mirror of https://github.com/morpheus65535/bazarr
403 lines
15 KiB
Python
403 lines
15 KiB
Python
"""
|
|
Footnotes Extension for Python-Markdown
|
|
=======================================
|
|
|
|
Adds footnote handling to Python-Markdown.
|
|
|
|
See <https://Python-Markdown.github.io/extensions/footnotes>
|
|
for documentation.
|
|
|
|
Copyright The Python Markdown Project
|
|
|
|
License: [BSD](https://opensource.org/licenses/bsd-license.php)
|
|
|
|
"""
|
|
|
|
from . import Extension
|
|
from ..blockprocessors import BlockProcessor
|
|
from ..inlinepatterns import InlineProcessor
|
|
from ..treeprocessors import Treeprocessor
|
|
from ..postprocessors import Postprocessor
|
|
from .. import util
|
|
from collections import OrderedDict
|
|
import re
|
|
import copy
|
|
import xml.etree.ElementTree as etree
|
|
|
|
FN_BACKLINK_TEXT = util.STX + "zz1337820767766393qq" + util.ETX
|
|
NBSP_PLACEHOLDER = util.STX + "qq3936677670287331zz" + util.ETX
|
|
RE_REF_ID = re.compile(r'(fnref)(\d+)')
|
|
|
|
|
|
class FootnoteExtension(Extension):
|
|
""" Footnote Extension. """
|
|
|
|
def __init__(self, **kwargs):
|
|
""" Setup configs. """
|
|
|
|
self.config = {
|
|
'PLACE_MARKER':
|
|
["///Footnotes Go Here///",
|
|
"The text string that marks where the footnotes go"],
|
|
'UNIQUE_IDS':
|
|
[False,
|
|
"Avoid name collisions across "
|
|
"multiple calls to reset()."],
|
|
"BACKLINK_TEXT":
|
|
["↩",
|
|
"The text string that links from the footnote "
|
|
"to the reader's place."],
|
|
"BACKLINK_TITLE":
|
|
["Jump back to footnote %d in the text",
|
|
"The text string used for the title HTML attribute "
|
|
"of the backlink. %d will be replaced by the "
|
|
"footnote number."],
|
|
"SEPARATOR":
|
|
[":",
|
|
"Footnote separator."]
|
|
}
|
|
super().__init__(**kwargs)
|
|
|
|
# In multiple invocations, emit links that don't get tangled.
|
|
self.unique_prefix = 0
|
|
self.found_refs = {}
|
|
self.used_refs = set()
|
|
|
|
self.reset()
|
|
|
|
def extendMarkdown(self, md):
|
|
""" Add pieces to Markdown. """
|
|
md.registerExtension(self)
|
|
self.parser = md.parser
|
|
self.md = md
|
|
# Insert a blockprocessor before ReferencePreprocessor
|
|
md.parser.blockprocessors.register(FootnoteBlockProcessor(self), 'footnote', 17)
|
|
|
|
# Insert an inline pattern before ImageReferencePattern
|
|
FOOTNOTE_RE = r'\[\^([^\]]*)\]' # blah blah [^1] blah
|
|
md.inlinePatterns.register(FootnoteInlineProcessor(FOOTNOTE_RE, self), 'footnote', 175)
|
|
# Insert a tree-processor that would actually add the footnote div
|
|
# This must be before all other treeprocessors (i.e., inline and
|
|
# codehilite) so they can run on the the contents of the div.
|
|
md.treeprocessors.register(FootnoteTreeprocessor(self), 'footnote', 50)
