在Python中从字符串中剥离HTML

from mechanize import Browser
br = Browser()
br.open('http://somewebpage')
html = br.response().readlines()
for line in html:
  print line

When printing a line in an HTML file, I’m trying to find a way to only show the contents of each HTML element and not the formatting itself. If it finds '<a href="whatever.com">some text</a>', it will only print ‘some text’, '<b>hello</b>' prints ‘hello’, etc. How would one go about doing this?

I always used this function to strip HTML tags, as it requires only the Python stdlib:

For Python 3:

from io import StringIO
from html.parser import HTMLParser

class MLStripper(HTMLParser):
    def __init__(self):
        super().__init__()
        self.reset()
        self.strict = False
        self.convert_charrefs= True
        self.text = StringIO()
    def handle_data(self, d):
        self.text.write(d)
    def get_data(self):
        return self.text.getvalue()

def strip_tags(html):
    s = MLStripper()
    s.feed(html)
    return s.get_data()

For Python 2:

from HTMLParser import HTMLParser
from StringIO import StringIO

class MLStripper(HTMLParser):
    def __init__(self):
        self.reset()
        self.text = StringIO()
    def handle_data(self, d):
        self.text.write(d)
    def get_data(self):
        return self.text.getvalue()

def strip_tags(html):
    s = MLStripper()
    s.feed(html)
    return s.get_data()

I haven’t thought much about the cases it will miss, but you can do a simple regex:

re.sub('<[^<]+?>', '', text)

For those that don’t understand regex, this searches for a string <...>, where the inner content is made of one or more (+) characters that isn’t a <. The ? means that it will match the smallest string it can find. For example given <p>Hello</p>, it will match <'p> and </p> separately with the ?. Without it, it will match the entire string <..Hello..>.

If non-tag < appears in html (eg. 2 < 3), it should be written as an escape sequence &... anyway so the ^< may be unnecessary.

You can use BeautifulSoup get_text() feature.

from bs4 import BeautifulSoup

html_str = '''
<td><a href="http://www.fakewebsite.com">Please can you strip me?</a>
<br/><a href="http://www.fakewebsite.com">I am waiting....</a>
</td>
'''
soup = BeautifulSoup(html_str)

print(soup.get_text()) 
#or via attribute of Soup Object: print(soup.text)

It is advisable to explicitly specify the parser, for example as BeautifulSoup(html_str, features="html.parser"), for the output to be reproducible.

Short version!

import re, cgi
tag_re = re.compile(r'(<!--.*?-->|<[^>]*>)')

# Remove well-formed tags, fixing mistakes by legitimate users
no_tags = tag_re.sub('', user_input)

# Clean up anything else by escaping
ready_for_web = cgi.escape(no_tags)

Regex source: MarkupSafe. Their version handles HTML entities too, while this quick one doesn’t.

Why can’t I just strip the tags and leave it?

It’s one thing to keep people from <i>italicizing</i> things, without leaving is floating around. But it’s another to take arbitrary input and make it completely harmless. Most of the techniques on this page will leave things like unclosed comments (<!--) and angle-brackets that aren’t part of tags (blah <<<><blah) intact. The HTMLParser version can even leave complete tags in, if they’re inside an unclosed comment.

What if your template is {{ firstname }} {{ lastname }}? firstname = '<a' and lastname = 'href="http://evil.com/">' will be let through by every tag stripper on this page (except @Medeiros!), because they’re not complete tags on their own. Stripping out normal HTML tags is not enough.

Django’s strip_tags, an improved (see next heading) version of the top answer to this question, gives the following warning:

Absolutely NO guarantee is provided about the resulting string being HTML safe. So NEVER mark safe the result of a strip_tags call without escaping it first, for example with escape().

Follow their advice!

To strip tags with HTMLParser, you have to run it multiple times.

It’s easy to circumvent the top answer to this question.

Look at this string (source and discussion):

<img<!-- --> src=x onerror=alert(1);//><!-- -->

The first time HTMLParser sees it, it can’t tell that the <img...> is a tag. It looks broken, so HTMLParser doesn’t get rid of it. It only takes out the <!-- comments -->, leaving you with

<img src=x onerror=alert(1);//>

This problem was disclosed to the Django project in March, 2014. Their old strip_tags was essentially the same as the top answer to this question. Their new version basically runs it in a loop until running it again doesn’t change the string:

# _strip_once runs HTMLParser once, pulling out just the text of all the nodes.

def strip_tags(value):
    """Returns the given HTML with all tags stripped."""
    # Note: in typical case this loop executes _strip_once once. Loop condition
    # is redundant, but helps to reduce number of executions of _strip_once.
    while '<' in value and '>' in value:
        new_value = _strip_once(value)
        if len(new_value) >= len(value):
            # _strip_once was not able to detect more tags
            break
        value = new_value
    return value

Of course, none of this is an issue if you always escape the result of strip_tags().

