def transform(self, html=None, pretty_print=True, **kwargs):
"""change the html and return it with CSS turned into style
attributes.
"""
if html is not None and self.html is not None:
raise TypeError("Can't pass html argument twice")
elif html is None and self.html is None:
raise TypeError("must pass html as first argument")
elif html is None:
html = self.html
if hasattr(html, "getroottree"):
# skip the next bit
root = html.getroottree()
page = root
tree = root
else:
if self.method == "xml":
parser = etree.XMLParser(ns_clean=False,
resolve_entities=False)
else:
parser = etree.HTMLParser()
stripped = html.strip()
tree = etree.fromstring(stripped, parser).getroottree()
page = tree.getroot()
# lxml inserts a doctype if none exists, so only include it in
# the root if it was in the original html.
root = tree if stripped.startswith(tree.docinfo.doctype) else page
assert page is not None
if self.disable_leftover_css:
head = None
else:
head = get_or_create_head(tree)
#
# style selectors
#
rules = []
index = 0
cssselector = ["style"]
if self.allow_network:
cssselector.append("link[rel~=stylesheet]")
for element in _create_cssselector(",".join(cssselector))(page):
# If we have a media attribute whose value is anything other than
# 'all' or 'screen', ignore the ruleset.
media = element.attrib.get("media")
if media and media not in ("all", "screen"):
continue
data_attribute = element.attrib.get(self.attribute_name)
if data_attribute:
if data_attribute == "ignore":
del element.attrib[self.attribute_name]
continue
else:
warnings.warn("Unrecognized %s attribute (%r)" %
(self.attribute_name, data_attribute))
is_style = element.tag == "style"
if is_style:
css_body = element.text
else:
href = element.attrib.get("href")
css_body = self._load_external(href)
these_rules, these_leftover = self._parse_style_rules(
css_body, index)
index += 1
rules.extend(these_rules)
parent_of_element = element.getparent()
if these_leftover or self.keep_style_tags:
if is_style:
style = element
else:
style = etree.Element("style")
style.attrib["type"] = "text/css"
if self.keep_style_tags:
style.text = css_body
else:
style.text = self._css_rules_to_string(these_leftover)
if self.method == "xml":
style.text = etree.CDATA(style.text)
if not is_style:
element.addprevious(style)
parent_of_element.remove(element)
elif not self.keep_style_tags or not is_style:
parent_of_element.remove(element)
# external style files
if self.external_styles and self.allow_network:
for stylefile in self.external_styles:
css_body = self._load_external(stylefile)
self._process_css_text(css_body, index, rules, head)
index += 1
# css text
if self.css_text:
for css_body in self.css_text:
self._process_css_text(css_body, index, rules, head)
index += 1
# rules is a tuple of (specificity, selector, styles), where
# specificity is a tuple ordered such that more specific
# rules sort larger.
rules.sort(key=operator.itemgetter(0))
# collecting all elements that we need to apply rules on
# id is unique for the lifetime of the object
# and lxml should give us the same everytime during this run
# item id -> {item: item, classes: [], style: []}
elements = {}
for _, selector, style in rules:
new_selector = selector
class_ = ""
if ":" in selector:
new_selector, class_ = re.split(":", selector, 1)
class_ = ":%s" % class_
# Keep filter-type selectors untouched.
if class_ in FILTER_PSEUDOSELECTORS or class_.startswith(
":nth-child"):
class_ = ""
else:
selector = new_selector
assert selector
try:
sel = _create_cssselector(selector)
except SelectorSyntaxError:
# TODO: this should be optional
next
items = sel(page)
if len(items):
# same so process it first
processed_style = csstext_to_pairs(
style, validate=not self.disable_validation)
for item in items:
item_id = id(item)
if item_id not in elements:
elements[item_id] = {
"item": item,
"classes": [],
"style": []
}
elements[item_id]["style"].append(processed_style)
elements[item_id]["classes"].append(class_)
# Now apply inline style
# merge style only once for each element
# crucial when you have a lot of pseudo/classes
# and a long list of elements
for _, element in elements.items():
final_style = merge_styles(
element["item"].attrib.get("style", ""),
element["style"],
element["classes"],
remove_unset_properties=self.remove_unset_properties,
)
if final_style:
# final style could be empty string because of
# remove_unset_properties
element["item"].attrib["style"] = final_style
self._style_to_basic_html_attributes(element["item"],
final_style,
force=True)
if self.remove_classes:
# now we can delete all 'class' attributes
for item in page.xpath("//@class"):
parent = item.getparent()
del parent.attrib["class"]
# Capitalize Margin properties
# To fix weird outlook bug
# https://www.emailonacid.com/blog/article/email-development/outlook.com-does-support-margins
if self.capitalize_float_margin:
for item in page.xpath("//@style"):
mangled = capitalize_float_margin(item)
