BoxLite 试图找个平衡点：
- 基于 libkrun ，每个容器跑在独立 microVM 里
- ~50ms 启动，内存开销小
- 无 daemon 架构，嵌入到应用里当库用
- 完整支持 OCI 镜像、网络、存储卷
技术栈：Rust + libkrun + virtio + gvproxy
主要场景：
- AI Agent 沙箱 - 让 Agent 随便折腾不怕炸
- 多租户隔离 - SaaS 场景下的强隔离
- macOS 本地开发 - 不装 Docker Desktop 也能跑 Linux 容器
目前 Python SDK 已经可用，Node.js 和 Go 在开发中。
GitHub: https://github.com/boxlite-labs/boxlite PyPI: pip install boxlite
求 star ，欢迎 PR ！
]]>什么是 FAML ？
FAML 是一种扩展自 TOML 的动态配置语言，专为需要运行时配置计算和更新的场景设计。它保留了 TOML 的简洁语法，同时增加了动态表达式、条件配置和运行时可变性等高级特性。
核心特性对比
| 特性 | TOML | KCL | PKL | FAML |
|---|---|---|---|---|
| 语法风格 | TOML 风格 | JSON 风格 | 结构体风格 | TOML 风格 |
| 动态表达式 | ❌ | ✅ | ✅ | ✅ |
| 条件配置 | ❌ | ✅ | ✅ | ✅ |
| 运行时修改 | ❌ | ❌ | ❌ | ✅ |
| 特殊数据类型 | ❌ | ✅ | ✅ | ✅ |
快速示例
基本语法
[server] port = 8080 host = "localhost" 动态表达式
[database] host = "localhost" port = 5432 connection_string = $"postgresql://{host}:{port}/mydb" 条件配置
[app] env = "production" @if env == "development" log_level = "debug" @if env == "production" log_level = "error" 特殊数据类型
[cache] ttl = 5 minutes max_size = 100 MB [network] timeout = 30 seconds buffer_size = 4 KB 复杂表达式
[user] age = 25 is_adult = age >= 18 welcome_message = is_adult ? $"Welcome, adult user!" : $"Welcome, young user!" 运行时动态修改
let mut cOnfig= FamlExpr::from_str(config_str)?; config["server"]["port"].set_int(9000); // 动态修改端口 let connection_string = config["database"]["connection_string"].evaluate()?.as_str(); // 自动更新连接字符串 新版本完全重构，采用更符合标准的 workspace 模式实现多应用
]]>两年没有 release,是什么原因？有没有类似的项目
]]>LZ 是一名在职的 Java boy ，花了上周一周的时间入门了 Rust 。周末使用 axum 写了个 web 项目，模拟了 SpringMVC 的架构。
2. 对比
均在我 mac 笔记本中进行测试
2.1 占用资源
这里可能不太公平。
Java 项目用了我之前的开源项目，虽然是个完整的项目但是也不是很大。没有打包成 jar ，使用 idea 启动，通过活动检测器根据 pid 查找发现占用内存494.2MB，JVM 配置：-Xms256m -Xmx4g 。
Rust 启动后发现内存占用为 2.0MB!!! 我人都傻了，虽然说是个很简单的项目，但是项目中必要的功能都有。
2.2 压测
为了进一步的看到差距，我开始了压测，代码如下。为了减少变量，没有加入 IO 操作，但也是正常的业务功能。
JAVA
@Override public AiUserVO loginTest(AiUserReq userReq) { CommonUtils.checkArgs(userReq.getAccountNo(),userReq.getPassword()); AiUser user = AiUser.builder().accountNo("admin").password("admin").name("admin").build(); user.setId(1941040891798581249L); //校验登录 // checkLogin(user,userReq); AiUserVO result = ConvertUtils.beanProcess(user, AiUserVO.class); //处理用户角色 // processUserRole(user,result); //生成 token String token = genToken(user); result.setToken(token); //放入 redis String key = CacheKeyEnum.format(CacheKeyEnum.ADMIN_LOGIN_KEY,user.getId(),token); SysCacheUserDto sysCacheUserDto = ConvertUtils.beanProcess(result, SysCacheUserDto.class); //删除原始 token // removeToken(user.getId()); // redisCacheComponent.set(key,JSONObject.toJSONString(sysCacheUserDto),TimeConstant.ONE_DAY * 12); return result; } RUST
