Chapitre 8. Traitement du HTML

8.1. Plonger

Je vois souvent sur comp.lang.python des questions comme «Comment faire une liste de tous les [en-têtes|images|liens] de mon document HTML ?» «Comment faire pour [parser|traduire|transformer] le texte d’un document HTML sans toucher aux balises ?» «Comment faire pour [ajouter|enlever|mettre entre guillemets] des attributs de mes balises HTML d’un coup ?» Ce chapitre répondra à toutes ces questions.

Voici un programme Python complet et fonctionnel en deux parties. La première partie, BaseHTMLProcessor.py, est un outil générique destiné à vous aider à traiter des fichiers HTML en parcourant les balises et les blocs de texte. La deuxième partie, dialect.py, est un exemple montrant comment utiliser BaseHTMLProcessor.py pour traduire le texte d’un document HTML sans toucher aux balises. Lisez les doc strings et les commentaires pour avoir une vue d’ensemble de ce qui se passe. Une grande partie va avoir l’air magique parce qu’il n’est pas évident de voir comment ces méthodes de classes sont appelées. Ne vous inquiétez pas, tout vous sera bientôt expliqué.

Exemple 8.1. `BaseHTMLProcessor.py`

Si vous ne l’avez pas déjà fait, vous pouvez télécharger cet exemple ainsi que les autres exemples du livre.


from sgmllib import SGMLParser
import htmlentitydefs

class BaseHTMLProcessor(SGMLParser):
    def reset(self):                       
        # extend (called by SGMLParser.__init__)
        self.pieces = []
        SGMLParser.reset(self)

    def unknown_starttag(self, tag, attrs):
        # called for each start tag
        # attrs is a list of (attr, value) tuples
        # e.g. for <pre class="screen">, tag="pre", attrs=[("class", "screen")]
        # Ideally we would like to reconstruct original tag and attributes, but
        # we may end up quoting attribute values that weren't quoted in the source
        # document, or we may change the type of quotes around the attribute value
        # (single to double quotes).
        # Note that improperly embedded non-HTML code (like client-side Javascript)
        # may be parsed incorrectly by the ancestor, causing runtime script errors.
        # All non-HTML code must be enclosed in HTML comment tags (<!-- code -->)
        # to ensure that it will pass through this parser unaltered (in handle_comment).
        strattrs = "".join([' %s="%s"' % (key, value) for key, value in attrs])
        self.pieces.append("<%(tag)s%(strattrs)s>" % locals())

    def unknown_endtag(self, tag):         
        # called for each end tag, e.g. for </pre>, tag will be "pre"
        # Reconstruct the original end tag.
        self.pieces.append("</%(tag)s>" % locals())

    def handle_charref(self, ref):         
        # called for each character reference, e.g. for "&#160;", ref will be "160"
        # Reconstruct the original character reference.
        self.pieces.append("&#%(ref)s;" % locals())

    def handle_entityref(self, ref):       
        # called for each entity reference, e.g. for "&copy;", ref will be "copy"
        # Reconstruct the original entity reference.
        self.pieces.append("&%(ref)s" % locals())
        # standard HTML entities are closed with a semicolon; other entities are not
        if htmlentitydefs.entitydefs.has_key(ref):
            self.pieces.append(";")

    def handle_data(self, text):           
        # called for each block of plain text, i.e. outside of any tag and
        # not containing any character or entity references
        # Store the original text verbatim.
        self.pieces.append(text)

    def handle_comment(self, text):        
        # called for each HTML comment, e.g. <!-- insert Javascript code here -->
        # Reconstruct the original comment.
        # It is especially important that the source document enclose client-side
        # code (like Javascript) within comments so it can pass through this
        # processor undisturbed; see comments in unknown_starttag for details.
        self.pieces.append("<!--%(text)s-->" % locals())

    def handle_pi(self, text):             
        # called for each processing instruction, e.g. <?instruction>
        # Reconstruct original processing instruction.
        self.pieces.append("<?%(text)s>" % locals())

    def handle_decl(self, text):
        # called for the DOCTYPE, if present, e.g.
        # <!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
        #     "http://www.w3.org/TR/html4/loose.dtd">
        # Reconstruct original DOCTYPE
        self.pieces.append("<!%(text)s>" % locals())

    def output(self):              
        """Return processed HTML as a single string"""
        return "".join(self.pieces)

Exemple 8.2. `dialect.py`


import re
from BaseHTMLProcessor import BaseHTMLProcessor

class Dialectizer(BaseHTMLProcessor):
    subs = ()

    def reset(self):
        # extend (called from __init__ in ancestor)
        # Reset all data attributes
        self.verbatim = 0
        BaseHTMLProcessor.reset(self)

    def start_pre(self, attrs):            
        # called for every <pre> tag in HTML source
        # Increment verbatim mode count, then handle tag like normal
        self.verbatim += 1                 
        self.unknown_starttag("pre", attrs)

    def end_pre(self):                     
        # called for every </pre> tag in HTML source
        # Decrement verbatim mode count
        self.unknown_endtag("pre")         
        self.verbatim -= 1                 

    def handle_data(self, text):                                        
        # override
        # called for every block of text in HTML source
        # If in verbatim mode, save text unaltered;
        # otherwise process the text with a series of substitutions
        self.pieces.append(self.verbatim and text or self.process(text))

    def process(self, text):
        # called from handle_data
        # Process text block by performing series of regular expression
        # substitutions (actual substitions are defined in descendant)
        for fromPattern, toPattern in self.subs:
            text = re.sub(fromPattern, toPattern, text)
        return text