item.getparent().attrib["style"] = mangled
# Add align attributes to images if they have a CSS float value of
# right or left. Outlook (both on desktop and on the web) are bad at
# understanding floats, but they do understand the HTML align attrib.
if self.align_floating_images:
for item in page.xpath("//img[@style]"):
image_css = cssutils.parseStyle(item.attrib["style"])
if image_css.float == "right":
item.attrib["align"] = "right"
elif image_css.float == "left":
item.attrib["align"] = "left"
#
# URLs
#
if self.base_url and not self.disable_link_rewrites:
if not urlparse(self.base_url).scheme:
raise ValueError("Base URL must have a scheme")
for attr in ("href", "src"):
for item in page.xpath("//@%s" % attr):
parent = item.getparent()
url = parent.attrib[attr]
if (attr == "href" and self.preserve_internal_links
and url.startswith("#")):
continue
if (attr == "src" and self.preserve_inline_attachments
and url.startswith("cid:")):
continue
if attr == "href" and url.startswith("tel:"):
continue
parent.attrib[attr] = urljoin(self.base_url, url)
if hasattr(html, "getroottree"):
return root
else:
kwargs.setdefault("method", self.method)
kwargs.setdefault("pretty_print", pretty_print)
kwargs.setdefault("encoding", "utf-8") # As Ken Thompson intended
out = etree.tostring(root, **kwargs).decode(kwargs["encoding"])
if self.method == "xml":
out = _cdata_regex.sub(
lambda m: "/*<![CDATA[*/%s/*]]>*/" % m.group(1), out)
if self.strip_important:
out = _importants.sub("", out)
return out
def transform(self, pretty_print=True, **kwargs):
"""change the self.html and return it with CSS turned into style
attributes.
"""
if hasattr(self.html, "getroottree"):
# skip the next bit
root = self.html.getroottree()
page = root
tree = root
else:
if self.method == 'xml':
parser = etree.XMLParser(
ns_clean=False,
resolve_entities=False
)
else:
parser = etree.HTMLParser()
stripped = self.html.strip()
tree = etree.fromstring(stripped, parser).getroottree()
page = tree.getroot()
# lxml inserts a doctype if none exists, so only include it in
# the root if it was in the original html.
root = tree if stripped.startswith(tree.docinfo.doctype) else page
assert page is not None
if self.disable_leftover_css:
head = None
else:
head = get_or_create_head(tree)
#
# style selectors
#
rules = []
index = 0
for element in CSSSelector('style,link[rel~=stylesheet]')(page):
# If we have a media attribute whose value is anything other than
# 'all' or 'screen', ignore the ruleset.
media = element.attrib.get('media')
if media and media not in ('all', 'screen'):
continue
data_attribute = element.attrib.get(self.attribute_name)
if data_attribute:
if data_attribute == 'ignore':
del element.attrib[self.attribute_name]
continue
else:
warnings.warn(
'Unrecognized %s attribute (%r)' % (
self.attribute_name,
data_attribute,
)
)
is_style = element.tag == 'style'
if is_style:
css_body = element.text
else:
href = element.attrib.get('href')
css_body = self._load_external(href)
these_rules, these_leftover = self._parse_style_rules(
css_body, index
)
index += 1
rules.extend(these_rules)
parent_of_element = element.getparent()
if these_leftover or self.keep_style_tags:
if is_style:
style = element
else:
style = etree.Element('style')
style.attrib['type'] = 'text/css'
if self.keep_style_tags:
style.text = css_body
else:
style.text = self._css_rules_to_string(these_leftover)
if self.method == 'xml':
style.text = etree.CDATA(style.text)
if not is_style:
element.addprevious(style)
parent_of_element.remove(element)
elif not self.keep_style_tags or not is_style:
parent_of_element.remove(element)
# external style files
if self.external_styles:
for stylefile in self.external_styles:
css_body = self._load_external(stylefile)
self._process_css_text(css_body, index, rules, head)
index += 1
# css text
if self.css_text:
for css_body in self.css_text:
self._process_css_text(css_body, index, rules, head)
