I've written a simple web scraper for a comic website. I'm running it on Ubuntu (Linux ubuntu 4.18.0-16-generic #17~18.04.1-Ubuntu
) but when I execute the script (permissions set to chmod ug+x
) I keep getting a series of errors with imported system libraries along with a confusing syntax error:
import-im6.q16: not authorized `time' @ error/constitute.c/WriteImage/1037.
import-im6.q16: not authorized `os' @ error/constitute.c/WriteImage/1037.
import-im6.q16: not authorized `sys' @ error/constitute.c/WriteImage/1037.
import-im6.q16: not authorized `re' @ error/constitute.c/WriteImage/1037.
import-im6.q16: not authorized `requests' @ error/constitute.c/WriteImage/1037.
from: can't read /var/mail/bs4
./poorlywrittenscraper.py: line 15: DEFAULT_DIR_NAME: command not found
./poorlywrittenscraper.py: line 16: syntax error near unexpected token `('
./poorlywrittenscraper.py: line 16: `COMICS_DIRECTORY = os.path.join(os.getcwd(), DEFAULT_DIR_NAME)'
Interestingly enough, when I run the same script via python3
it fires up, creates the folder, fetches the images but... does not save them. o.O
Any idea what am I missing here or how to fix this?
Here's the full code of the script:
"""
A simple image downloader for poorlydrawnlines.com/archive
"""
import time
import os
import sys
import re
import concurrent.futures
import requests
from bs4 import BeautifulSoup as bs
DEFAULT_DIR_NAME = "poorly_created_folder"
COMICS_DIRECTORY = os.path.join(os.getcwd(), DEFAULT_DIR_NAME)
LOGO = """
a Python comic(al) scraper for poorlydwarnlines.com
__
.-----.-----.-----.----.| |.--.--.
| _ | _ | _ | _|| || | |
| __|_____|_____|__| |__||___ |
|__| |_____|
__ __ __
.--.--.--.----.|__| |_| |_.-----.-----.
| | | | _|| | _| _| -__| |
|________|__| |__|____|____|_____|__|__|
.-----.----.----.---.-.-----.-----.----.
|__ --| __| _| _ | _ | -__| _|
|_____|____|__| |___._| __|_____|__|
|__|
version: 0.4 | author: baduker | https://github.com/baduker
"""
ARCHIVE_URL = "http://www.poorlydrawnlines.com/archive/"
COMIC_PATTERN = re.compile(r'http://www.poorlydrawnlines.com/comic/.+')
def download_comics_menu(comics_found):
"""
Main download menu, takes number of available comics for download
"""
print("
The scraper has found {} comics.".format(len(comics_found)))
print("How many comics do you want to download?")
print("Type 0 to exit.")
while True:
try:
comics_to_download = int(input(">> "))
except ValueError:
print("Error: expected a number. Try again.")
continue
if comics_to_download > len(comics_found) or comics_to_download < 0:
print("Error: incorrect number of comics to download. Try again.")
continue
elif comics_to_download == 0:
sys.exit()
return comics_to_download
def grab_image_src_url(session, url):
"""
Fetches urls with the comic image source
"""
response = session.get(url)
soup = bs(response.text, 'html.parser')
for i in soup.find_all('p'):
for img in i.find_all('img', src=True):
return img['src']
def download_and_save_comic(session, url):
"""
Downloads and saves the comic image
"""
file_name = url.split('/')[-1]
with open(os.path.join(COMICS_DIRECTORY, file_name), "wb") as file:
response = session.get(url)
file.write(response.content)
def fetch_comics_from_archive(session):
"""
Grabs all urls from the poorlydrawnlines.com/archive and parses for only those that link to published comics
"""
response = session.get(ARCHIVE_URL)
soup = bs(response.text, 'html.parser')
comics = [url.get("href") for url in soup.find_all("a")]
return [url for url in comics if COMIC_PATTERN.match(url)]
def download_comic(session, url):
"""
Download progress information
"""
print("Downloading: {}".format(url))
url = grab_image_src_url(session, url)
download_and_save_comic(session, url)
def main():
"""
Encapsulates and executes all methods in the main function
"""
print(LOGO)
session = requests.Session()
comics = fetch_comics_from_archive(session)
comics_to_download = download_comics_menu(comics)
try:
os.mkdir(DEFAULT_DIR_NAME)
except OSError as exc:
sys.exit("Failed to create directory (error_no {})".format(exc.error_no))
start = time.time()
with concurrent.futures.ThreadPoolExecutor() as executor:
executor.map(lambda url: download_comic(session, url), comics[:comics_to_download])
executor.shutdown()
end = time.time()
print("Finished downloading {} comics in {:.2f} sec.".format(comics_to_download, end - start))
if __name__ in "__main__":
main()
See Question&Answers more detail:
os 与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…