pub async fn login(user:UserReq)->Res<UserVO> { // 查询逻辑 let user_do = match query_user(&user.account_no, &user.password).await { Some(user_do) => user_do, NOne=> return Res::build(FAIL), }; let mut user_vo:UserVO = user_do.into(); //加密 let token = match security_utils::aes_encrypt(&user_vo.id.to_string()) { Ok(token) => token, Err(_) => return Res::build(FAIL), }; user_vo.token = token; let key = SYSTEM_USER.key().format(&[&user_vo.id.to_string()]); //删除缓存 GlobalCache::del(&key); //放入缓存 GlobalCache::set(key,serde_json::to_string(&user_vo).unwrap()); Res::success(user_vo) } /** * 查询用户 */ async fn query_user(account_no:&str,pwd:&str)->Option<UserDO> { let user = UserDO{ id:1912753992656158721, name:"admin".to_string(), account_no:"admin".to_string(), status:true, }; Some(user) } 压测配置如下
Name: 线程组-30 Number of Threads: 200 Ramp-Up Period: 30 Loop Count: 1 Scheduler: Duration: 90 Name: 线程组-90 Number of Threads: 200 Ramp-Up Period: 30 Loop Count: 1 Scheduler: Startup Delay: 90 Duration: 90 压测流程如下
阶段一：0~30 秒启动 200 线程 → 持续到第 90 秒 阶段二：从第 90 秒开始，30 秒内启动 300 线程 → 持续到第 180 秒 压测结果对比如下
| 指标 | RUST | JAVA | 差异 |
|---|---|---|---|
| 平均响应时间 | 3~4 ms | 6~8 ms | ↑ 上升 50%~100% |
| 最大响应时间 | 17 ms | 26 ms | ↑ 上升 53% |
| 99% 百分位 | 11~15 ms | 15~20 ms | ↑ 上升 30%~45% |
| 吞吐量 | 6.7 → 10.0 req/sec | 6.7 → 10.0 req/sec | ✅ 相同 |
| 异常率 | 0% | 0% | ✅ 相同 |
| 数据包大小 | 294 B | 489 B | ↑ 增加 66% |
]]>
- 于是发现了 loco ，这是根据 rails 风格搞的。用起来很顺手
想起它骨子里的坚定，克制，却又锋利
我敬畏，但也安心
它有清晰的边界感，每一个指针都有归属，每一次借用都遵循秩序
它不给我侥幸，却给我前所未有的安全
它的世界里没有野蛮的随意，没有未定义的空洞，只有规则下的极致自由
它能在不失控的前提下奔跑到极限，并发如风，却稳若磐石
Rust ，我想你了
]]>https://github.com/auv-sh/avav
av top av list avavavaavavavavavavavavavavavavavavavaavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavavavaavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavavavaavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavavvavavavavavavavav
]]>Moritz 首先承认了 Rust 的优点：理论上，Rust 通过其所有权系统和借用检查器，提供了远超 C++ 的默认安全性。这正是大家喜爱 Rust 的原因。
但问题在于，像 CedarDB 这样的高性能数据库，其开发工作远不止是处理业务逻辑。它需要深入到硬件的毛细血管中，榨干每一滴性能。这意味着：
使用大量底层的 CPU 特性。
实现复杂的侵入式数据结构。
进行带有验证的、乐观且激进的内存访问。
在这些场景下，Rust 的安全检查反而成了“束缚”。为了完成任务，你必须使用 Rust 提供的“逃生舱口”——unsafe 关键字。
而 Moritz 抛出的重磅炸弹正在于此：
“一旦你在 Rust 中写下 unsafe 代码，所有的赌注都将失效。在 unsafe 代码中，你遇到未定义行为（ UB, Undefined Bahavior ）的风险，甚至比在 C++ 中还要高。”
https://x.com/iavins/status/1946598328533430702 ]]>
详情见： https://github.com/chunhuitrue/nsave
Nsave 是一个抓取并保存数据包的工具。它持续不断地抓取数据包，并保存到本地。可以根据条件查询链接、数据包并导出成 pcap 文件。可以通过 pcap 或者 af_xdp 来捕获数据包。主要特点是它不基于单个数据包，而是基于流来作索引，可以大幅减少索引所占的磁盘空间。
提醒
- 目前是开发阶段，不要应用在关键的生产环境。
- 需要配置单独的抓包网卡。管理网卡不能用于抓包，否则你会失去连接，因为加载的 xdp 程序会把网卡上所有的数据包都截获，不再流入内核。
WASM 计算器 - 使用 WebAssembly 和 Rust 技术构建的高性能在线计算器应用，支持基础运算、百分比计算，界面美观，运行速度快，是您日常计算的理想选择。
地址👉: 在线地址
后续：
学习 wasm 的练手项目，新手上路，轻拍
]]>使用软件的人，应该不太关心软件使用什么语言开发的吧？
]]>曾以为 rust 能改变 linux ，取代 C
现在 rust for linux 写个接口能被干
曾以为 rust 能借着 wasm 升天
现在 photoshop web ，figma 用 C++
曾以为 rust 的 deno 能干掉 nodejs
现在 deno repl 看个彩色 log
曾以为 rust 重写的 cli 是未来，alias, lsd, dust, dog...
