def clean_text(self, html_text):
text_without_tags = w3lib.html.remove_tags(html_text)
text_without_escape_chars = w3lib.html.replace_escape_chars(
text_without_tags)
text_without_escape_chars = html.unescape(text_without_escape_chars)
text_without_whitespace = text_without_escape_chars.strip()
return text_without_whitespace
def filteHTML(string):
content = remove_comments(string) # 过滤注释
content = html.unescape(content) # 去掉实体字符
content = content.replace(' ', '')
content = content.replace(' ', '')
content = content.replace(' ', '')
return content