|
| 1 | +import re |
| 2 | +import sys |
| 3 | + |
| 4 | + |
| 5 | +def get_urls(file_path): |
| 6 | + """[start method to fire extracting urls process] |
| 7 | + |
| 8 | + Arguments: |
| 9 | + file_path {[str]} -- [target text file path] |
| 10 | + """ |
| 11 | + text = read_text_file(file_path) |
| 12 | + urls = extract_urls(text) |
| 13 | + export_urls(urls, file_path) |
| 14 | + |
| 15 | + |
| 16 | +def read_text_file(file_path): |
| 17 | + """[summary] |
| 18 | + |
| 19 | + Arguments: |
| 20 | + file_path {[str]} -- [target text file path] |
| 21 | + |
| 22 | + Returns: |
| 23 | + [str] -- [file content to works on] |
| 24 | + """ |
| 25 | + with open(file_path) as f: |
| 26 | + text = f.read() |
| 27 | + return text |
| 28 | + |
| 29 | + |
| 30 | +def extract_urls(text): |
| 31 | + """[summary] |
| 32 | + |
| 33 | + Arguments: |
| 34 | + text {[str]} -- [file content to works on] |
| 35 | + |
| 36 | + Returns: |
| 37 | + [list] -- [extracted urls] |
| 38 | + """ |
| 39 | + url_regex_pattern = r"(?:(?:https?|ftp):\/\/)?[\w/\-?=%.]+\.[\w/\-?=%.]+" |
| 40 | + urls = re.findall(url_regex_pattern, text) |
| 41 | + return urls |
| 42 | + |
| 43 | + |
| 44 | +def export_urls(urls, file_path): |
| 45 | + """[summary] |
| 46 | + |
| 47 | + Arguments: |
| 48 | + urls {[list]} -- [extracted urls] |
| 49 | + file_path {[str]} -- [result text file path] |
| 50 | + """ |
| 51 | + with open(file_path.replace(".txt", "_links.txt"), "w") as f: |
| 52 | + f.write("\n".join(urls)) |
| 53 | + |
| 54 | + |
| 55 | +if __name__ == "__main__": |
| 56 | + file_path = sys.argv[1] |
| 57 | + get_urls(file_path) |
0 commit comments