现在 unalias, ls, du, dig... ]]>
2 、枚举类型的变量看不到内部具体的值 ]]>
667.3 = note: "cc" "-m64" "/tmp/rustcXFCL9g/symbols.o" "<1 object files omitted>" "-Wl,--as-needed" "-Wl,-Bstatic" "/tmp/rustcXFCL9g/{libv8-dbef52e4bc4b6f8f.rlib,libskia_bindings-279c509e56943a3f.rlib,libzstd_sys-356ea57c902a259b.rlib}.rlib" "<sysroot>/lib/rustlib/x86_64-unknown-linux-gnu/lib/{libcompiler_builtins-*}.rlib" "-Wl,-Bdynamic" "-ldl" "-ldl" "-lssl" "-lcrypto" "-ldav1d" "-lstdc++" "-lfontconfig" "-lfreetype" "-lEGL" "-lGL" "-lwayland-egl" "-lGLESv2" "-lgcc_s" "-lutil" "-lrt" "-lpthread" "-lm" "-ldl" "-lc" "-Wl,--eh-frame-hdr" "-Wl,-z,noexecstack" "-L" "/usr/lib" "-L" "/usr/local/lib/x86_64-linux-gnu" "-L" "/home/app/target/release/build/skia-bindings-60ededebb9dd8e7e/out/skia" "-L" "/home/app/target/release/build/zstd-sys-e9bd7677eed9b254/out" "-L" "/home/app/target/release/gn_out/obj" "-L" "<sysroot>/lib/rustlib/x86_64-unknown-linux-gnu/lib" "-o" "/home/app/target/release/deps/render_http-e7adb05509d015b0" "-Wl,--gc-sections" "-pie" "-Wl,-z,relro,-z,now" "-Wl,-O1" "-Wl,--strip-debug" "-nodefaultlibs" 667.3 = note: some arguments are omitted. use `--verbose` to show all linker arguments 667.3 = note: /usr/bin/ld: /tmp/rustcXFCL9g/libskia_bindings-279c509e56943a3f.rlib(libicu.uloc.o): in function `ulocimp_getParent': 667.3 uloc.cpp:(.text.ulocimp_getParent+0x0): multiple definition of `ulocimp_getParent'; /tmp/rustcXFCL9g/libv8-dbef52e4bc4b6f8f.rlib(uloc.o):./../../../../third_party/icu/source/common/uloc.cpp:1685: first defined here 667.3 collect2: error: ld returned 1 exit status 有没有大佬提供帮助
]]>写到一半的时候发现还是有很多内容不熟悉，看文档的时候都知道是咋回事，真正自己上手写的时候就懵逼
这种感觉真的太受挫了！
]]>
如果 web 项目用到 gRpc 的话，通过 protoc 创建的这些服务，只能通过 layer 贯穿到整个服务中吗？？
那正常的一个项目假如有几十上百个服务的话，那不是这个.layer(Extension(Arc::new(model::AppState::new(cate, topic, tera)))); 会很臃肿，因为要把这几十上百个服务都累加进去？
]]>但是我正阻塞在该对象上进行读取，又怎么可能通过 drop 该对象来触发关闭呢？
请别告诉我要非阻塞调用，或者轮询调用，或者其他解决方案，我只是觉得接口这么设计我有点无法理解。
]]>models-cat 是 ModelScope Hub 的非官方 Rust 客户端，设计灵感来自 hf-hub。models-cat 源自一个简单的需求：“编写一个 Rust 桌面端 AI APP ，需要下载模型和数据集，但是没有合适的 Rust 客户端。”
什么时候需要 models-cat 下载模型？主要有三个原因：
- 网络原因，无法使用 hf-hub 访问 huggingface。
- 虽然使用 hf-hub 也可以从 hf-mirror 下载模型，但稳定性和下载速度无法保证。
- 将模型托管在 ModelScope Hub 上，可以保证稳定性和下载速度。但 hf-hub 不兼容 ModelScope, 需要使用 models-cat 。
顺便说下，ModelScope 托管模型真香，国内下载速度贼快、稳定、免费。
功能特性
- 模型/数据集下载与缓存管理, 同步或异步下载
- 支持并发安全访问文件
- 本地缓存校验（ SHA256 ）
- 下载进度回调
使用示例
同步下载：
