What does middleware and app.use actually mean in Expressjs?

After simplifying things, a web server can be seen as a function that takes in a request and outputs a response. So if you view a web server as a function, you could organize it into several pieces and separate them into smaller functions so that the composition of them will be the original function.

Middlewares are the smaller functions that you can compose with others and the obvious benefit is that you can reuse them.

middleware

I'm halfway through separating the concept of middleware in a new project.

Middleware allows you to define a stack of actions that you should flow through. Express servers themselves are a stack of middlewares.

// express
var app = express();
// middleware
var stack = middleware();

Then you can add layers to the middleware stack by calling .use

// express
app.use(express.static(..));
// middleware
stack.use(function(data, next) {
  next();
});

A layer in the middleware stack is a function, which takes n parameters (2 for express, req & res) and a next function.

Middleware expects the layer to do some computation, augment the parameters and then call next.

A stack doesn't do anything unless you handle it. Express will handle the stack every time an incoming HTTP request is caught on the server. With middleware you handle the stack manually.

// express, you need to do nothing
// middleware
stack.handle(someData);

A more complete example :

var middleware = require("../src/middleware.js");

var stack = middleware(function(data, next) {
    data.foo = data.data*2;
    next();
}, function(data, next) {
    setTimeout(function() {
        data.async = true;
        next();
    }, 100)
}, function(data) {
    console.log(data);
});

stack.handle({
    "data": 42
})

In express terms you just define a stack of operations you want express to handle for every incoming HTTP request.

In terms of express (rather than connect) you have global middleware and route specific middleware. This means you can attach a middleware stack to every incoming HTTP requests or only attach it to HTTP requests that interact with a certain route.

Advanced examples of express & middleware :

// middleware 

var stack = middleware(function(req, res, next) {
    users.getAll(function(err, users) {
        if (err) next(err);
        req.users = users;
        next();  
    });
}, function(req, res, next) {
    posts.getAll(function(err, posts) {
        if (err) next(err);
        req.posts = posts;
        next();
    })
}, function(req, res, next) {
    req.posts.forEach(function(post) {
        post.user = req.users[post.userId];
    });

    res.render("blog/posts", {
        "posts": req.posts
    });
});

var app = express.createServer();

app.get("/posts", function(req, res) {
   stack.handle(req, res); 
});

// express

var app = express.createServer();

app.get("/posts", [
    function(req, res, next) {
        users.getAll(function(err, users) {
            if (err) next(err);
            req.users = users;
            next();  
        });
    }, function(req, res, next) {
        posts.getAll(function(err, posts) {
            if (err) next(err);
            req.posts = posts;
            next();
        })
    }, function(req, res, next) {
        req.posts.forEach(function(post) {
            post.user = req.users[post.userId];
        });

        res.render("blog/posts", {
            "posts": req.posts
        });
    }
], function(req, res) {
   stack.handle(req, res); 
});

I add a late answer to add something not mentioned in the previous answers.

By now it should be clear that middleware is/are function(s) run between the client request and the server answer. The most common middleware functionality needed are error managing, database interaction, getting info from static files or other resources. To move on the middleware stack the next callback must be called, you can see it in the end of middleware function to move to the next step in the flow.

You can use the app.use approach and have a flow like this:

var express = require('express'),
    app = express.createServer(),                                                                                                                                                 
    port = 1337;

function middleHandler(req, res, next) {
    console.log("execute middle ware");
    next();
}

app.use(function (req, res, next) {
    console.log("first middle ware");                                                                                                             
    next();
});

app.use(function (req, res, next) {
    console.log("second middle ware");                                                                                                             
    next();
});

app.get('/', middleHandler, function (req, res) {
    console.log("end middleware function");
    res.send("page render finished");
});

app.listen(port);
console.log('start server');

but you can also use another approach and pass each middleware as function arguments. Here is a example from the MooTools Nodejs website where midleware gets the Twitter, Github and Blog flow before the response is sent back to the client. Note how the functions are passed as arguments in app.get('/', githubEvents, twitter, getLatestBlog, function(req, res){. Using app.get will only be called for GET requests, app.use will be called for all requests.

// github, twitter & blog feeds
var githubEvents = require('./middleware/githubEvents')({
    org: 'mootools'
});
var twitter = require('./middleware/twitter')();
var blogData = require('./blog/data');
function getLatestBlog(req, res, next){
    blogData.get(function(err, blog) {
        if (err) next(err);
        res.locals.lastBlogPost = blog.posts[0];
        next();
    });
}

// home
app.get('/', githubEvents, twitter, getLatestBlog, function(req, res){
    res.render('index', {
        title: 'MooTools',
        site: 'mootools',
        lastBlogPost: res.locals.lastBlogPost,
        tweetFeed: res.locals.twitter
    });
});