys, values, or both — powered by Rayon for parallel execution
• Object search with multiple property filters, nested property paths, per-row apply, and apply-all
• Property path autocomplete in object search
• Right-click context menu: copy JSONPath, value, or raw JSON subtree
• Keyboard navigation in the tree (Arrow keys, Enter)
• Recent files list (last 5, persisted in localStorage)
• Search filters persisted per file in localStorage
• Drag and drop a JSON file from Finder/Explorer directly onto the window
• Status bar with node count, file size, and path
• Dark UI with Tailwind CSS

---

Requirements

• macOS 12+ (Apple Silicon or Intel) — primary target
• Rust stable toolchain
• Node.js 22+
• npm 10+

---

Install

npm install

---

Usage

Launch in development mode (hot-reload):

npm run tauri:dev

Build a release bundle:

npm run tauri build

On macOS, npm run tauri build now creates the .app with Tauri and then generates the .dmg from an external staging directory to avoid the create-dmg temporary-image recursion issue.

---

Search

JsonGUI provides two search modes: Text and Objects.

Text search

Use text search when you want to scan the whole document for:

• keys only
• values only
• both keys and values

Available options:

• case sensitive
• regex
• exact match
• scope path, to limit the search to a subtree such as $.users.0
• sort by relevance or file order

Object search

Use object search when you want to find objects matching one or more property conditions.

Each row contains:

• an enable/disable checkbox
• a property path
• an operator
• a value, when required
• an Apply button for that single row

You can also use Apply all to run all enabled rows together as an AND query.

Supported operators:

• contains
• equals
• regex
• exists

When exists is selected, the value field is hidden because no comparison value is needed.

Nested property paths

Property paths can target nested data using dot notation.

Examples:

• marketing_lingua
• content.mainImage
• content.mainImage.0.url
• product.details.title

In other words, you can use key.key for nested object lookups, and continue deeper as needed. Array indexes can be included as path segments as well.

Object search options

Object search also supports:

• separate case sensitivity for property keys and property values
• path autocomplete based on existing keys in the current file
• per-file filter persistence in localStorage
• scope path limitation, so object search can run inside a specific subtree only

Typical examples

Find objects where a nested URL contains a domain:

content.mainImage.0.url contains example.com

Find objects where a property exists:

marketing_lingua exists

Find objects that match multiple conditions:

marketing_lingua contains Acciaio
finish equals Lucido

---

Development

# TypeScript type check only
npx tsc --noEmit

# Rust check only
cd src-tauri && cargo check

# Rust tests
cd src-tauri && cargo test

---

Benchmarks

Rust (Criterion)

Measures parsing, BFS + DTO build, IPC serialization and build_raw on synthetic data:

cd src-tauri && cargo bench
# HTML reports: src-tauri/target/criterion/

To benchmark against a real JSON file, create .bench.env in the project root (already in .gitignore) with the path to your large file:

cp .bench.env.example .bench.env
# edit .bench.env:
# BENCH_JSON_PATH=/path/to/large.json

cd src-tauri && cargo bench -- real_file

GitHub Actions perf smoke test

A lightweight CI workflow is available in /.github/workflows/performance.yml. It generates a deterministic JSON dataset of about 64 MiB, then measures:

• load via JsonIndex::from_file
• search_text over values
• search_regex over values
• search_objects with a nested property path filter
• expand_all via the same BFS helper used by the Rust benchmarks
• structural memory metrics from the built index (heap_bytes_estimate, bytes per node, bytes per input byte)

Each run publishes:

• a Markdown summary directly in the GitHub Actions job summary
• a rust-perf-smoke artifact containing perf-ci.json and perf-ci-summary.md

This is meant to track trends over time from the Actions panel, not to act as a strict pass/fail performance budget: GitHub-hosted runners are noisy.

For tagged releases, the same benchmark snapshot is also appended to the draft release notes by /.github/workflows/release.yml, so the release page contains the measured load, search_text, search_regex, search_objects, expand_all, and structural memory numbers for that version.

You can run the same smoke test locally with:

cd src-tauri
cargo run --release --example perf_ci -- \
  --size-mib 64 \
  --iterations 3 \
  --sample-path target/perf-ci-sample.json \
  --output target/perf-ci.json

JavaScript (frontend)

Measures buildVisibleNodes, Map copy cost, and streaming overhead on synthetic trees:

npx tsx src/bench/perf.mts

Architecture

Frontend (React + Zustand)
  TreeNode (lazy expand)   SearchPanel
          |                     |
          +------Tauri IPC------+
                     |
Backend (Rust)
  JsonIndex (arena Vec<Node>)
  SearchEngine (rayon parallel)
  FileLoader (sonic-rs SIMD parser)

---

License

MIT