index += 1
# rules is a tuple of (specificity, selector, styles), where
# specificity is a tuple ordered such that more specific
# rules sort larger.
rules.sort(key=operator.itemgetter(0))
# collecting all elements that we need to apply rules on
# id is unique for the lifetime of the object
# and lxml should give us the same everytime during this run
# item id -> {item: item, classes: [], style: []}
elements = {}
for _, selector, style in rules:
new_selector = selector
class_ = ''
if ':' in selector:
new_selector, class_ = re.split(':', selector, 1)
class_ = ':%s' % class_
# Keep filter-type selectors untouched.
if class_ in FILTER_PSEUDOSELECTORS:
class_ = ''
else:
selector = new_selector
sel = CSSSelector(selector)
items = sel(page)
if len(items):
# same so process it first
processed_style = csstext_to_pairs(style)
for item in items:
item_id = id(item)
if item_id not in elements:
elements[item_id] = {
'item': item,
'classes': [],
'style': [],
}
elements[item_id]['style'].append(processed_style)
elements[item_id]['classes'].append(class_)
# Now apply inline style
# merge style only once for each element
# crucial when you have a lot of pseudo/classes
# and a long list of elements
for _, element in elements.items():
final_style = merge_styles(
element['item'].attrib.get('style', ''),
element['style'],
element['classes'],
remove_unset_properties=self.remove_unset_properties,
)
if final_style:
# final style could be empty string because of
# remove_unset_properties
element['item'].attrib['style'] = final_style
self._style_to_basic_html_attributes(
element['item'],
final_style,
force=True
)
if self.remove_classes:
# now we can delete all 'class' attributes
for item in page.xpath('//@class'):
parent = item.getparent()
del parent.attrib['class']
# Add align attributes to images if they have a CSS float value of
# right or left. Outlook (both on desktop and on the web) are bad at
# understanding floats, but they do understand the HTML align attrib.
if self.align_floating_images:
for item in page.xpath('//img[@style]'):
image_css = cssutils.parseStyle(item.attrib['style'])
if image_css.float == 'right':
item.attrib['align'] = 'right'
elif image_css.float == 'left':
item.attrib['align'] = 'left'
#
# URLs
#
if self.base_url:
if not urlparse(self.base_url).scheme:
raise ValueError('Base URL must have a scheme')
for attr in ('href', 'src'):
for item in page.xpath("//@%s" % attr):
parent = item.getparent()
url = parent.attrib[attr]
if (
attr == 'href' and self.preserve_internal_links and
url.startswith('#')
):
continue
if (
attr == 'src' and self.preserve_inline_attachments and
url.startswith('cid:')
):
continue
parent.attrib[attr] = urljoin(self.base_url, url)
if hasattr(self.html, "getroottree"):
return root
else:
kwargs.setdefault('method', self.method)
kwargs.setdefault('pretty_print', pretty_print)
kwargs.setdefault('encoding', 'utf-8') # As Ken Thompson intended
out = etree.tostring(root, **kwargs).decode(kwargs['encoding'])
if self.method == 'xml':
out = _cdata_regex.sub(
lambda m: '/*<![CDATA[*/%s/*]]>*/' % m.group(1),
out
)
if self.strip_important:
out = _importants.sub('', out)
return out
def test_csstext_to_pairs(self):
csstext = 'font-size:1px'
parsed_csstext = csstext_to_pairs(csstext)
self.assertEqual(('font-size', '1px'), parsed_csstext[0])
def test_important_rule(self):
# No exception after #133
csstext = 'font-size:1px !important'
parsed_csstext = csstext_to_pairs(csstext)
self.assertEqual(('font-size', '1px'), parsed_csstext[0])
def transform(self, pretty_print=True, **kwargs):
"""change the self.html and return it with CSS turned into style
attributes.
