登录

nippy

taoenssoEPL-1.0活跃维护

Clojure 的快速序列化库，支持压缩、加密和 EDN 格式，用来高效存取 Clojure 数据结构。

Fast serialization library for Clojure

v3.9.0发布于 2026-09-06 11:11 UTC1,116star62fork同步于 2026-11-07 07:33 UTC
序列化Clojure压缩加密EDN
对比 GitHub 官网clojurecompressionednencryptionepl

怎么上手

由 README 提炼，执行前请确认
原文 
(require '[taoensso.nippy :as nippy])
(nippy/thaw (nippy/freeze <my-value>))
硬件要求按 README 说明

README 未说明

下载：源码 zip·GitHub Releases（安装包）本站不镜像安装包

项目说明

由 README 自动翻译提炼
没人工点评？写下你的经验

纯 Clojure 的高性能序列化库，替代慢速 reader 把数据快速写入数据库

Clojure 的快速序列化库，支持压缩、加密和 EDN 格式，用来高效存取 Clojure 数据结构。

核心能力

  • 用 freeze/thaw 一对函数完成任意 Clojure 值的序列化与还原
  • 覆盖全部标准数据类型，并支持扩展自定义类型
  • 可选 LZ4 或 Zstandard 智能压缩
  • 可选 AES128 加密
  • 提供 thaw 转换器，便于检查与变换反序列化数据

容易踩的坑

压缩、加密、Java Serializable 回退和 Reader 回退都是可选项，需要按需开启；滚动部署时要按版本逐步升级（如 v3.5→v3.6→v3.7），先部署读支持再写入；Nippy 不保证字节输出稳定，不要依赖具体字节结果

内容由模型根据 README 生成，可能不完整或不准确，以官方文档为准。看原文 README

最近发版

5 个版本v3.9.0
全部 release
  1. v3.9.0v3.9.0 (2026-09-06)最新正式版2026-09-06 · 2026-09-06 11:11 UTC

    This promotes v3.9.0-RC1 to final with a small fix and a few extra optimizations.

    Major performance release. Includes automatic keyword caching, optional shared dictionaries, and many other optimizations that together yield up to 2x faster performance in several common cases.

    As always, please test carefully before updating against production data, and report any unexpected problems - thank you! 🙏

    - Peter Taoussanis

    BREAKING

    Users of low-level freeze-to-out!/freeze-to-bb! and thaw-from-in!/thaw-from-bb!, please note:

    If your freeze calls are wrapped in with-cache, your thaw calls MUST likewise be wrapped (with matching session boundaries). This has always been true, but is extra important now that Nippy has automatic keyword caching enabled by default [03725ac].

    Rolling updates

    Updating Nippy with a rolling deployment (>1 coexisting Nippy version)?

    Then please ensure you update step by step (e.g. v3.5 → v3.6 → v3.7) so that read support is fully deployed before the corresponding writes might happen.

    Shared dictionaries are opt-in: deploy v3.9 readers with matching dictionaries BEFORE enabling dictionary writes.

    New byte output

    v3.9 changes the byte output of several common Nippy data types, including repeated keywords and tightly packed strings, collections, and longs.

    Almost no one should care about this, but mentioning for the tiny minority of users that depend on Nippy's byte output remaining stable.

    Since v3.9.0-RC1 (2026-08-21)

    • [fix] Cache: respect custom freezers [8f48e3c]
    • [prf] Thaw: reduce cache allocations [67cceaa]
    • [prf] Arrays: streamline indexed loops [4bec13b]
    • [prf] Deftypes: streamline field reads [7d177f3]
    • [doc] Benchmarks: update for v3.9.0 [715f78f]

    Since v3.8.1 (2026-07-28)