Update 19 March, 2015: There was a bug in Django versions before 1.4.20, 1.6.11, 1.7.7, and 1.8c1. These versions could enter an infinite loop in the strip_tags() function. The fixed version is reproduced above. More details here.

Good things to copy or use

My example code doesn’t handle HTML entities – the Django and MarkupSafe packaged versions do.

My example code is pulled from the excellent MarkupSafe library for cross-site scripting prevention. It’s convenient and fast (with C speedups to its native Python version). It’s included in Google App Engine, and used by Jinja2 (2.7 and up), Mako, Pylons, and more. It works easily with Django templates from Django 1.7.

Django’s strip_tags and other html utilities from a recent version are good, but I find them less convenient than MarkupSafe. They’re pretty self-contained, you could copy what you need from this file.

If you need to strip almost all tags, the Bleach library is good. You can have it enforce rules like “my users can italicize things, but they can’t make iframes.”

Understand the properties of your tag stripper! Run fuzz tests on it! Here is the code I used to do the research for this answer.

sheepish note – The question itself is about printing to the console, but this is the top Google result for “python strip html from string”, so that’s why this answer is 99% about the web.

I needed a way to strip tags and decode HTML entities to plain text. The following solution is based on Eloff’s answer (which I couldn’t use because it strips entities).

from HTMLParser import HTMLParser
import htmlentitydefs

class HTMLTextExtractor(HTMLParser):
    def __init__(self):
        HTMLParser.__init__(self)
        self.result = [ ]

    def handle_data(self, d):
        self.result.append(d)

    def handle_charref(self, number):
        codepoint = int(number[1:], 16) if number[0] in (u'x', u'X') else int(number)
        self.result.append(unichr(codepoint))

    def handle_entityref(self, name):
        codepoint = htmlentitydefs.name2codepoint[name]
        self.result.append(unichr(codepoint))

    def get_text(self):
        return u''.join(self.result)

def html_to_text(html):
    s = HTMLTextExtractor()
    s.feed(html)
    return s.get_text()

A quick test:

html = u'<a href="#">Demo <em>(&not; \u0394&#x03b7;&#956;&#x03CE;)</em></a>'
print repr(html_to_text(html))

Result:

u'Demo (\xac \u0394\u03b7\u03bc\u03ce)'

Error handling:

• Invalid HTML structure may cause an HTMLParseError.
• Invalid named HTML entities (such as &#apos;, which is valid in XML and XHTML, but not plain HTML) will cause a ValueError exception.
• Numeric HTML entities specifying code points outside the Unicode range acceptable by Python (such as, on some systems, characters outside the Basic Multilingual Plane) will cause a ValueError exception.

Security note: Do not confuse HTML stripping (converting HTML into plain text) with HTML sanitizing (converting plain text into HTML). This answer will remove HTML and decode entities into plain text – that does not make the result safe to use in a HTML context.

Example: <script>alert("Hello");</script> will be converted to <script>alert("Hello");</script>, which is 100% correct behavior, but obviously not sufficient if the resulting plain text is inserted as-is into a HTML page.

The rule is not hard: Any time you insert a plain-text string into HTML output, you should always HTML escape it (using cgi.escape(s, True)), even if you “know” that it doesn’t contain HTML (e.g. because you stripped HTML content).

(However, the OP asked about printing the result to the console, in which case no HTML escaping is needed.)

Python 3.4+ version: (with doctest!)

import html.parser

class HTMLTextExtractor(html.parser.HTMLParser):
    def __init__(self):
        super(HTMLTextExtractor, self).__init__()
        self.result = [ ]

    def handle_data(self, d):
        self.result.append(d)

    def get_text(self):
        return ''.join(self.result)

def html_to_text(html):
    """Converts HTML to plain text (stripping tags and converting entities).
    >>> html_to_text('<a href="#">Demo<!--...--> <em>(&not; \u0394&#x03b7;&#956;&#x03CE;)</em></a>')
    'Demo (\xac \u0394\u03b7\u03bc\u03ce)'

    "Plain text" doesn't mean result can safely be used as-is in HTML.
    >>> html_to_text('&lt;script&gt;alert("Hello");&lt;/script&gt;')
    '<script>alert("Hello");</script>'

    Always use html.escape to sanitize text before using in an HTML context!