"""
if hasattr(self.html, "getroottree"):
# skip the next bit
root = self.html.getroottree()
page = root
tree = root
else:
if self.method == 'xml':
parser = etree.XMLParser(ns_clean=False,
resolve_entities=False)
else:
parser = etree.HTMLParser()
stripped = self.html.strip()
tree = etree.fromstring(stripped, parser).getroottree()
page = tree.getroot()
# lxml inserts a doctype if none exists, so only include it in
# the root if it was in the original html.
root = tree if stripped.startswith(tree.docinfo.doctype) else page
assert page is not None
if self.disable_leftover_css:
head = None
else:
head = get_or_create_head(tree)
#
# style selectors
#
rules = []
index = 0
for element in CSSSelector('style,link[rel~=stylesheet]')(page):
# If we have a media attribute whose value is anything other than
# 'all' or 'screen', ignore the ruleset.
media = element.attrib.get('media')
if media and media not in ('all', 'screen'):
continue
data_attribute = element.attrib.get(self.attribute_name)
if data_attribute:
if data_attribute == 'ignore':
del element.attrib[self.attribute_name]
continue
else:
warnings.warn('Unrecognized %s attribute (%r)' % (
self.attribute_name,
data_attribute,
))
is_style = element.tag == 'style'
if is_style:
css_body = element.text
else:
href = element.attrib.get('href')
css_body = self._load_external(href)
these_rules, these_leftover = self._parse_style_rules(
css_body, index)
index += 1
rules.extend(these_rules)
parent_of_element = element.getparent()
if these_leftover or self.keep_style_tags:
if is_style:
style = element
else:
style = etree.Element('style')
style.attrib['type'] = 'text/css'
if self.keep_style_tags:
style.text = css_body
else:
style.text = self._css_rules_to_string(these_leftover)
if self.method == 'xml':
style.text = etree.CDATA(style.text)
if not is_style:
element.addprevious(style)
parent_of_element.remove(element)
elif not self.keep_style_tags or not is_style:
parent_of_element.remove(element)
# external style files
if self.external_styles:
for stylefile in self.external_styles:
css_body = self._load_external(stylefile)
self._process_css_text(css_body, index, rules, head)
index += 1
# css text
if self.css_text:
for css_body in self.css_text:
self._process_css_text(css_body, index, rules, head)
index += 1
# rules is a tuple of (specificity, selector, styles), where
# specificity is a tuple ordered such that more specific
# rules sort larger.
rules.sort(key=operator.itemgetter(0))
# collecting all elements that we need to apply rules on
# id is unique for the lifetime of the object
# and lxml should give us the same everytime during this run
# item id -> {item: item, classes: [], style: []}
elements = {}
for _, selector, style in rules:
new_selector = selector
class_ = ''
if ':' in selector:
new_selector, class_ = re.split(':', selector, 1)
class_ = ':%s' % class_
# Keep filter-type selectors untouched.
if class_ in FILTER_PSEUDOSELECTORS:
class_ = ''
else:
selector = new_selector
sel = CSSSelector(selector)
items = sel(page)
if len(items):
# same so process it first
processed_style = csstext_to_pairs(style)
for item in items:
item_id = id(item)
if item_id not in elements:
elements[item_id] = {
'item': item,
'classes': [],
'style': [],
}
elements[item_id]['style'].append(processed_style)
elements[item_id]['classes'].append(class_)
# Now apply inline style
# merge style only once for each element
# crucial when you have a lot of pseudo/classes
# and a long list of elements
for _, element in elements.items():
final_style = merge_styles(
element['item'].attrib.get('style', ''),
element['style'],
element['classes'],
remove_unset_properties=self.remove_unset_properties,
)
if final_style:
# final style could be empty string because of
# remove_unset_properties
element['item'].attrib['style'] = final_style
self._style_to_basic_html_attributes(element['item'],
final_style,
force=True)
if self.remove_classes:
# now we can delete all 'class' attributes
for item in page.xpath('//@class'):
parent = item.getparent()
del parent.attrib['class']
# Capitalize Margin properties
# To fix weird outlook bug
# https://www.emailonacid.com/blog/article/email-development/outlook.com-does-support-margins
if self.capitalize_float_margin:
for item in page.xpath('//@style'):
mangled = capitalize_float_margin(item)