    • [fix] write-long: support Long/MIN_VALUE [2115360]
    • [fix] Cache: respect custom freezers [8f48e3c]
    • [new] Add shared dictionaries [da69caf]
    • [new] Update benchmark results [74eebdc, df391d2]
    • [prf] Cache: auto-cache keywords [03725ac]
    • [prf] Freeze: enable tight packing by default [2705bb1]
    • [prf] Freeze: copy-free compression [e36a2ff]
    • [prf] Freeze: fast-path common elements [bbff6c6]
    • [prf] Freeze: streamline common paths [50a56ac]
    • [prf] Freeze: optimize default header copy [79f17cd]
    • [prf] Thaw: reduce cache allocations [67cceaa]
    • [prf] Arrays: streamline indexed loops [4bec13b]
    • [prf] Deftypes: streamline field reads [7d177f3]
    • [prf] Thaw: direct transient loop for read-into [5f05346]
    • [prf] Thaw: decode keywords from bytes [bff8c29]
    • [prf] Thaw: build lists directly [5a417fc]
    • [prf] Thaw: decode strings in place [2ac10be]
    • [prf] Thaw: bulk-build short vectors [7fde2b1]
    • [prf] Thaw: avoid header copies [d119f59]
    • [prf] Crypto: reduce buffer copies [1b62a1d]
    • [prf] Cache: use ArrayList for read-cached [ecb1206]
    • [prf] Cache: identity maps for keywords [ac4958e]
    • [prf] Cache: use mutable Java state [eecda2d]
    • [doc] Benchmarks: update for v3.9.0 [715f78f]
    • [doc] Update rolling deployment info [5ea1728]
    • [doc] Remove unnecessary details [97f0252]
    • [nop] Bump Encore from v3.169.1 to v3.171.0 [d44feec]
    • [nop] Streaming: simplify dispatch [86a4c20]
  2. v3.9.0-RC1v3.9.0-RC1 (2026-08-21)pre-release2026-08-21 · 2026-08-21 05:15 UTC
    查看更新说明

    This promotes v3.9.0-beta1 to release candidate with no changes.

    Major performance release. Includes new automatic keyword caching, and many other optimizations that together yield up to 2x faster performance in several common cases.

    Intended as a non-breaking release, but touches a lot of core code. As always please test carefully before using for anything important, and please report any unexpected problems - thank you! 🙏

    - Peter Taoussanis

    Rolling updates

    Updating Nippy with a rolling deployment (>1 coexisting Nippy version)?

    Then please ensure you update step by step (e.g. v3.5 → v3.6 → v3.7) so that read support is fully deployed before the corresponding writes might happen.

    New byte output

    v3.9 changes the byte output of several common Nippy data types.

    Almost no one should care about this, but mentioning for the tiny minority of users that depend on Nippy's byte output remaining stable.

    Since v3.9.0-beta1 (2026-08-06)

    No changes

    Since v3.8.1 (2026-07-28)

    • [fix] write-long: support Long/MIN_VALUE [2115360]
    • [new] Add shared dictionaries [da69caf]
    • [new] Update benchmark results [df391d2]
    • [prf] Cache: auto-cache keywords [03725ac]
    • [prf] Freeze: enable tight packing by default [2705bb1]
    • [prf] Freeze: copy-free compression [e36a2ff]
    • [prf] Freeze: fast-path common elements [bbff6c6]
    • [prf] Freeze: streamline common paths [50a56ac]
    • [prf] Freeze: optimize default header copy [79f17cd]
    • [prf] Thaw: direct transient loop for read-into [5f05346]
    • [prf] Thaw: decode keywords from bytes [bff8c29]
    • [prf] Thaw: build lists directly [5a417fc]
    • [prf] Thaw: decode strings in place [2ac10be]
    • [prf] Thaw: bulk-build short vectors [7fde2b1]
    • [prf] Thaw: avoid header copies [d119f59]
    • [prf] Crypto: reduce buffer copies [1b62a1d]
    • [prf] Cache: use ArrayList for read-cached [ecb1206]
    • [prf] Cache: identity maps for keywords [ac4958e]
    • [prf] Cache: use mutable Java state [eecda2d]
  3. v3.9.0-beta1v3.9.0-beta1 (2026-08-06)pre-release2026-08-06 · 2026-08-06 07:38 UTC
    查看更新说明

    BETA: for early testers only, not yet recommended for production.

    Major performance release. Includes new automatic keyword caching, and many other optimizations that together yield up to 2x faster performance in several common cases.

    Intended as a non-breaking release, but touches a lot of core code. As always please test carefully before using for anything important, and please report any unexpected problems - thank you! 🙏

    - Peter Taoussanis

    Rolling updates

    Updating Nippy with a rolling deployment (>1 coexisting Nippy version)?

    Then please ensure you update step by step (e.g. v3.5 → v3.6 → v3.7) so that read support is fully deployed before the corresponding writes might happen.

    New byte output

    v3.9 changes the byte output of several common Nippy data types.

    Almost no one should care about this, but mentioning for the tiny minority of users that depend on Nippy's byte output remaining stable.