class ChefDialectizer(Dialectizer):
    """convert HTML to Swedish Chef-speak
    
    based on the classic chef.x, copyright (c) 1992, 1993 John Hagerman
    """
    subs = ((r'a([nu])', r'u\1'),
            (r'A([nu])', r'U\1'),
            (r'a\B', r'e'),
            (r'A\B', r'E'),
            (r'en\b', r'ee'),
            (r'\Bew', r'oo'),
            (r'\Be\b', r'e-a'),
            (r'\be', r'i'),
            (r'\bE', r'I'),
            (r'\Bf', r'ff'),
            (r'\Bir', r'ur'),
            (r'(\w*?)i(\w*?)$', r'\1ee\2'),
            (r'\bow', r'oo'),
            (r'\bo', r'oo'),
            (r'\bO', r'Oo'),
            (r'the', r'zee'),
            (r'The', r'Zee'),
            (r'th\b', r't'),
            (r'\Btion', r'shun'),
            (r'\Bu', r'oo'),
            (r'\BU', r'Oo'),
            (r'v', r'f'),
            (r'V', r'F'),
            (r'w', r'w'),
            (r'W', r'W'),
            (r'([a-z])[.]', r'\1.  Bork Bork Bork!'))

class FuddDialectizer(Dialectizer):
    """convert HTML to Elmer Fudd-speak"""
    subs = ((r'[rl]', r'w'),
            (r'qu', r'qw'),
            (r'th\b', r'f'),
            (r'th', r'd'),
            (r'n[.]', r'n, uh-hah-hah-hah.'))

class OldeDialectizer(Dialectizer):
    """convert HTML to mock Middle English"""
    subs = ((r'i([bcdfghjklmnpqrstvwxyz])e\b', r'y\1'),
            (r'i([bcdfghjklmnpqrstvwxyz])e', r'y\1\1e'),
            (r'ick\b', r'yk'),
            (r'ia([bcdfghjklmnpqrstvwxyz])', r'e\1e'),
            (r'e[ea]([bcdfghjklmnpqrstvwxyz])', r'e\1e'),
            (r'([bcdfghjklmnpqrstvwxyz])y', r'\1ee'),
            (r'([bcdfghjklmnpqrstvwxyz])er', r'\1re'),
            (r'([aeiou])re\b', r'\1r'),
            (r'ia([bcdfghjklmnpqrstvwxyz])', r'i\1e'),
            (r'tion\b', r'cioun'),
            (r'ion\b', r'ioun'),
            (r'aid', r'ayde'),
            (r'ai', r'ey'),
            (r'ay\b', r'y'),
            (r'ay', r'ey'),
            (r'ant', r'aunt'),
            (r'ea', r'ee'),
            (r'oa', r'oo'),
            (r'ue', r'e'),
            (r'oe', r'o'),
            (r'ou', r'ow'),
            (r'ow', r'ou'),
            (r'\bhe', r'hi'),
            (r've\b', r'veth'),
            (r'se\b', r'e'),
            (r"'s\b", r'es'),
            (r'ic\b', r'ick'),
            (r'ics\b', r'icc'),
            (r'ical\b', r'ick'),
            (r'tle\b', r'til'),
            (r'll\b', r'l'),
            (r'ould\b', r'olde'),
            (r'own\b', r'oune'),
            (r'un\b', r'onne'),
            (r'rry\b', r'rye'),
            (r'est\b', r'este'),
            (r'pt\b', r'pte'),
            (r'th\b', r'the'),
            (r'ch\b', r'che'),
            (r'ss\b', r'sse'),
            (r'([wybdp])\b', r'\1e'),
            (r'([rnt])\b', r'\1\1e'),
            (r'from', r'fro'),
            (r'when', r'whan'))

def translate(url, dialectName="chef"):
    """fetch URL and translate using dialect
    
    dialect in ("chef", "fudd", "olde")"""
    import urllib                      
    sock = urllib.urlopen(url)         
    htmlSource = sock.read()           
    sock.close()                       
    parserName = "%sDialectizer" % dialectName.capitalize()
    parserClass = globals()[parserName]                    
    parser = parserClass()                                 
    parser.feed(htmlSource)
    parser.close()         
    return parser.output() 

def test(url):
    """test all dialects against URL"""
    for dialect in ("chef", "fudd", "olde"):
        outfile = "%s.html" % dialect
        fsock = open(outfile, "wb")
        fsock.write(translate(url, dialect))
        fsock.close()
        import webbrowser
        webbrowser.open_new(outfile)

if __name__ == "__main__":
    test("http://diveintopython.org/odbchelper_list.html")

Exemple 8.3. Sortie de `dialect.py`

Ce script effectue la traduction de la Section 3.2, «Présentation des listes» en pseudo-Chef Suédois (des Muppets), pseudo-Elmer Fudd (de Bugs Bunny) et en pseudo-ancien Anglais (librement adapté de The Canterbury Tales de Chaucer). Si vous regardez le source HTML de la sortie, vous verrez que toutes les balises HTML et les attributs sont intacts mais que le texte entre les balises a été «traduit» dans le pseudo-langage. Si vous regardez plus attentivement, vous verrez qu’en fait, seuls les titres et les paragraphes ont été traduits, les listing de code et les exemples d’écrans ont été laissé intacts.

<div class="abstract">
<p>Lists awe <span class="application">Pydon</span>'s wowkhowse datatype.
If youw onwy expewience wif wists is awways in
<span class="application">Visuaw Basic</span> ow (God fowbid) de datastowe
in <span class="application">Powewbuiwdew</span>, bwace youwsewf fow
<span class="application">Pydon</span> wists.</p>
</div>

Plongez au coeur de Python