item.getparent().attrib['style'] = mangled
# Add align attributes to images if they have a CSS float value of
# right or left. Outlook (both on desktop and on the web) are bad at
# understanding floats, but they do understand the HTML align attrib.
if self.align_floating_images:
for item in page.xpath('//img[@style]'):
image_css = cssutils.parseStyle(item.attrib['style'])
if image_css.float == 'right':
item.attrib['align'] = 'right'
elif image_css.float == 'left':
item.attrib['align'] = 'left'
#
# URLs
#
if self.base_url:
if not urlparse(self.base_url).scheme:
raise ValueError('Base URL must have a scheme')
for attr in ('href', 'src'):
for item in page.xpath("//@%s" % attr):
parent = item.getparent()
url = parent.attrib[attr]
if (attr == 'href' and self.preserve_internal_links
and url.startswith('#')):
continue
if (attr == 'src' and self.preserve_inline_attachments
and url.startswith('cid:')):
continue
if attr == 'href' and url.startswith('tel:'):
continue
parent.attrib[attr] = urljoin(self.base_url, url)
if hasattr(self.html, "getroottree"):
return root
else:
kwargs.setdefault('method', self.method)
kwargs.setdefault('pretty_print', pretty_print)
kwargs.setdefault('encoding', 'utf-8') # As Ken Thompson intended
out = etree.tostring(root, **kwargs).decode(kwargs['encoding'])
if self.method == 'xml':
out = _cdata_regex.sub(
lambda m: '/*<![CDATA[*/%s/*]]>*/' % m.group(1), out)
if self.strip_important:
out = _importants.sub('', out)
return out
def test_csstext_to_pairs(self):
csstext = "font-size:1px"
parsed_csstext = csstext_to_pairs(csstext)
self.assertEqual(("font-size", "1px"), parsed_csstext[0])
def transform(self, html=None, pretty_print=True, **kwargs):
"""change the html and return it with CSS turned into style
attributes.
"""
if html is not None and self.html is not None:
raise TypeError("Can't pass html argument twice")
elif html is None and self.html is None:
raise TypeError("must pass html as first argument")
elif html is None:
html = self.html
if hasattr(html, "getroottree"):
# skip the next bit
root = html.getroottree()
page = root
tree = root
else:
if self.method == "xml":
parser = etree.XMLParser(ns_clean=False, resolve_entities=False)
else:
parser = etree.HTMLParser()
stripped = html.strip()
tree = etree.fromstring(stripped, parser).getroottree()
page = tree.getroot()
# lxml inserts a doctype if none exists, so only include it in
# the root if it was in the original html.
root = tree if stripped.startswith(tree.docinfo.doctype) else page
assert page is not None
if self.disable_leftover_css:
head = None
else:
head = get_or_create_head(tree)
#
# style selectors
#
rules = []
index = 0
cssselector = ["style"]
if self.allow_network:
cssselector.append("link[rel~=stylesheet]")
for element in _create_cssselector(",".join(cssselector))(page):
# If we have a media attribute whose value is anything other than
# 'all' or 'screen', ignore the ruleset.
media = element.attrib.get("media")
if media and media not in ("all", "screen"):
continue
data_attribute = element.attrib.get(self.attribute_name)
if data_attribute:
if data_attribute == "ignore":
del element.attrib[self.attribute_name]
continue
else:
warnings.warn(
"Unrecognized %s attribute (%r)"
% (self.attribute_name, data_attribute)
)
is_style = element.tag == "style"
if is_style:
css_body = element.text
else:
href = element.attrib.get("href")
css_body = self._load_external(href)
these_rules, these_leftover = self._parse_style_rules(css_body, index)
index += 1
rules.extend(these_rules)
parent_of_element = element.getparent()
if these_leftover or self.keep_style_tags:
if is_style:
style = element
else:
style = etree.Element("style")
style.attrib["type"] = "text/css"
if self.keep_style_tags:
style.text = css_body
else:
style.text = self._css_rules_to_string(these_leftover)
if self.method == "xml":
style.text = etree.CDATA(style.text)
if not is_style:
element.addprevious(style)
parent_of_element.remove(element)
elif not self.keep_style_tags or not is_style:
parent_of_element.remove(element)
# external style files
if self.external_styles and self.allow_network:
for stylefile in self.external_styles:
css_body = self._load_external(stylefile)
self._process_css_text(css_body, index, rules, head)
index += 1
# css text
if self.css_text:
for css_body in self.css_text:
self._process_css_text(css_body, index, rules, head)