    Since v3.8.1 (2026-07-28)

    • [fix] write-long: support Long/MIN_VALUE [2115360]
    • [new] Add shared dictionaries [da69caf]
    • [new] Update benchmark results [df391d2]
    • [prf] Cache: auto-cache keywords [03725ac]
    • [prf] Freeze: enable tight packing by default [2705bb1]
    • [prf] Freeze: copy-free compression [e36a2ff]
    • [prf] Freeze: fast-path common elements [bbff6c6]
    • [prf] Freeze: streamline common paths [50a56ac]
    • [prf] Freeze: optimize default header copy [79f17cd]
    • [prf] Thaw: direct transient loop for read-into [5f05346]
    • [prf] Thaw: decode keywords from bytes [bff8c29]
    • [prf] Thaw: build lists directly [5a417fc]
    • [prf] Thaw: decode strings in place [2ac10be]
    • [prf] Thaw: bulk-build short vectors [7fde2b1]
    • [prf] Thaw: avoid header copies [d119f59]
    • [prf] Crypto: reduce buffer copies [1b62a1d]
    • [prf] Cache: use ArrayList for read-cached [ecb1206]
    • [prf] Cache: identity maps for keywords [ac4958e]
    • [prf] Cache: use mutable Java state [eecda2d]
  4. v3.9.0-alpha2v3.9.0-alpha2 (2026-07-29)pre-release2026-07-29 · 2026-07-29 03:39 UTC
    查看更新说明

    EARLY PREVIEW: only for early testers, not suitable for production.

    Major performance release. Includes new automatic keyword caching, and many other optimizations that together yield up to 2x faster performance in several common cases.

    Intended as a non-breaking release, but touches a lot of core code. As always please test carefully before using for anything important, and please report any unexpected problems - thank you! 🙏

    - Peter Taoussanis

    Rolling updates

    Updating Nippy with a rolling deployment (>1 coexisting Nippy version)?

    Then please ensure you update step by step (e.g. v3.5 → v3.6 → v3.7) so that read support is fully deployed before the corresponding writes might happen.

    New byte output

    v3.9 changes the byte output of several common Nippy data types.

    Almost no one should care about this, but mentioning for the tiny minority of users that depend on Nippy's byte output remaining stable.

    Since v3.8.1 (2026-07-28)

    • [new] Update benchmark results [df391d2]
    • [prf] Cache: auto-cache keywords [03725ac]
    • [prf] Freeze: enable tight packing by default [2705bb1]
    • [prf] Freeze: copy-free compression [e36a2ff]
    • [prf] Freeze: fast-path common elements [bbff6c6]
    • [prf] Freeze: streamline common paths [50a56ac]
    • [prf] Freeze: optimize default header copy [79f17cd]
    • [prf] Thaw: direct transient loop for read-into [5f05346]
    • [prf] Thaw: decode keywords from bytes [bff8c29]
    • [prf] Thaw: build lists directly [5a417fc]
    • [prf] Thaw: decode strings in place [2ac10be]
    • [prf] Thaw: bulk-build short vectors [7fde2b1]
    • [prf] Thaw: avoid header copies [d119f59]
    • [prf] Crypto: reduce buffer copies [1b62a1d]
    • [prf] Cache: use ArrayList for read-cached [ecb1206]
    • [prf] Cache: identity maps for keywords [ac4958e]
    • [prf] Cache: use mutable Java state [eecda2d]
  5. v3.8.1v3.8.1 (2026-07-28)2026-07-28 · 2026-07-28 08:53 UTC
    查看更新说明

    Hotfix release, should be a safe update for users on v3.8.0 - but as always please test carefully before using in production, and please report any unexpected problems - thank you! 🙏

    - Peter Taoussanis

    Since v3.8.0 (2026-07-26)

    • [fix] freeze-to-out!: restore size-unlimited streaming [2613609]
    • [fix] with-bb: cap buffer growth at max array size [930a3cb]
    • [sec] Thaw: validate buffered allocation lengths [ca0fa4f]
    • [fix] Thaw: reject negative collection counts [b2e16e9]
    • [fix] Crypto: reject truncated encrypted payloads [fe8ab01]
    • [fix] with-cache: restore outer cache [c368c4a]
    • [fix] freeze-to-out!: restore cache on failure [bb8078b]
    • [fix] freeze-to-bb!: restore state on failure [c0b2228]
    • [fix] write-sz: serialize before writing header [7bc1e3e]

用户点评

登录用户的真实使用记录，可按版本和硬件筛选
共 0 条

    还没有人点评。第一个说出真实体验的人最有价值。

    README 原文

    未翻译
    在 GitHub 打开


    API | Wiki | Support | Latest release: v3.9.0 (2026-09-06)

    Clj tests Graal tests

    Nippy

    Fast serialization library for Clojure

    Clojure's rich data types are awesome. And its reader allows you to take your data just about anywhere. But the reader can be painfully slow when you've got a lot of data to crunch (like when you're serializing to a database).

