用Javascript大批量收集网站数据

最近为了写论文，要大批量收集慕课网的相关用户数据（因为用户个人主页是公开的），故而写了一个插件进行收集。需要在慕课网控制台输入。最后收集了3000多份数据。

/*

收集项	收集标准
用户编号	慕课网用户编号是七位数递增的，范围在5000000以内。故而本次收集将在1000000至5000000内分区段进行数据的随机收集。
学习时长	慕课有记录用户的总共学习时长，如110时34分。为方便后期统计，将该数据转化为以分钟为单位的数据。
积分	积分反映的是用户参与网站社交的程度。
经验	相比学习时长，经验更能直接反映课程完成程度。
学习课程数	为避免用户参加课程而没有进行学习，课程学习数以学习进度超过5%的课程为准入标准。
平均完成度	指学生课程的平均完成度。
完成课程数	为课程完成度80%及以上的课程。
参加路径数	慕课网的学习路径的参与数目。
路径完成度	学习路径的完成程度。

*/



var userInfo = function(user,link,i){
    var user = user;
    var link = link;
    this.name = i;
    var _this_ = this;
    this.getBasicInfo(user);
    this.getCourseNum(user);
    this.getRoute(user,link);

    setTimeout(function(){
        console.log(_this_.name+"	"+_this_.learnTime+"	"+_this_.credit+"	"+_this_.mp+"	"+_this_.CourseNum+"	"+_this_.learnAve+"	"+_this_.finish+"	"+_this_.routeNum+"	"+_this_.routeAve);
        user.close();
    },10000);
};

userInfo.prototype ={
//便捷取值函数;
getElem : function(user,className,tag) {
            var infor =user.document.getElementsByClassName(className)[0].getElementsByTagName(tag)[0].innerHTML;
            return infor;
        },
//获取基本信息：名字、学习时长、积分、经验值。
getBasicInfo:function(d){
                    var user = d;
                    //get time;
                    var time =this.getElem(user,"u-info-learn","em");
                    var cutHour = /d+(?=W{2})/g;
                    var hour = parseInt(time.match(cutHour))*60;
                    var cutMinute = /d+(?!W{2})/g; 
                    var minute = parseInt(time.match(cutMinute));
                    if(isNaN(hour)){
                        var learnTime = minute;
                    }else{
                        var learnTime = hour+minute;
                    }
                    this.learnTime = learnTime;
                    //get credit;
                    var credit = parseFloat(this.getElem(user,"u-info-credit","em"));
                    this.credit = credit;
                    //get mp;
                    var mp = parseFloat(this.getElem(user,"u-info-mp","em"));
                    this.mp = mp;

    },
//获得课程数量和课程完成度
getCourseNum:function(d){
                var self = this;
                var user = d;
                var CourseNum = 0,
                    learnSum = 0,
                    finish = 0;

                //获得一个页面的课程数量和完成度
                
                function getOnepage(obj){
                    if(obj.document.getElementsByClassName("course-one").length > 0){
                        var course = obj.document.getElementsByClassName("course-one");
                        for(i = 0;i<course.length;i++){
                            var    hasLearn = course[i].getElementsByClassName("i-left")[0].innerHTML,
                                cutWord = /WW/g;
                            var hasLearn =parseFloat(hasLearn.replace(cutWord,""));
                            if(hasLearn > 5){
                                CourseNum = CourseNum + 1;
                                learnSum = learnSum + hasLearn;
                                if(hasLearn > 80){
                                    finish = finish + 1;
                                }
                            };
                        };
                        self.CourseNum = CourseNum;
                        if(CourseNum !== 0){
                        self.learnAve = (learnSum/CourseNum).toFixed(2);
                        }
                        self.finish = finish;
                    }
                };

                //获取当前页面
                
                getOnepage(user);

                //获取其他页面
                if(user.document.getElementsByClassName("text-page-tag").length > 0){
                    var pages = user.document.getElementsByClassName("text-page-tag");
                    if(pages.length > 1){
                        for(i=1;i<pages.length;i++){
                            !function(i){
                            var page = window.open(pages[i].getAttribute("href"));
                                setTimeout(function get(){
                                    if(page.document.getElementsByClassName("course-one").length > 0){
                                            getOnepage(page);
                                            page.close();
                                            self.CourseNum = CourseNum;
                                            self.learnAve = (learnSum/CourseNum).toFixed(2);
                                            self.finish = finish;
                                    }else{
                                        console.log("page" + i +"didn't load!");
                                    }
                                },5000);//要若有些页面没有在4S内打开，导致无法取值怎么办？
                            }(i);
                        };
                    };
                }

                //获取其他页面列表 
                if(user.document.getElementsByClassName("page").length > 0){
                var lastPage =user.document.getElementsByClassName("page")[0].lastChild.getAttribute("href");
                var allPage = parseInt(lastPage.match(/d$/));
                if(allPage>7){
                    var otherLink = lastPage.match(/^.*=(?=d)/g);
                    for(i=8;i <= allPage;i++){    
                        !function(i){
                        var page = window.open(otherLink+i);
                            setTimeout(function get(){
                                if(page.document.getElementsByClassName("course-one").length > 0){
                                        getOnepage(page);
                                        page.close();
                                        self.CourseNum = CourseNum;
                                        self.learnAve = (learnSum/CourseNum).toFixed(2);
                                        self.finish = finish;
                                }else{
                                    console.log("page" + i +"didn't load!");
                                }
                            },4000);//记得找个好的网络，防止有些页面没有在4S内打开，导致无法取值。
                        }(i);
                    };
                };
                }

    },
//获取参加路径数
getRoute : function(d,link){
            var user = d;
            var self = this;
            var currPage = link;
            var openPage = window.open(currPage.replace("courses","plans"))
            setTimeout(function(){
                if(openPage.document.getElementsByClassName("plans-item").length > 0){
                    var route = openPage.document.getElementsByClassName("plans-item");
                    var routeNum = route.length;
                    var routeLearn = openPage.document.getElementsByClassName("plans-list-progress");
                    var routeSum = 0,learnSum = 0;
                    for(i=0;i<routeNum;i++){
                        var hasLearn = parseFloat(routeLearn[i].innerHTML.replace(/W+/,""));
                        if(hasLearn>2){
                            routeSum = routeSum + hasLearn;
                            learnSum +=1;
                        };
                    };
                    if(learnSum > 0 ){
                        routeAve = routeSum/learnSum;
                        self.routeAve = routeAve;
                    }
                    self.routeNum =routeNum;
                };
                openPage.close();
            },5000)
            
    }
};

//打开页面
var n = Math.round(Math.random()*1000000+4000000);//取3000000-4000000之间的随机数
var min = n;
function openLink(){
    if(n< min+1000){ //取1000份数据
        var link = "http://www.imooc.com/u/"+ n +"/courses";
        var user = window.open(link);
        setTimeout(function(){
            if(user.document.getElementsByClassName("user-name").length > 0){
                new userInfo(user,link,n);
            }else{
                console.log(n+" page didn't exit!")
            };
        },4000);
        setTimeout(function(){n = n+1;openLink();},4000)
    }
};
openLink();