index += 1
# rules is a tuple of (specificity, selector, styles), where
# specificity is a tuple ordered such that more specific
# rules sort larger.
rules.sort(key=operator.itemgetter(0))
# collecting all elements that we need to apply rules on
# id is unique for the lifetime of the object
# and lxml should give us the same everytime during this run
# item id -> {item: item, classes: [], style: []}
elements = {}
for _, selector, style in rules:
new_selector = selector
class_ = ""
if ":" in selector:
new_selector, class_ = re.split(":", selector, 1)
class_ = ":%s" % class_
# Keep filter-type selectors untouched.
if class_ in FILTER_PSEUDOSELECTORS or class_.startswith(":nth-child"):
class_ = ""
else:
selector = new_selector
assert selector
sel = _create_cssselector(selector)
items = sel(page)
if len(items):
# same so process it first
processed_style = csstext_to_pairs(style)
for item in items:
item_id = id(item)
if item_id not in elements:
elements[item_id] = {"item": item, "classes": [], "style": []}
elements[item_id]["style"].append(processed_style)
elements[item_id]["classes"].append(class_)
# Now apply inline style
# merge style only once for each element
# crucial when you have a lot of pseudo/classes
# and a long list of elements
for _, element in elements.items():
final_style = merge_styles(
element["item"].attrib.get("style", ""),
element["style"],
element["classes"],
remove_unset_properties=self.remove_unset_properties,
)
if final_style:
# final style could be empty string because of
# remove_unset_properties
element["item"].attrib["style"] = final_style
self._style_to_basic_html_attributes(
element["item"], final_style, force=True
)
if self.remove_classes:
# now we can delete all 'class' attributes
for item in page.xpath("//@class"):
parent = item.getparent()
del parent.attrib["class"]
# Capitalize Margin properties
# To fix weird outlook bug
# https://www.emailonacid.com/blog/article/email-development/outlook.com-does-support-margins
if self.capitalize_float_margin:
for item in page.xpath("//@style"):
mangled = capitalize_float_margin(item)
item.getparent().attrib["style"] = mangled
# Add align attributes to images if they have a CSS float value of
# right or left. Outlook (both on desktop and on the web) are bad at
# understanding floats, but they do understand the HTML align attrib.
if self.align_floating_images:
for item in page.xpath("//img[@style]"):
image_css = cssutils.parseStyle(item.attrib["style"])
if image_css.float == "right":
item.attrib["align"] = "right"
elif image_css.float == "left":
item.attrib["align"] = "left"
#
# URLs
#
if self.base_url and not self.disable_link_rewrites:
if not urlparse(self.base_url).scheme:
raise ValueError("Base URL must have a scheme")
for attr in ("href", "src"):
for item in page.xpath("//@%s" % attr):
parent = item.getparent()
url = parent.attrib[attr]
if (
attr == "href"
and self.preserve_internal_links
and url.startswith("#")
):
continue
if (
attr == "src"
and self.preserve_inline_attachments
and url.startswith("cid:")
):
continue
if attr == "href" and url.startswith("tel:"):
continue
parent.attrib[attr] = urljoin(self.base_url, url)
if hasattr(html, "getroottree"):
return root
else:
kwargs.setdefault("method", self.method)
kwargs.setdefault("pretty_print", pretty_print)
kwargs.setdefault("encoding", "utf-8") # As Ken Thompson intended
out = etree.tostring(root, **kwargs).decode(kwargs["encoding"])
if self.method == "xml":
out = _cdata_regex.sub(
lambda m: "/*<![CDATA[*/%s/*]]>*/" % m.group(1), out
)
if self.strip_important:
out = _importants.sub("", out)
return out
