Fast, Enjoyable & Secure NoSQL Database

Dart CI codecov Pub Version License

PubStats Popularity PubStats Rank PubStats Dependents

Hive is a lightweight and blazing fast key-value database written in pure Dart. Inspired by Bitcask.

Documentation & Samples 📖

Features

  • 🚀 Cross platform: mobile, desktop, browser
  • ⚡ Great performance (see benchmarks)
  • ❤️ Simple, powerful, & intuitive API
  • 🔒 Strong encryption built in
  • 🎈 NO native dependencies
  • 🔋 Batteries included

New features in Hive CE

Hive CE is a spiritual continuation of Hive v2 with the following new features:

  • Tooling
    • Hive CE Inspector DevTools extension to browse and search the contents of your boxes
  • Platforms
    • Isolate support through IsolatedHive
    • Flutter web WASM support
  • Code generation
    • Generate adapters with the GenerateAdapters annotation instead of annotating every type and field
    • Generate adapters for classes outside the current package
    • Register every generated adapter in one call with the HiveRegistrar extension
    • Support for Freezed, constructor parameter defaults, and named imports
  • Types
    • Built-in support for Set and Duration
    • DateTime and Duration keep microsecond precision
    • Type IDs up to 65439 instead of 223
  • Performance
    • Faster batch writes and opening of large boxes, including encrypted boxes
    • Lazy boxes open several times faster
    • IsolatedHive individual writes are faster than Hive CE
    • See the benchmarks

Getting Started

Hive CE requires Dart 3. Ensure that you have the following in your pubspec.yaml file:

environment:
  sdk: ^3.0.0

Usage

You can use Hive just like a map. Writes are readable immediately, but a write can be lost if the process is killed before its future completes. On the VM, flush() waits for pending writes and flushes them to disk.

import 'package:hive_ce/hive_ce.dart';

void example() {
  final box = Hive.box('myBox');
  box.put('name', 'David');
  final name = box.get('name');
  print('Name: $name');
}

Benchmarks

The benchmarks were run on an M3 Max MacBook Pro. You can see the benchmark code here.

Speed

All times are averaged over several runs.

Individual writes

Writing values one at a time, then flushing once. Hive v4, sqflite, drift, and ObjectBox commit each write. Runs that took longer than 5 minutes were stopped.

Database 1,000 10,000 100,000 1,000,000
Hive CE 14 ms 144 ms 1.45 s 14.92 s
IsolatedHive 11 ms 119 ms 1.20 s 12.65 s
Hive v2 15 ms 144 ms 1.50 s 20.23 s
Hive v4 129 ms 1.60 s 17.82 s 228.49 s
sqflite 1.70 s 14.26 s 94.92 s > 5 min
drift 266 ms 3.63 s 44.00 s > 5 min
sembast 62 ms 642 ms 7.45 s 72.72 s
ObjectBox 4.11 s 41.97 s > 5 min > 5 min

Batch writes

Writing all values with a single addAll, then flushing once:

Database 1,000 10,000 100,000 1,000,000
Hive CE 0 ms 9 ms 101 ms 1.10 s
IsolatedHive 1 ms 10 ms 126 ms 1.47 s
Hive v2 0 ms 8 ms 111 ms 3.10 s
Hive v4 2 ms 26 ms 257 ms 2.88 s
sqflite 7 ms 56 ms 581 ms 6.48 s
drift 5 ms 40 ms 443 ms 4.56 s
sembast 17 ms 169 ms 1.71 s 17.43 s
ObjectBox 4 ms 6 ms 29 ms 269 ms

Opening a database

Opening a database and reading every value:

Database 1,000 10,000 100,000 1,000,000
Hive CE 1 ms 9 ms 114 ms 1.25 s
IsolatedHive 1 ms 13 ms 134 ms 1.41 s
Hive v2 1 ms 11 ms 145 ms 3.68 s
Hive v4 1 ms 12 ms 154 ms 1.61 s
sqflite 0 ms 7 ms 95 ms 1.04 s
drift 1 ms 13 ms 166 ms 1.86 s
sembast 2 ms 31 ms 343 ms 3.87 s
ObjectBox 0 ms 1 ms 9 ms 103 ms

Encrypted and lazy boxes

Hive v2 compared to Hive CE. Lazy boxes only load keys when opened, so the lazy open time doesn't include reading values.

Values Encrypted batch write Encrypted open Lazy open
10,000 40 ms → 23 ms 27 ms → 20 ms 6 ms → 5 ms
100,000 413 ms → 256 ms 304 ms → 211 ms 84 ms → 59 ms
1,000,000 6.06 s → 2.53 s 5.51 s → 2.17 s 3.30 s → 634 ms

Database size

Database 1,000 10,000 100,000 1,000,000
Hive CE 0.11 MB 1.10 MB 10.97 MB 109.67 MB
Hive v4 1.00 MB 5.00 MB 30.00 MB 290.00 MB
sqflite 0.04 MB 0.25 MB 2.49 MB 24.94 MB
drift 0.03 MB 0.24 MB 2.40 MB 24.02 MB
sembast 0.24 MB 2.41 MB 24.21 MB 243.08 MB
ObjectBox 0.24 MB 1.51 MB 14.16 MB 141.02 MB

Database size in Hive v4 is directly affected by the length of field names in model classes.

Memory usage

Peak memory usage with 1,000,000 values, including native memory, averaged over three runs. Each run is a new process. Hive CE and Hive v2 keep every value of a regular box in memory while it is open. Lazy boxes only keep keys in memory and read values with getAt.

Database Batch write Open Open and read
Hive CE 595 MB 471 MB 486 MB
Hive CE (lazy) 501 MB 241 MB 369 MB
Hive v2 595 MB 469 MB 485 MB
Hive v4 481 MB 1 MB 1,396 MB
sqflite 1,673 MB 1 MB 558 MB
drift 644 MB 1 MB 784 MB
sembast 1,790 MB 1,128 MB 1,400 MB
ObjectBox 292 MB 1 MB 272 MB

Libraries

hive
hive_ce