如何用Python下载百度指数的数据

大家好我是小小明，今天给大家演示如何使用python直接采集百度指数的数据。

百度指数(Baidu Index) 是以百度海量网民行为数据为基础的数据分析平台，它能够能够告诉用户：某个关键词在百度的搜索规模有多大，一段时间内的涨跌态势以及相关的新闻舆论变化，关注这些词的网民是什么样的，分布在哪里，同时还搜了哪些相关的词。

百分十先生分享过如何使用uiautomation采集百度指数：百度指数 如何批量获取？

不过个人感觉这方法好像有点杀鸡用牛刀，对于网页使用selenium完全足以，当然对于专门针对selenium进行反爬检测的网页就需要特殊修改。

本文不演示如何使用UI自动化工具采集百度指数，为了采集更简单将直接读取并解析接口。

关于uiautomation，PC端的UI自动化可以查看教程：Windows桌面程序自动化控制之uiautomation模块全面讲解

打开百度指数发现查看指数必须要先登录，比如我们对比一个python和Java最近一周的指数：

当鼠标移动到每天的坐标上时会显示当天的数据，例如：

如果我们采用UI自动化的方式，至少得模拟移动到每天的坐标。

打开开发者工具，重新查询发现获取数据的接口：

实际的指数数据就存储在这个data字段中，但是以某种加密方式加密了。

然后注意第二个接口的某个参数与当前接口返回的数据某个值一致。

此时我全局搜索decrypt，找到了加密函数：

此时打上断点重新搜索，可以看到传入该函数的t参数与ptbk接口返回的值一致：

说明我们只需要将这段js翻译为python来解密加密数据即可。

下面我们总结一下指数数据获取的思路：

1. 通过index接口获取uniqid和加密后的指数数据userIndexes

2. 通过ptbk接口传入uniqid获取密钥key

3. 通过解密函数根据密钥key解密userIndexes

下面我们分别用代码来实现，首先获取指数数据：

import requestsimport jsonheaders = {    "Connection": "keep-alive",    "Accept": "application/json, text/plain, */*",    "User-Agent": "Mozilla/5.0 (Windows NT 10.0; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/86.0.4240.198 Safari/537.36",    "Sec-Fetch-Site": "same-origin",    "Sec-Fetch-Mode": "cors",    "Sec-Fetch-Dest": "empty",    "Referer": "https://index.baidu.com/v2/main/index.html",    "Accept-Language": "zh-CN,zh;q=0.9",    "Cookie": cookie,}words = "[[{"name":"python","wordType":1}],[{"name":"java","wordType":1}]]"start = "2021-11-15"end = "2021-11-21"url = f"http://index.baidu.com/api/SearchApi/index?area=0&word={words}&area=0&startDate={start}&endDate={end}"res = requests.get(url, headers=headers)data = res.json()["data"]data

cookie需要在登录后复制粘贴获取，就是请求中的这段字符串（直接复制粘贴即可）：

结果：

{"userIndexes": [{"word": [{"name": "python", "wordType": 1}],   "all": {"startDate": "2021-11-15",    "endDate": "2021-11-21",    "data": "WQ3Q-nWQ.yGnWQ.y3nW3yQsnWW.Q-nysXV3ny.-VG"},   "pc": {"startDate": "2021-11-15",    "endDate": "2021-11-21",    "data": "y3yVXny3yWyny3GWWny3QyVnyQG33nXGsQn-..G"},   "wise": {"startDate": "2021-11-15",    "endDate": "2021-11-21",    "data": "XWVXnXQ-XnX3XWnX-WynX3X3n--XynsQyG"},   "type": "day"},  {"word": [{"name": "java", "wordType": 1}],   "all": {"startDate": "2021-11-15",    "endDate": "2021-11-21",    "data": "-XW.n-ssXnXG3GnXG..nXyyGnVQyWn.QQQ"},   "pc": {"startDate": "2021-11-15",    "endDate": "2021-11-21",    "data": ".VVVn.3Xsn.XX3n.-VWn.sW3nQG-snWVWQ"},   "wise": {"startDate": "2021-11-15",    "endDate": "2021-11-21",    "data": "QW.XnQW-WnQG3VnQyXQnQQ-VnQWW.nWsyG"},   "type": "day"}], "generalRatio": [{"word": [{"name": "python", "wordType": 1}],   "all": {"avg": 21565, "yoy": -24, "qoq": 7},   "pc": {"avg": 12470, "yoy": -32, "qoq": 3},   "wise": {"avg": 9095, "yoy": -10, "qoq": 12}},  {"word": [{"name": "java", "wordType": 1}],   "all": {"avg": 8079, "yoy": -23, "qoq": 11},   "pc": {"avg": 4921, "yoy": -33, "qoq": 6},   "wise": {"avg": 3157, "yoy": "-", "qoq": 18}}], "uniqid": "5f0a123915325e28d9f055409955c9ad"}