    Nippy is a mature, high-performance drop-in alternative to the reader.

    It is used at scale by Carmine, Faraday, PigPen, Onyx, XTDB, Datalevin, and others.

    Why Nippy?

    • Small, simple pure-Clojure library
    • Terrific performance: the best for Clojure that I'm aware of
    • Comprehensive support for all standard data types
    • Easily extendable to custom data types
    • Robust test suite incl. coverage of every supported type
    • Mature and widely used in production for 12+ years
    • Optional auto fallback to Java Serializable for safe types
    • Optional auto fallback to Clojure Reader (including tagged literals)
    • Optional smart compression with LZ4 or Zstandard
    • Optional encryption with AES128
    • Tools for easy + robust integration into 3rd-party libraries, etc.
    • Powerful thaw transducer for flexible data inspection and transformation

    Quick example

    Nippy's super easy to use:

    (require '[taoensso.nippy :as nippy])

;; Freeze any Clojure value
(nippy/freeze <my-value>) ; => Serialized byte[]

;; Thaw the byte[] to get back the original value:
(nippy/thaw (nippy/freeze <my-value>)) ; => <my-value>

    See the wiki for more.

    Operational considerations

    Data longevity

    Nippy is widely used to store long-lived data and promises (as always) that data serialized today should be readable by all future versions of Nippy.

    But please note that the converse is not generally true:

    • Nippy vX should be able to read all data from Nippy vY<=X (backwards compatibility)
    • Nippy vX may/not be able to read all data from Nippy vY>X (forwards compatibility)

    Rolling updates and rollback

    From time to time, Nippy may introduce:

    • Support for serializing new types
    • Optimizations to the serialization of pre-existing types

    To help better support rolling deployments and rollback, Nippy will always introduce such changes over two version releases:

    • Release 1: to add read support for the new types
    • Release 2: to add write support for the new types

    So if you use rolling deployments please ensure you update step by step (e.g. v3.5 → v3.6 → v3.7) so that read support is fully deployed before the corresponding writes might happen.

    Stability of byte output

    It has never been an objective of Nippy to offer predictable byte output, and I'd generally recommend against depending on specific byte output.

    However, I know that a small minority of users do have specialized needs in this area.

    So starting with Nippy v3.4, Nippy's release notes will always clearly indicate if any changes to byte output are expected.

    Performance

    Nippy is pretty fast. Latest benchmark results:

    benchmarks-png

    PRs welcome to include other alternatives in the bench suite!

    Documentation

    • Wiki (getting started, usage, etc.)
    • API reference via cljdoc

    Funding

    You can help support continued work on this project and others, thank you!! 🙏

    License

    Copyright © 2012-2026 Peter Taoussanis.
    Licensed under EPL 1.0 (same as Clojure).

    相关项目

    同分类里 star 最高的项目
    对比全部 →
    CyberChef
    gchq·JavaScript

    网页版的数据处理工具，把加密、编码、压缩、哈希和解析等操作组合成流水线，方便安全分析人员快速转换和排查数据。

    数据转换+4压缩解压
    36366+254.2kv11.5.02026-11-07 12:52 UTC
    digital_video_introduction
    leandromoreira·Jupyter Notebook

    用 Jupyter Notebook 带你动手理解视频技术：从图像、帧类型到 AV1/VP9/H265 编解码与 FFmpeg 编码实践，适合入门视频压缩与流媒体。

    视频编解码+4压缩解压
    16384+21.4k1.0.12026-11-05 12:00 UTC
    borg
    borgbackup·Python

    去重备份归档工具，支持压缩和认证加密，可通过 SSH 远程备份，适合需要高效增量备份和隐私保护的数据场景。

    去重备份+4压缩解压
    13918+48821.4.52026-11-10 21:58 UTC
    lz4
    lz4·C

    极快的无损压缩算法，压缩和解压速度远超 gzip，适合对延迟敏感的场景，比如日志传输、内存数据压缩和嵌入式存储。

    无损压缩+4压缩解压
    12750+471.7kv1.11.02026-11-10 23:05 UTC