    HTMLParser will do its best to make sense of invalid HTML.
    >>> html_to_text('x < y &lt z <!--b')
    'x < y < z '

    Unrecognized named entities are included as-is. '&apos;' is recognized,
    despite being XML only.
    >>> html_to_text('&nosuchentity; &apos; ')
    "&nosuchentity; ' "
    """
    s = HTMLTextExtractor()
    s.feed(html)
    return s.get_text()

Note that HTMLParser has improved in Python 3 (meaning less code and better error handling).

There’s a simple way to this:

def remove_html_markup(s):
    tag = False
    quote = False
    out = ""

    for c in s:
            if c == '<' and not quote:
                tag = True
            elif c == '>' and not quote:
                tag = False
            elif (c == '"' or c == "'") and tag:
                quote = not quote
            elif not tag:
                out = out + c

    return out

The idea is explained here: http://youtu.be/2tu9LTDujbw

You can see it working here: http://youtu.be/HPkNPcYed9M?t=35s

PS – If you’re interested in the class(about smart debugging with python) I give you a link: http://www.udacity.com/overview/Course/cs259/CourseRev/1. It’s free!

You’re welcome! :)

If you need to preserve HTML entities (i.e. &), I added “handle_entityref” method to Eloff’s answer.

from HTMLParser import HTMLParser

class MLStripper(HTMLParser):
    def __init__(self):
        self.reset()
        self.fed = []
    def handle_data(self, d):
        self.fed.append(d)
    def handle_entityref(self, name):
        self.fed.append('&%s;' % name)
    def get_data(self):
        return ''.join(self.fed)

def html_to_text(html):
    s = MLStripper()
    s.feed(html)
    return s.get_data()

If you want to strip all HTML tags the easiest way I found is using BeautifulSoup:

from bs4 import BeautifulSoup  # Or from BeautifulSoup import BeautifulSoup

def stripHtmlTags(htmlTxt):
    if htmlTxt is None:
            return None
        else:
            return ''.join(BeautifulSoup(htmlTxt).findAll(text=True)) 

I tried the code of the accepted answer but I was getting “RuntimeError: maximum recursion depth exceeded”, which didn’t happen with the above block of code.

An lxml.html-based solution (lxml is a native library and can be more performant than a pure python solution).

Remove ALL tags

from lxml import html


## from file-like object or URL
tree = html.parse(file_like_object_or_url)

## from string
tree = html.fromstring('safe <script>unsafe</script> safe')

print(tree.text_content().strip())

### OUTPUT: 'safe unsafe safe'

Remove ALL tags with pre-sanitizing HTML (dropping some tags)

from lxml import html
from lxml.html.clean import clean_html

tree = html.fromstring("""<script>dangerous</script><span class="item-summary">
                            Detailed answers to any questions you might have
                        </span>""")

## text only
print(clean_html(tree).text_content().strip())

### OUTPUT: 'Detailed answers to any questions you might have'

Also see http://lxml.de/lxmlhtml.html#cleaning-up-html for what exactly the lxml.cleaner does.

If you need more control over what exactly is sanitized before converting to text then you might want to use the lxml Cleaner explicitly by passing the options you want in the constructor, e.g:

cleaner = Cleaner(page_structure=True,
                  meta=True,
                  embedded=True,
                  links=True,
                  style=True,
                  processing_instructions=True,
                  inline_style=True,
                  scripts=True,
                  javascript=True,
                  comments=True,
                  frames=True,
                  forms=True,
                  annoying_tags=True,
                  remove_unknown_tags=True,
                  safe_attrs_only=True,
                  safe_attrs=frozenset(['src','color', 'href', 'title', 'class', 'name', 'id']),
                  remove_tags=('span', 'font', 'div')
                  )
sanitized_html = cleaner.clean_html(unsafe_html)

If you need more control over how plain text is generated then instead of text_content() you can use lxml.etree.tostring:

plain_bytes = tostring(tree, method='text', encoding='utf-8')
print(plain.decode('utf-8'))

Here is a simple solution that strips HTML tags and decodes HTML entities based on the amazingly fast lxml library:

from lxml import html

def strip_html(s):
    return str(html.fromstring(s).text_content())

strip_html('Ein <a href="">sch&ouml;ner</a> Text.')  # Output: Ein schöner Text.

The Beautiful Soup package does this immediately for you.

from bs4 import BeautifulSoup

soup = BeautifulSoup(html)
text = soup.get_text()
print(text)

Here’s my solution for python 3.

import html
import re

def html_to_txt(html_text):
    ## unescape html
    txt = html.unescape(html_text)
    tags = re.findall("<[^>]+>",txt)
    print("found tags: ")
    print(tags)
    for tag in tags:
        txt=txt.replace(tag,'')
    return txt

Not sure if it is perfect, but solved my use case and seems simple.

