有没有办法优化for循环？Selenium 需要很长时间才能抓取 38 页

Question

我正在尝试通过 Selenium 和 python抓取https://arxiv.org/search/?query=healthcare&searchtype=allI 。for 循环执行时间太长。我尝试使用无头浏览器和 PhantomJS 进行抓取，但它不会抓取抽象字段（需要通过单击更多按钮来扩展抽象字段）import pandas as pdimport seleniumimport reimport timefrom selenium.common.exceptions import NoSuchElementExceptionfrom selenium.webdriver import Firefoxbrowser = Firefox()url_healthcare = 'https://arxiv.org/search/?query=healthcare&searchtype=all'browser.get(url_healthcare)dfs = []for i in range(1, 39):    articles = browser.find_elements_by_tag_name('li[class="arxiv-result"]')    for article in articles:        title = article.find_element_by_tag_name('p[class="title is-5 mathjax"]').text        arxiv_id = article.find_element_by_tag_name('a').text.replace('arXiv:','')        arxiv_link = article.find_elements_by_tag_name('a')[0].get_attribute('href')         pdf_link = article.find_elements_by_tag_name('a')[1].get_attribute('href')        authors = article.find_element_by_tag_name('p[class="authors"]').text.replace('Authors:','')        try:                link1 = browser.find_element_by_link_text('▽ More')                link1.click()        except:                time.sleep(0.1)        abstract = article.find_element_by_tag_name('p[class="abstract mathjax"]').text        date = article.find_element_by_tag_name('p[class="is-size-7"]').text        date = re.split(r"Submitted|;",date)[1]        tag = article.find_element_by_tag_name('div[class="tags is-inline-block"]').text.replace('\n', ',')                try:            doi = article.find_element_by_tag_name('div[class="tags has-addons"]').text            doi = re.split(r'\s', doi)[1]         except NoSuchElementException:            doi = 'None'        all_combined = [title, arxiv_id, arxiv_link, pdf_link, authors, abstract, date, tag, doi]        dfs.append(all_combined)    print('Finished Extracting Page:', i)

Answer 1

以下实现在16 秒内实现了这一目标。

为加快执行进程，我采取了以下措施：

• 完全删除Selenium（无需点击）

• 对于abstract, 使用BeautifulSoup的输出并稍后对其进行处理

• 添加multiprocessing以显着加快该过程

from multiprocessing import Process, Manager

import requests 

from bs4 import BeautifulSoup

import re

import time

start_time = time.time()

def get_no_of_pages(showing_text):

    no_of_results = int((re.findall(r"(\d+,*\d+) results for all",showing_text)[0].replace(',','')))

    pages = no_of_results//200 + 1

    print("total pages:",pages)

    return pages 

def clean(text):

    return text.replace("\n", '').replace("  ",'')

def get_data_from_page(url,page_number,data):

    print("getting page",page_number)

    response = requests.get(url+"start="+str(page_number*200))

    soup = BeautifulSoup(response.content, "lxml")

    

    arxiv_results = soup.find_all("li",{"class","arxiv-result"})

    for arxiv_result in arxiv_results:

        paper = {} 

        paper["titles"]= clean(arxiv_result.find("p",{"class","title is-5 mathjax"}).text)

        links = arxiv_result.find_all("a")

        paper["arxiv_ids"]= links[0].text.replace('arXiv:','')

        paper["arxiv_links"]= links[0].get('href')

        paper["pdf_link"]= links[1].get('href')

        paper["authors"]= clean(arxiv_result.find("p",{"class","authors"}).text.replace('Authors:',''))

        split_abstract = arxiv_result.find("p",{"class":"abstract mathjax"}).text.split("▽ More\n\n\n",1)

        if len(split_abstract) == 2:

            paper["abstract"] = clean(split_abstract[1].replace("△ Less",''))

        else: 

            paper["abstract"] = clean(split_abstract[0].replace("△ Less",''))

        paper["date"] = re.split(r"Submitted|;",arxiv_results[0].find("p",{"class":"is-size-7"}).text)[1]

        paper["tag"] = clean(arxiv_results[0].find("div",{"class":"tags is-inline-block"}).text) 

        doi = arxiv_results[0].find("div",{"class":"tags has-addons"})       

        if doi is None:

            paper["doi"] = "None"

        else:

            paper["doi"] = re.split(r'\s', doi.text)[1] 

        data.append(paper)

    

    print(f"page {page_number} done")

if __name__ == "__main__":

    url = 'https://arxiv.org/search/?searchtype=all&query=healthcare&abstracts=show&size=200&order=-announced_date_first&'

    response = requests.get(url+"start=0")

    soup = BeautifulSoup(response.content, "lxml")

    with Manager() as manager:

        data = manager.list()  

        processes = []

        get_data_from_page(url,0,data)

        showing_text = soup.find("h1",{"class":"title is-clearfix"}).text

        for i in range(1,get_no_of_pages(showing_text)):

            p = Process(target=get_data_from_page, args=(url,i,data))

            p.start()

            processes.append(p)

        for p in processes:

            p.join()

        print("Number of entires scraped:",len(data))

        stop_time = time.time()

        print("Time taken:", stop_time-start_time,"seconds")

输出：

>>> python test.py

getting page 0

page 0 done

total pages: 10

getting page 1

getting page 4

getting page 2

getting page 6

getting page 5

getting page 3

getting page 7

getting page 9

getting page 8

page 9 done

page 4 done

page 1 done

page 6 done

page 2 done

page 7 done

page 3 done

page 5 done

page 8 done

Number of entires scraped: 1890

Time taken: 15.911492586135864 seconds

Answer 2

您可以根据要求尝试一下美丽的汤做法。无需点击更多链接。

from requests import get

from bs4 import BeautifulSoup

# you can change the size to retrieve all the results at one shot.

url = 'https://arxiv.org/search/?query=healthcare&searchtype=all&abstracts=show&order=-announced_date_first&size=50&start=0'

response = get(url,verify = False)

soup = BeautifulSoup(response.content, "lxml")

#print(soup)

queryresults = soup.find_all("li", attrs={"class": "arxiv-result"})

for result in queryresults:

    title = result.find("p",attrs={"class": "title is-5 mathjax"})

    print(title.text)

#If you need full abstract content - try this (you do not need to click on more button

    for result in queryresults:

        abstractFullContent = result.find("span",attrs={"class": "abstract-full has-text-grey-dark mathjax"})

        print(abstractFullContent.text)

输出：

 Interpretable Deep Learning for Automatic Diagnosis of 12-lead Electrocardiogram

            

  Leveraging Technology for Healthcare and Retaining Access to Personal Health Data to Enhance Personal Health and Well-being

  Towards new forms of particle sensing and manipulation and 3D imaging on a smartphone for healthcare applications