这些数据中，wise表示移动端，all表示pc端+移动端。userIndexes是指数详情数据，generalRatio是概览数据。

下面我们只关心各个关键字的整体表现。

下面我们获取uniqid并获取ptbk：

uniqid = data["uniqid"]res = requests.get(    f"http://index.baidu.com/Interface/ptbk?uniqid={uniqid}", headers=headers)ptbk = res.json()["data"]ptbk

"LV.7yF-s30WXGQn.65+1-874%2903,"

下面我将下面这段Js代码翻译为python：

decrypt: function(t, e) {	if (t) {		for (var n = t.split(""), i = e.split(""), a = {}, r = [], o = 0; o < n.length / 2; o++)			a[n[o]] = n[n.length / 2 + o];		for (var s = 0; s < e.length; s++)			r.push(a[i[s]]);		return r.join("")	}}

python代码：

def decrypt(ptbk, index_data):    n = len(ptbk)//2    a = dict(zip(ptbk[:n], ptbk[n:]))    return "".join([a[s] for s in index_data])

然后我们遍历每个关键字解密出对应的指数数据：

for userIndexe in data["userIndexes"]:    name = userIndexe["word"][0]["name"]    index_data = userIndexe["all"]["data"]    r = decrypt(ptbk, index_data)    print(name, r)

python 23438,23510,23514,24137,22538,17964,15860java 8925,8779,9040,9055,9110,6312,5333

检查实际网页中的数据发现确实一致：

那么我们就可以轻松获取任意指定关键字的指数数据。下面我将其整体封装一下，完整代码为：

import requestsimport jsonfrom datetime import date, timedeltaheaders = {    "Connection": "keep-alive",    "Accept": "application/json, text/plain, */*",    "User-Agent": "Mozilla/5.0 (Windows NT 10.0; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/86.0.4240.198 Safari/537.36",    "Sec-Fetch-Site": "same-origin",    "Sec-Fetch-Mode": "cors",    "Sec-Fetch-Dest": "empty",    "Referer": "https://index.baidu.com/v2/main/index.html",    "Accept-Language": "zh-CN,zh;q=0.9",    "Cookie": cookie,}def decrypt(ptbk, index_data):    n = len(ptbk)//2    a = dict(zip(ptbk[:n], ptbk[n:]))    return "".join([a[s] for s in index_data])def get_index_data(keys, start=None, end=None):    words = [[{"name": key, "wordType": 1}] for key in keys]    words = str(words).replace(" ", "").replace(""", "/"")    today = date.today()    if start is None:        start = str(today-timedelta(days=8))    if end is None:        end = str(today-timedelta(days=2))    url = f"http://index.baidu.com/api/SearchApi/index?area=0&word={words}&area=0&startDate={start}&endDate={end}"    print(words, start, end)    res = requests.get(url, headers=headers)    data = res.json()["data"]    uniqid = data["uniqid"]    url = f"http://index.baidu.com/Interface/ptbk?uniqid={uniqid}"    res = requests.get(url, headers=headers)    ptbk = res.json()["data"]    result = {}    result["startDate"] = start    result["endDate"] = end    for userIndexe in data["userIndexes"]:        name = userIndexe["word"][0]["name"]        tmp = {}        index_all = userIndexe["all"]["data"]        index_all_data = [int(e) for e in decrypt(ptbk, index_all).split(",")]        tmp["all"] = index_all_data        index_pc = userIndexe["pc"]["data"]        index_pc_data = [int(e) for e in decrypt(ptbk, index_pc).split(",")]        tmp["pc"] = index_pc_data        index_wise = userIndexe["wise"]["data"]        index_wise_data = [int(e)                           for e in decrypt(ptbk, index_wise).split(",")]        tmp["wise"] = index_wise_data        result[name] = tmp    return result

测试一下：

get_index_data(["python", "java"])

{"startDate": "2021-11-15", "endDate": "2021-11-21", "python": {"all": [23438, 23510, 23514, 24137, 22538, 17964, 15860],  "pc": [14169, 14121, 14022, 14316, 13044, 9073, 8550],  "wise": [9269, 9389, 9492, 9821, 9494, 8891, 7310]}, "java": {"all": [8925, 8779, 9040, 9055, 9110, 6312, 5333],  "pc": [5666, 5497, 5994, 5862, 5724, 3087, 2623],  "wise": [3259, 3282, 3046, 3193, 3386, 3225, 2710]}}

结果非常不错。