You can use either a different HTML parser (like lxml, or Beautiful Soup) — one that offers functions to extract just text. Or, you can run a regex on your line string that strips out the tags. See Python docs for more.

I have used Eloff’s answer successfully for Python 3.1 [many thanks!].

I upgraded to Python 3.2.3, and ran into errors.

The solution, provided here thanks to the responder Thomas K, is to insert super().__init__() into the following code:

def __init__(self):
    self.reset()
    self.fed = []

… in order to make it look like this:

def __init__(self):
    super().__init__()
    self.reset()
    self.fed = []

… and it will work for Python 3.2.3.

Again, thanks to Thomas K for the fix and for Eloff’s original code provided above!

You can write your own function:

def StripTags(text):
     finished = 0
     while not finished:
         finished = 1
         start = text.find("<")
         if start >= 0:
             stop = text[start:].find(">")
             if stop >= 0:
                 text = text[:start] + text[start+stop+1:]
                 finished = 0
     return text

The solutions with HTML-Parser are all breakable, if they run only once:

html_to_text('<<b>script>alert("hacked")<</b>/script>

results in:

<script>alert("hacked")</script>

what you intend to prevent. if you use a HTML-Parser, count the Tags until zero are replaced:

from HTMLParser import HTMLParser

class MLStripper(HTMLParser):
    def __init__(self):
        self.reset()
        self.fed = []
        self.containstags = False

    def handle_starttag(self, tag, attrs):
       self.containstags = True

    def handle_data(self, d):
        self.fed.append(d)

    def has_tags(self):
        return self.containstags

    def get_data(self):
        return ''.join(self.fed)

def strip_tags(html):
    must_filtered = True
    while ( must_filtered ):
        s = MLStripper()
        s.feed(html)
        html = s.get_data()
        must_filtered = s.has_tags()
    return html

This is a quick fix and can be even more optimized but it will work fine. This code will replace all non empty tags with “” and strips all html tags form a given input text .You can run it using ./file.py input output

    #!/usr/bin/python
import sys

def replace(strng,replaceText):
    rpl = 0
    while rpl > -1:
        rpl = strng.find(replaceText)
        if rpl != -1:
            strng = strng[0:rpl] + strng[rpl + len(replaceText):]
    return strng


lessThanPos = -1
count = 0
listOf = []

try:
    #write File
    writeto = open(sys.argv[2],'w')

    #read file and store it in list
    f = open(sys.argv[1],'r')
    for readLine in f.readlines():
        listOf.append(readLine)         
    f.close()

    #remove all tags  
    for line in listOf:
        count = 0;  
        lessThanPos = -1  
        lineTemp =  line

            for char in lineTemp:

            if char == "<":
                lessThanPos = count
            if char == ">":
                if lessThanPos > -1:
                    if line[lessThanPos:count + 1] != '<>':
                        lineTemp = replace(lineTemp,line[lessThanPos:count + 1])
                        lessThanPos = -1
            count = count + 1
        lineTemp = lineTemp.replace("&lt","<")
        lineTemp = lineTemp.replace("&gt",">")                  
        writeto.write(lineTemp)  
    writeto.close() 
    print "Write To --- >" , sys.argv[2]
except:
    print "Help: invalid arguments or exception"
    print "Usage : ",sys.argv[0]," inputfile outputfile"

A python 3 adaption of søren-løvborg’s answer

from html.parser import HTMLParser
from html.entities import html5

class HTMLTextExtractor(HTMLParser):
    """ Adaption of http://stackoverflow.com/a/7778368/196732 """
    def __init__(self):
        super().__init__()
        self.result = []

    def handle_data(self, d):
        self.result.append(d)

    def handle_charref(self, number):
        codepoint = int(number[1:], 16) if number[0] in (u'x', u'X') else int(number)
        self.result.append(unichr(codepoint))

    def handle_entityref(self, name):
        if name in html5:
            self.result.append(unichr(html5[name]))

    def get_text(self):
        return u''.join(self.result)

def html_to_text(html):
    s = HTMLTextExtractor()
    s.feed(html)
    return s.get_text()

For one project, I needed so strip HTML, but also css and js. Thus, I made a variation of Eloffs answer:

class MLStripper(HTMLParser):
    def __init__(self):
        self.reset()
        self.strict = False
        self.convert_charrefs= True
        self.fed = []
        self.css = False
    def handle_starttag(self, tag, attrs):
        if tag == "style" or tag=="script":
            self.css = True
    def handle_endtag(self, tag):
        if tag=="style" or tag=="script":
            self.css=False
    def handle_data(self, d):
        if not self.css:
            self.fed.append(d)
    def get_data(self):
        return ''.join(self.fed)

def strip_tags(html):
    s = MLStripper()
    s.feed(html)
    return s.get_data()

