convert html to rst POC

convert_html_to_rst.py

# -*- coding: utf-8 -*-

from io import StringIO
from lxml import etree


pre_str = """

.. code-block::

%s
"""

img_str = """

.. thumbnail:: %s
   :class: %s

"""

links_to_add = {}

def require_text(fn):
    def _require_text(elem):
        if not elem.text:
            return None
        return fn(elem)
    return _require_text

def convert_link(elem):
    links_to_add[elem.text] = elem.get('href', '')
    return '`%s`_' % elem.text

@require_text
def convert_strong(elem):
    return '**%s**' % elem.text

@require_text
def convert_pre(elem):
    pre_content = StringIO(elem.text.decode('utf-8'))
    def _convert_pre(text_io):
        for line in text_io:
            yield '   %s' % line
    return pre_str % u''.join(_convert_pre(pre_content))

def convert_br(elem):
    return '\n'

def convert_br_tail(elem):
    if elem.getparent().tag == 'pre':
        if elem.tail:
            return '   %s' % elem.tail
    if elem.tail:
        return elem.tail
    return ''

def convert_p(elem):
    if elem.tail is not None:
        return '\n%s' % elem.tail
    return '\n'

def convert_img(elem):
    return img_str % (elem.get('src'), elem.get('class'))

@require_text
def convert_null(elem):
    return elem.text

def convert_null_tail(elem):
    if elem.tail is not None:
        return elem.tail
    return None

def convert_pre_tail(elem):
    if elem.tail is not None:
        return '\n%s' % elem.tail
    return '\n'

converter_registry = {
        'a': convert_link,
        'strong': convert_strong,
        'pre': convert_pre,
        'br': convert_br,
        'img': convert_img,
        }

converter_registry_tail = {
        'p': convert_p,
        'br': convert_br_tail,
        'pre': convert_pre_tail,
        }

def get_root(html):
    parser = etree.HTMLParser()
    tree = etree.parse(StringIO(html), parser)
    return tree.getroot()

def recursive_runner(elem):
    converter = converter_registry.get(elem.tag, convert_null)
    res = converter(elem)
    if res:
        yield res
    for elm in elem.iterchildren():
        # there is no yield from in python2
        for x in recursive_runner(elm):
            yield x
    converter = converter_registry_tail.get(elem.tag, convert_null_tail)
    res = converter(elem)
    if res:
        yield res