|
|
|
|
# Insert a tree-processor that will run after inline is done.
|
|
# In this tree-processor we want to check our duplicate footnote tracker
|
|
# And add additional backrefs to the footnote pointing back to the
|
|
# duplicated references.
|
|
md.treeprocessors.register(FootnotePostTreeprocessor(self), 'footnote-duplicate', 15)
|
|
|
|
# Insert a postprocessor after amp_substitute processor
|
|
md.postprocessors.register(FootnotePostprocessor(self), 'footnote', 25)
|
|
|
|
def reset(self):
|
|
""" Clear footnotes on reset, and prepare for distinct document. """
|
|
self.footnotes = OrderedDict()
|
|
self.unique_prefix += 1
|
|
self.found_refs = {}
|
|
self.used_refs = set()
|
|
|
|
def unique_ref(self, reference, found=False):
|
|
""" Get a unique reference if there are duplicates. """
|
|
if not found:
|
|
return reference
|
|
|
|
original_ref = reference
|
|
while reference in self.used_refs:
|
|
ref, rest = reference.split(self.get_separator(), 1)
|
|
m = RE_REF_ID.match(ref)
|
|
if m:
|
|
reference = '%s%d%s%s' % (m.group(1), int(m.group(2))+1, self.get_separator(), rest)
|
|
else:
|
|
reference = '%s%d%s%s' % (ref, 2, self.get_separator(), rest)
|
|
|
|
self.used_refs.add(reference)
|
|
if original_ref in self.found_refs:
|
|
self.found_refs[original_ref] += 1
|
|
else:
|
|
self.found_refs[original_ref] = 1
|
|
return reference
|
|
|
|
def findFootnotesPlaceholder(self, root):
|
|
""" Return ElementTree Element that contains Footnote placeholder. """
|
|
def finder(element):
|
|
for child in element:
|
|
if child.text:
|
|
if child.text.find(self.getConfig("PLACE_MARKER")) > -1:
|
|
return child, element, True
|
|
if child.tail:
|
|
if child.tail.find(self.getConfig("PLACE_MARKER")) > -1:
|
|
return child, element, False
|
|
child_res = finder(child)
|
|
if child_res is not None:
|
|
return child_res
|
|
return None
|
|
|
|
res = finder(root)
|
|
return res
|
|
|
|
def setFootnote(self, id, text):
|
|
""" Store a footnote for later retrieval. """
|
|
self.footnotes[id] = text
|
|
|
|
def get_separator(self):
|
|
""" Get the footnote separator. """
|
|
return self.getConfig("SEPARATOR")
|
|
|
|
def makeFootnoteId(self, id):
|
|
""" Return footnote link id. """
|
|
if self.getConfig("UNIQUE_IDS"):
|
|
return 'fn%s%d-%s' % (self.get_separator(), self.unique_prefix, id)
|
|
else:
|
|
return 'fn{}{}'.format(self.get_separator(), id)
|
|
|
|
def makeFootnoteRefId(self, id, found=False):
|
|
""" Return footnote back-link id. """
|
|
if self.getConfig("UNIQUE_IDS"):
|
|
return self.unique_ref('fnref%s%d-%s' % (self.get_separator(), self.unique_prefix, id), found)
|
|
else:
|
|
return self.unique_ref('fnref{}{}'.format(self.get_separator(), id), found)
|
|
|
|
def makeFootnotesDiv(self, root):
|
|
""" Return div of footnotes as et Element. """
|
|
|
|
if not list(self.footnotes.keys()):
|
|
return None
|
|
|
|
div = etree.Element("div")
|
|
div.set('class', 'footnote')
|
|
etree.SubElement(div, "hr")
|
|
ol = etree.SubElement(div, "ol")
|
|
surrogate_parent = etree.Element("div")
|
|
|
|
for index, id in enumerate(self.footnotes.keys(), start=1):
|
|
li = etree.SubElement(ol, "li")
|
|
li.set("id", self.makeFootnoteId(id))