Here’s a solution similar to the currently accepted answer (https://stackoverflow.com/a/925630/95989), except that it uses the internal HTMLParser class directly (i.e. no subclassing), thereby making it significantly more terse:

def strip_html(text):
    parts = []                                                                      
    parser = HTMLParser()                                                           
    parser.handle_data = parts.append                                               
    parser.feed(text)                                                               
    return ''.join(parts)

I’m parsing Github readmes and I find that the following really works well:

import re
import lxml.html

def strip_markdown(x):
    links_sub = re.sub(r'\[(.+)\]\([^\)]+\)', r'\1', x)
    bold_sub = re.sub(r'\*\*([^*]+)\*\*', r'\1', links_sub)
    emph_sub = re.sub(r'\*([^*]+)\*', r'\1', bold_sub)
    return emph_sub

def strip_html(x):
    return lxml.html.fromstring(x).text_content() if x else ''

And then

readme = """<img src="https://raw.githubusercontent.com/kootenpv/sky/master/resources/skylogo.png" />

            sky is a web scraping framework, implemented with the latest python versions in mind (3.4+). 
            It uses the asynchronous `asyncio` framework, as well as many popular modules 
            and extensions.

            Most importantly, it aims for **next generation** web crawling where machine intelligence 
            is used to speed up the development/maintainance/reliability of crawling.

            It mainly does this by considering the user to be interested in content 
            from *domains*, not just a collection of *single pages*
            ([templating approach](#templating-approach))."""

strip_markdown(strip_html(readme))

Removes all markdown and html correctly.

Using BeautifulSoup, html2text or the code from @Eloff, most of the time, it remains some html elements, javascript code…

So you can use a combination of these libraries and delete markdown formatting (Python 3):

import re
import html2text
from bs4 import BeautifulSoup
def html2Text(html):
    def removeMarkdown(text):
        for current in ["^[ #*]{2,30}", "^[ ]{0,30}\d\\\.", "^[ ]{0,30}\d\."]:
            markdown = re.compile(current, flags=re.MULTILINE)
            text = markdown.sub(" ", text)
        return text
    def removeAngular(text):
        angular = re.compile("[{][|].{2,40}[|][}]|[{][*].{2,40}[*][}]|[{][{].{2,40}[}][}]|\[\[.{2,40}\]\]")
        text = angular.sub(" ", text)
        return text
    h = html2text.HTML2Text()
    h.images_to_alt = True
    h.ignore_links = True
    h.ignore_emphasis = False
    h.skip_internal_links = True
    text = h.handle(html)
    soup = BeautifulSoup(text, "html.parser")
    text = soup.text
    text = removeAngular(text)
    text = removeMarkdown(text)
    return text

It works well for me but it can be enhanced, of course…

Simple code!. This will remove all kind of tags and content inside of it.

def rm(s):
    start=False
    end=False
    s=' '+s
    for i in range(len(s)-1):
        if i<len(s):
            if start!=False:
                if s[i]=='>':
                    end=i
                    s=s[:start]+s[end+1:]
                    start=end=False
            else:
                if s[i]=='<':
                    start=i
    if s.count('<')>0:
        self.rm(s)
    else:
        s=s.replace('&nbsp;', ' ')
        return s

But it won’t give full result if text contains <> symbols inside it.

# This is a regex solution.
import re
def removeHtml(html):
  if not html: return html
  # Remove comments first
  innerText = re.compile('<!--[\s\S]*?-->').sub('',html)
  while innerText.find('>')>=0: # Loop through nested Tags
    text = re.compile('<[^<>]+?>').sub('',innerText)
    if text == innerText:
      break
    innerText = text

  return innerText.strip()

This method works flawlessly for me and requires no additional installations:

import re
import htmlentitydefs

def convertentity(m):
    if m.group(1)=='#':
        try:
            return unichr(int(m.group(2)))
        except ValueError:
            return '&#%s;' % m.group(2)
        try:
            return htmlentitydefs.entitydefs[m.group(2)]
        except KeyError:
            return '&%s;' % m.group(2)

def converthtml(s):
    return re.sub(r'&(#?)(.+?);',convertentity,s)

html =  converthtml(html)
html.replace(" ", " ") ## Get rid of the remnants of certain formatting(subscript,superscript,etc).