use models_cat::{download_model_with_progress, ProgressBarWrapper}; download_model_with_progress( "BAAI/bge-small-zh-v1.5", "model.safetensors", ProgressBarWrapper::default(), ).unwrap(); 异步下载：
use models_cat::asynchronous::{download_model_with_progress, ProgressBarWrapper}; download_model_with_progress( "BAAI/bge-small-zh-v1.5", "model.safetensors", ProgressBarWrapper::default(), ).await.unwrap(); 异步下载需开启特性tokio特性：
models-cat = { version = "*", features = ["tokio"] } 从 ModelScope 的托管仓库 BAAI/bge-small-zh-v1.5 下载模型到本地，默认保存在[HOME_DIR].cache/modelscope/hub/models--BAAI--bge-small-zh-v1.5/目录下。
使用快捷函数时，可通过环境变量MODELS_CAT_CACHE_DIR设置本地缓存路径。或者使用ModelsCat，在初始化时传入本地缓存路径。
pub async fn list( State(state): State<ArcAppState>, Query(frm): Query<form::subject::ListForAdmin>, ) -> Result<resp::JsonResp<model::subject::SubjectPaginate>> { let handler_name = "admin/subject/list"; let p = get_pool(&state); let subjects = model::subject::Subject::list( &*p, &model::subject::SubjectListFilter { pq: model::subject::SubjectPaginateReq { page: frm.pq.page(), page_size: frm.pq.page_size(), }, order: None, is_del: frm.is_del(), status: frm.status, name: frm.name, slug: frm.slug, }, ) .await .map_err(Error::from) .map_err(log_error(handler_name))?; Ok(resp::ok(subjects)) } #[derive(Debug, Default, Deserialize, Serialize, sqlx::FromRow, Db)] #[db(table = subjects, pk = id, del_field = is_del)] pub struct Subject { #[db(find)] #[db(skip_update)] pub id: String, #[db(list_opt)] #[db(list_opt_like)] pub name: String, #[db(find)] #[db(exists)] #[db(list_opt)] #[db(list_opt_like)] pub slug: String, pub summary: String, #[db(find_opt)] #[db(list_opt)] pub is_del: bool, pub cover: String, #[db(list_opt)] pub status: Status, pub price: Decimal, pub pin: i32, } 作为 Java 开发，看到这个 list 方法本能的想跳转过去看下这个 model::subject::Subject::list 方法的实现咋写的，但是跳转过去是宏？ 那我咋知道最后是咋写的呢？
实在是懵逼～
]]>axum：web framwork sqlx：sqlboy
写着写着感觉用 rust 写 web 项目总感觉哪里味不对，一直在反思这 rust 写 web 应用真是适合吗？
就目前 orm 框架 sqlx 、seaorm 、diesel 都还在讨论哪个牛逼～
不像 Java ，mybatis 用起来那么爽～ 我既想要 orm 的字段映射到对象，也想要自定义查询的灵活～
]]>Client 是基于 reqwest 的分支，修复了 HTTP 版本协商问题，并增强了对 WebSocket （ HTTP1/HTTP2 ）的支持，同时优化了常用 API 和连接池的性能。
- Features
-
完美模仿 Chrome/Safari/Firefox 的 TLS/HTTP2 指纹配置。
-
TLS 后端使用 BoringSSL ，该分支像 Golang utls 一样访问低级 TLS 配置功能，同时避免了 utls 的致命缺陷：utls issue #274。
-
HTTP2 后端在原 h2 分支基础上实现了对低级 Priority/Headers frame 及
pseudo-header排列的访问。
- Repository
- Client : https://github.com/0x676e67/rquest
- BoringSSL: https://github.com/0x676e67/boring2
- HTTP2: https://github.com/0x676e67/http2
salvo 和 axum 框架，并同时支持创建「单应用」和「多应用」 安装
