Introduction

Though I am transitioning from teaching to data science, I still continue to teach via private tuition. I am currently helping somebody prepare for the STEP, so I wanted to obtain as many past papers and solutions as I could. To do this manually would have required hours of clicking on links and ‘Saving as…’, so I decided to automate it with python.

Downloading the step papers

This was straightforward. The website stepdatabase.com has past papers on them, and the urls and naming system they use is systematic and 100% consistent. I was able to download the papers using a simple loop with a wget command. See the code below.

Downloading the answers

This was less straightforward. This website provides answers to the STEP papers but the urls did not have a consistent format, so I could not just use wget again. Therefore, I used some web-scraping tools to help me.

There was a considerable learning curve, as I had not done any web scraping before. There were actually a couple of moments where I was going to give up. However, I persisted in the knowledge that if I want to be successful in a tech role, I will encounter such difficulties, and the only way to improve is to perservere.

In the end, I succeeded and the final code is given below. I will not go through the full process of trying things out, failing, tweaking, de-bugging, etc., but I will provide some of the key learning points for me.

  • BeautifulSoup is only suitable for statically generated pages. For dynamic ones, you can use Selenium.
  • The dir function in python is extremely handy. I have to thank this Kaggle tutorial for introducing me to this function. On two or three occasions, I wanted to do something, and by looking at the list of methods, I was able to find one which worked. The example I remember is find_elements_by_partial_link_text.
  • StackOverFlow is extremely handy. I am not sure what I would have done without it.
  • Don’t make assumptions. I assumed that the answers would all have different file names, but that was not always the case. This meant that previously downloaded were sometimes over-written by later answers. Fortunately, the was apparent in the first few minutes of running the program, so I could stop the program, and quickly modify it to add my own naming convention.

Code

The code to download the past papers is:

import subprocess

for i in range(87,119):
    for j in range(1,4):
        i = i%100
        url = f"https://stepdatabase.maths.org/database/db/{i:02}/{i:02}-S{j}.pdf"
        subprocess.run(["wget", url])

for i in range(87,119):
    for j in range(1,4):
        i = i%100
        url = f"https://stepdatabase.maths.org/database/db/{i:02}/{i:02}-S{j}.tex"
        subprocess.run(["wget", url])

The (ugly) code I created to download the solutions is:

from bs4 import BeautifulSoup
from selenium import webdriver
import requests
import re

# open browser and go to the url
browser = webdriver.Chrome()
URL = 'https://mei.org.uk/step-aea-solutions'
browser.get(URL)

# click button, which opens new tab, so move to new tab
browser.find_element_by_xpath('//button[text()="STEP Solutions"]').click()
browser.switch_to_window(browser.window_handles[1])

# click on link to go to next page
browser.find_element_by_link_text('STEP past paper worked solutions').click()

# obtain list of links in this page
# each of these refers to a group of step papers, e.g. STEP I, 2016-2019
elements = browser.find_elements_by_partial_link_text('STEP solutions')
groups = []
for element in elements:
    groups.append(element.get_attribute("href"))

# define regex pattern to help identify correct links
pattern = r"STEP (1|I)+: \d\d\d\d"

# loop through links in groups
for url_group in groups:
    # open the link
    browser.get(url_group)
    
    # obtain list of links and names of papers in this group.
    # this requires the regex pattern from above
    papers = []
    for paper in browser.find_elements_by_partial_link_text('STEP '):
        if re.match(pattern,paper.get_property('textContent')):
            papers.append([paper.get_attribute("href"), paper.get_property('textContent')])
    
    # loop through the list of papers
    for url_paper, paper_name in papers:
        # open link for an individual paper
        browser.get(url_paper)

        # obtain list of links for answers to individual questions
        questions = browser.find_elements_by_partial_link_text('Question ')

        # loop through the questions
        for question in questions:
            
            # open link to answer for individual question
            question.click()
            browser.switch_to_window(browser.window_handles[2])

            # download image
            # note that the url you end on is different to the url you use to get to this page
            url = browser.current_url
            filename = paper_name+'-'+url.split('/')[-1]
            r = requests.get(url, allow_redirects=True)
            open(filename, 'wb').write(r.content)

            # close browser and switch back to page with list of questions
            browser.close()
            browser.switch_to_window(browser.window_handles[1])