Executive Process

Node.js provide http module to create HTTP service, used to respond to user requests, such as Node.js official website provide examples of creating HTTP services:

const http = require('http');

const hostname = '127.0.0.1';
const port = 3000;

const server = http.createServer((req, res) => {
  res.statusCode = 200;
  res.setHeader('Content-Type', 'text/plain');
  res.end('Hello World\n');
});

server.listen(port, hostname, () => {
  console.log(`Server running at http://${hostname}:${port}/`);
});

ThinkJS also use http.createServer to create service, so the whole process contains two parts to start service and respond to user requests.

Start system service

• Run npm start or node development.js.
• Instantiate ThinkJS Application class，and call run method.
• Depending on the environment (Master process, Worker process, command-line calls) to deal with different logic.
• If it is Master process
  • Load config file and produce think.config think.logger object.
  • Load src/bootstrap/master.js file.
  • If config to watch service, then start to listen files changes within folder src/.
  • After file changed, if config to compile source code, it will do the compilation and output to app/ folder.
  • According to workers configuration to decide the number of Worker to fork. When the worker process has started, the appReady event is fired. (Can be captured via think.app.on("appReady")).
  • If the file has a new modification, it will trigger the compilation, and then kill all Worker processes and re-fork.
• If it is Worker process
  • Load the configuration file to generate the think.config andthink.logger objects.
  • Extend is loaded to provide more functionality for the framework, the configuration file is src/config/extend.js.
  • Get the list of modules for the current project, on think.app.modules, or empty array if single module.
  • Load the controller file (src/controller/*.js) in the project and place it on thethink.app.controllers object.
  • Load the logic file (src/logic/*.js) in the project and place it on the think.app.logics object.
  • Load the project's model file (src/model/*.js) and place it on the think.app.models object.
  • load the service file (src/service/*.js) in the project and place it on the think.app.services object.
  • Load the route configuration file src/config/router.js on the think.app.routers object.
  • Load the validation configuration file src/config/validator.js on the think.app.validators object.
  • Load the middleware configuration file src/config/middleware.js and register it with the think.app.use method.
  • Loads the timing task configuration file src/config/crontab.js and registers the timing task service.
  • Load the src/bootstrap/worker.js startup file.
  • Listen for the onUncaughtException andonUnhandledRejection errors in the process and process them. You can customize these two erroneous handlers by configuring src/config.js.
  • Wait for the pre-launch handler for think.beforeStartServer registration, where you can register some transactions before starting the service.
  • If you have customized the create service configuration createServer, execute this function createServer(port, host, callback) to create the service.
  • If there is no customization, start the service with think.app.listen.
  • The appReady event is fired when service startup is complete, elsewhere can listened to this event via think.app.on("appReady").
  • The created service is assigned to the think.app.server object.

After the service starts, the following log is printed:

[2017-07-02 13:36:40.646] [INFO] - Server running at http://127.0.0.1:8360
[2017-07-02 13:36:40.649] [INFO] - ThinkJS version: 3.0.0-beta1
[2017-07-02 13:36:40.649] [INFO] - Enviroment: development  #current running environment
[2017-07-02 13:36:40.649] [INFO] - Workers: 8   #worker process number

User request processing

When the user requests service, it will be processed through the following steps.

• The request arrives at the webserver (eg: nginx) and forwards the request to the node service via the reverse proxy. If you access the node service directly through the port, then there is no such step.
• Node service to receive user requests, the Master process will be forwarded to the corresponding Worker process.
• Worker process through the registration of the middleware to handle the user's request:
  • meta to handle some common information such as: setting request timeout, sending ThinkJS version number, sending processing time, etc.
  • resource Processing static resource requests, static resources are placed under www/static/, if a static resource is hit, this middleware will response resource and stop the following middleware.
  • trace Handles some error messages, prints detailed error messages in the development environment, and the production environment simply reports a generic error.
  • payload Handling user-uploaded data, including: form data, files, etc. After the analysis is completed, put the data on request.body object for later accessing.
  • router Parse the route to match a Controller's Action, the result is saved on ctx.controller andctx.action for subsequent processing . If the project is a multi-module structure then there is ctx.module.
  • logic According to the controller and action parsed out, call the corresponding method in logic.
    • Instantiate the logic class and pass ctx into it. If there is no then skip this.
    • Execute __before method, if it returnsfalse, it will not execute all subsequent logic (end prematurely)
    • If the xxxAction method exists then it will be executed. If the result isfalse, then all subsequent logic will not be executed
    • If the xxxAction method does not exist, then try the__call method
    • Execute the __after method, and if it returns false, no subsequent logic will execute
    • By the method returns false to block the implementation of subsequent logic
  • controller According to the resolution of the controller and action, call the corresponding method in the controller.
    • Specific call strategy is exactly the same with logic
    • If not, then the current request returns 404
  • When the action is done, you can put the result on the this.body property and return it to the user.
• When Worker error, trigger onUncaughtException or onUnhandledRejection event, or Worker abnormal exit, Master will capture an error, re-fork a new Worker process, and kill the current process.

We can see that all the user requests are handled through middleware. In specific projects, we add assemble more middleware to handle user's request accordingly.