wizardoc / wizard Goto Github PK

原先用的是 JSON 传输，准备切到 protobuf

Currently, the mode field only does the one thing that automatically adds prefixes depending on whether it is a DEV environment, but in some condition, it's not very fixable for developers, in instance, I need to custom the prefix of request URL sometimes, through the @wizardoc/http-request allow us to change the prefix in HTTPFactory but the mode is very useful now, it just kind of like the mode field is redundant.

I'm thinking about that if it is possible to replace the mode field with replace field directly, that might allow developers can pass the prefix of request URL in server config.

Now Archie has added the follow and unfollow interface, and we need to call them and respond to the client(website, App or H5).

U can refer to the interface implementation for login and register in Node middleware.

@Put('login')
  login(@Body() loginBody: LoginBody) {
    const query = gql`
      mutation login($username: String!, $password: String!) {
        login(username: $username, password: $password) {
          user {
            id
            username
            displayName
            avatar
            registerTime
            loginTime
          }
          token
        }
      }
    `;

    return this.http.sendQuery(query, loginBody);
  }

/user/login路径打开不显示

需要一个统一的部署控制平台来监控 docker 的部署情况（wizard, archie）

添加

feat: msg
fix: msg

这种形式的msg内容兼容

在windows下特殊字符无法识别处理

babel 不会去读 tsconfig.json，这使得 react-ts-di 不能发挥作用，在 wizard 里，是用 decorator 驱动来做 DI 的，因此，升级受阻，决定回滚

目前的行为是在拦截器里发生 Error 不会抛到最外层

Motivation

For a better user experience, the better way to do that is when the user first reach the server will return a rendered skeleton view to them, if u wanna learn more, please move to this issue.

Also, the skeleton page is building by React or something like JSX framework, so I have to mix the skeleton code with the business code that makes them coupled to each other heavily, seems like It's not a good idea though a lot of projects do that. We need a fantastic solution!

Solution

We used the node server to request the rendered page of the current project, and use skeleton component instead of elements that like img, text or something like that after the page was rendered, then export the HTML as a page after rendered for the first screen render via innerHTML or something like that.

Now we have decoupled skeleton code and business code, and the skeleton code is generated automatically by node server, so we even don't have to write skeleton code over and over again for each page, we just need to write once and run anywhere!

• plugin system that allows users to create their own plugin to enhance rules of validation.
• custom rules. The user can create a custom rule of validation to validate the commit message.
• configuration driver. The user just needs to provide a config file in the root by package.json side, and the config file will tell the CLI how to validate the commit message that contains validation rules and some plugins, just like .eslintrc or something like that.

用 WebWorker 改成 C/S 架构抽离

Parcel only build successfully if has .parcel-cache, so we just need to run parcel build twice to ensure compilation successfully, the first compilation is used to generate the .parcel-cache file, and the second compilation will only compile successfully with cache file that I have ro run parcel build || parcel build to compile source code.

Please consider that upload the app image in DockerHub when build on GitHub Actions CI, and pull the image from DockerHub using mirror when on the Tencent server

SSR is difficult to integrate into the current project, so we need a skeleton screen as the first render screen, but the client was being supplied by NginX, so we might migrate the assets of client build to node middleware, and the server will supply Skeleton HTML when users reached.