1 | import unicodedata
|
---|
2 | from datetime import datetime
|
---|
3 | import psycopg2
|
---|
4 | import config_read
|
---|
5 | from bs4 import BeautifulSoup
|
---|
6 | from selenium import webdriver
|
---|
7 | import requests
|
---|
8 |
|
---|
9 | import sys
|
---|
10 |
|
---|
11 | file_path = 'outputfile.txt'
|
---|
12 | sys.stdout = open(file_path, "w")
|
---|
13 |
|
---|
14 | # Call to read the configuration file and connect to database
|
---|
15 | cinfo = config_read.get_databaseconfig("../postgresdb.config")
|
---|
16 | db_connection = psycopg2.connect(
|
---|
17 | database=cinfo[0],
|
---|
18 | host=cinfo[1],
|
---|
19 | user=cinfo[2],
|
---|
20 | password=cinfo[3]
|
---|
21 | )
|
---|
22 | cur = db_connection.cursor()
|
---|
23 |
|
---|
24 |
|
---|
25 | def scrape_function(driver1, i):
|
---|
26 | offer_shop = "Tehnomarket" # offer shop
|
---|
27 | last_updated = datetime.now().date()
|
---|
28 | is_validated = False
|
---|
29 |
|
---|
30 | tehnomarket_html = driver1.page_source
|
---|
31 | soup1 = BeautifulSoup(tehnomarket_html, 'html.parser')
|
---|
32 | active_li = soup1.find('div', {'class': 'adjust-elems pagination pagination-centered'}).find('li',
|
---|
33 | {'class': 'active'})
|
---|
34 |
|
---|
35 | print('page: ' + active_li.get_text())
|
---|
36 |
|
---|
37 | if int(active_li.get_text().strip()) == i:
|
---|
38 | phones = soup1.find('ul', {'class': 'products products-display-grid thumbnails'}).find_all('li', {
|
---|
39 | 'class': 'span4 product-fix'})
|
---|
40 |
|
---|
41 | for phone in phones:
|
---|
42 | offer_url = phone.find('a').get('href')
|
---|
43 | offer_name = phone.find('div', {'class': 'product-name'}).get_text().strip()
|
---|
44 | price = int(phone.find('div', {'class': 'product-price clearfix'}).find('strong') \
|
---|
45 | .get_text().replace('ден.', '').replace(',', '').strip())
|
---|
46 |
|
---|
47 | response2 = requests.get(offer_url)
|
---|
48 | soup2 = BeautifulSoup(response2.content, 'html.parser')
|
---|
49 |
|
---|
50 | image = soup2.find('div', {'id': 'product_gallery'}).find('img')
|
---|
51 |
|
---|
52 | image_url = None
|
---|
53 | if image is not None:
|
---|
54 | image_url = image.get('src')
|
---|
55 |
|
---|
56 | details = soup2.find('div', {'class': 'product-desc'}).get_text().split('\n')
|
---|
57 |
|
---|
58 | brand = details[2].strip().capitalize()
|
---|
59 | offer_shop_code = details[4].strip()
|
---|
60 |
|
---|
61 | specifications = []
|
---|
62 | for info in soup2.find_all('span', {'class': 'info'}):
|
---|
63 | specifications.append(info.get_text())
|
---|
64 |
|
---|
65 | print(brand)
|
---|
66 | print(offer_name)
|
---|
67 | print()
|
---|
68 | print()
|
---|
69 |
|
---|
70 | offer_description = '\n'.join(specifications)
|
---|
71 |
|
---|
72 | insert_script = 'INSERT INTO phone_offers (offer_shop, brand, offer_name, price, image_url, offer_url,' \
|
---|
73 | 'offer_description, offer_shop_code, last_updated, is_validated)' \
|
---|
74 | ' VALUES (%s, %s, %s, %s, %s, %s, %s, %s, %s, %s);'
|
---|
75 | insert_value = (offer_shop, brand, offer_name, price, image_url, offer_url, offer_description,
|
---|
76 | offer_shop_code, last_updated, is_validated)
|
---|
77 | cur.execute(insert_script, insert_value)
|
---|
78 | db_connection.commit()
|
---|
79 | else:
|
---|
80 | driver1.implicitly_wait(30)
|
---|
81 | scrape_function(driver1, i)
|
---|
82 |
|
---|
83 |
|
---|
84 | for i in range(1, 6):
|
---|
85 | tehnomarket_url = 'https://tehnomarket.com.mk/category/4109/mobilni-telefoni#page/' + str(i)
|
---|
86 | # print(anhoch_url)
|
---|
87 |
|
---|
88 | # selenium is used because of the dynamic content of the page
|
---|
89 | driver1 = webdriver.Safari(executable_path='/usr/bin/safaridriver')
|
---|
90 | driver1.get(tehnomarket_url)
|
---|
91 |
|
---|
92 | scrape_function(driver1, i)
|
---|
93 | # closing the driver so the safari instance can pair with another webdriver session
|
---|
94 | driver1.close()
|
---|
95 |
|
---|
96 | cur.close()
|
---|
97 | db_connection.close()
|
---|