Cargo install yiirs 地址
]]>其实中说 FnOne 是至少调用一次。
还有 Rust 圣经中也同样这么说。 内容如下：
所有的闭包都自动实现了 FnOnce 特征，因此任何一个闭包都至少可以被调用一次
但 Kimi 的解释是：
FnOnce 的设计目标是确保闭包在调用时可以安全地消耗变量的所有权，而不是强制要求闭包必须被调用。换句话说：
“只能调用一次”：闭包被设计为只能调用一次，以避免多次调用导致的错误。 “至少调用一次”：这不是 FnOnce 的要求。闭包可以选择不被调用，但一旦调用，就不能再次调用。 大致意思就是“只能调用一次”
我的理解也是只能调用一次。
ps: 前几天看到有个推荐 Rust 教程的页面，说《 Rust 语言圣经》错误太多。我现在找不到出处了。请问有哪些是错误的？
]]>最好有 DEX 方向的 ]]>
cargo build --target aarch64-unknown-linux-gnu -r 也试过 cargo 配置文件里加
[target.aarch64-unknown-linux-gnu] linker = "aarch64-linux-gnu-gcc" rustflags = ["-L", "/usr/lib/gcc/aarch64-linux-gnu/13.1.0/tmp/usr/lib/"] 都解决不了，放到 raspi 编译一次要很久。
]]>比如：
- mise: 管理一些环境版本；
- justfile: makefile 的替代；
- uv: python 包/虚拟环境管理
- blink.cmp: neovim 的一个补全插件
- fd: 文件查找，find 的替代
- rg: ripgrep, grep 的替代
比较突出的特点是比同类的其他工具都更快，用起来也确实很明显。我自己主要用 C++/Python ，想拿 rust 业余玩下（主要也是做些自己用的命令行小工具吧）。有啥上手的小项目推荐吗？（上面几个粗略瞄了眼有点大）
]]>首先看书中图 5-9 的所有权树，讲了对于可变和不可变引用时的访问关系，我以为自己理解了，可是后面看到这个例子，结果又看不明白，代码是这样的：
... 也可以再次借用一个可变引用： let mut v = (136, 139); let m = &mut v; let m0 = &mut m.0; // ok：从可变引用重新借用可变引用 *m0 = 137; let r1 = &m.1; // ok：从可变引用重新借用共享引用,并且和 m0 没有重叠部分 v.1; // error：通过其他路经访问仍然是禁止的 println!("{}", r1); // r1 在这里使用 其实从第一个 ok 注释那里我就没搞明白，不是说变量只能同时存在一个可变引用吗，为什么有可变引用(“引用”为名词)m 的情况下还能从 m 那再可变引用（“引用”为动词）一个 m0 ？
然后我问 gpt ，它是这么回答的：
...
嵌套 mut 引用的场景
假设你有一个可变变量，并希望对其创建嵌套的可变引用。以下 是一些常见的情况及其分析。
情况 1：对整个变量的可变引用，然后尝试对部分值的可变引用
fn main() { let mut x = (1, 2); let r1 = &mut x; // 对整个变量 `x` 的可变引用 let r2 = &mut r1.0; // 尝试对 `x` 的一部分创建可变引用 *r2 = 42; // 修改部分值 r1.1 = 99; // 修改另一部分 } 错误原因： Rust 编译器会报错，提示 不能同时借用 x 的多个可变引用。这是因为：
r1 是对整个 x 的可变引用，它控制了 x 的独占访问权。 当你尝试对 r1.0 （ x 的某个字段）创建可变引用时，r1 的作用域依然活跃，Rust 无法区分你是否会同时操作 x 的其他部分。
错误示例的编译器消息（可能类似于）：
error[E0499]: cannot borrow `r1.0` as mutable more than once at a time ...
但事实上，我尝试了类似代码，并不会报错:
fn main() { let mut v = (136, 139); let m = &mut v; let m0 = &mut m.0; *m0 = 137; m.1 = 42; println!("{:?}", v) } 其运行结果是:
(137, 42) 很多资料都没有讲解关于这种嵌套的引用会发生什么结果，也不知道什么情况是可以什么是不行，我没看过这例子时还以为压根不能创建 m0 呢，所以想着找 GPT 问问，结果告诉的内容是错的。
请问该如何解释上面书中的代码，如何更好地理解 Rust 在这种嵌套情况下创建引用的做法是否成功？请各位赐教
]]>// This function only gets compiled if the target OS is linux #[cfg(target_os = "linux")] fn 在跑 linux() { println!("你在跑 linux!"); } // And this function only gets compiled if the target OS is *not* linux #[cfg(not(target_os = "linux"))] fn 在跑 linux() { println!("你昧在跑 linux!"); } fn main() { 在跑 linux(); println!("确定?"); if cfg!(target_os = "linux") { println!("是。的确!"); } else { println!("是。不是!"); } }