|
|
# Parse footnote with surrogate parent as li cannot be used.
|
|
# List block handlers have special logic to deal with li.
|
|
# When we are done parsing, we will copy everything over to li.
|
|
self.parser.parseChunk(surrogate_parent, self.footnotes[id])
|
|
for el in list(surrogate_parent):
|
|
li.append(el)
|
|
surrogate_parent.remove(el)
|
|
backlink = etree.Element("a")
|
|
backlink.set("href", "#" + self.makeFootnoteRefId(id))
|
|
backlink.set("class", "footnote-backref")
|
|
backlink.set(
|
|
"title",
|
|
self.getConfig("BACKLINK_TITLE") % (index)
|
|
)
|
|
backlink.text = FN_BACKLINK_TEXT
|
|
|
|
if len(li):
|
|
node = li[-1]
|
|
if node.tag == "p":
|
|
node.text = node.text + NBSP_PLACEHOLDER
|
|
node.append(backlink)
|
|
else:
|
|
p = etree.SubElement(li, "p")
|
|
p.append(backlink)
|
|
return div
|
|
|
|
|
|
class FootnoteBlockProcessor(BlockProcessor):
|
|
""" Find all footnote references and store for later use. """
|
|
|
|
RE = re.compile(r'^[ ]{0,3}\[\^([^\]]*)\]:[ ]*(.*)$', re.MULTILINE)
|
|
|
|
def __init__(self, footnotes):
|
|
super().__init__(footnotes.parser)
|
|
self.footnotes = footnotes
|
|
|
|
def test(self, parent, block):
|
|
return True
|
|
|
|
def run(self, parent, blocks):
|
|
""" Find, set, and remove footnote definitions. """
|
|
block = blocks.pop(0)
|
|
m = self.RE.search(block)
|
|
if m:
|
|
id = m.group(1)
|
|
fn_blocks = [m.group(2)]
|
|
|
|
# Handle rest of block
|
|
therest = block[m.end():].lstrip('\n')
|
|
m2 = self.RE.search(therest)
|
|
if m2:
|
|
# Another footnote exists in the rest of this block.
|
|
# Any content before match is continuation of this footnote, which may be lazily indented.
|
|
before = therest[:m2.start()].rstrip('\n')
|
|
fn_blocks[0] = '\n'.join([fn_blocks[0], self.detab(before)]).lstrip('\n')
|
|
# Add back to blocks everything from begining of match forward for next iteration.
|
|
blocks.insert(0, therest[m2.start():])
|
|
else:
|
|
# All remaining lines of block are continuation of this footnote, which may be lazily indented.
|
|
fn_blocks[0] = '\n'.join([fn_blocks[0], self.detab(therest)]).strip('\n')
|
|
|
|
# Check for child elements in remaining blocks.
|
|
fn_blocks.extend(self.detectTabbed(blocks))
|
|
|
|
footnote = "\n\n".join(fn_blocks)
|
|
self.footnotes.setFootnote(id, footnote.rstrip())
|
|
|
|
if block[:m.start()].strip():
|
|
# Add any content before match back to blocks as separate block
|
|
blocks.insert(0, block[:m.start()].rstrip('\n'))
|
|
return True
|
|
# No match. Restore block.
|
|
blocks.insert(0, block)
|
|
return False
|
|
|
|
def detectTabbed(self, blocks):
|
|
""" Find indented text and remove indent before further proccesing.
|
|
|
|
Returns: a list of blocks with indentation removed.
|
|
"""
|
|
fn_blocks = []
|
|
while blocks:
|
|
if blocks[0].startswith(' '*4):
|
|
block = blocks.pop(0)
|
|
# Check for new footnotes within this block and split at new footnote.
|
|
m = self.RE.search(block)
|
|
if m:
|
|
# Another footnote exists in this block.
|
|
# Any content before match is continuation of this footnote, which may be lazily indented.
|
|
before = block[:m.start()].rstrip('\n')
|
|
fn_blocks.append(self.detab(before))
|
|
# Add back to blocks everything from begining of match forward for next iteration.
|
|
blocks.insert(0, block[m.start():])
|
|
# End of this footnote.
|
|
break
|
|
else:
|
|
# Entire block is part of this footnote.
|
|
fn_blocks.append(self.detab(block))
|
|
else:
|
|
# End of this footnote.
|
|
break
|
|
return fn_blocks
|
|
|
|
def detab(self, block):
|
|
""" Remove one level of indent from a block.
|
|
|
|
Preserve lazily indented blocks by only removing indent from indented lines.
|
|
"""
|
|
lines = block.split('\n')
|
|
for i, line in enumerate(lines):
|
|
if line.startswith(' '*4):
|
|
lines[i] = line[4:]
|
|
return '\n'.join(lines)
|
|
|
|
|
|
class FootnoteInlineProcessor(InlineProcessor):
|
|
""" InlinePattern for footnote markers in a document's body text. """
|
|
|
|
def __init__(self, pattern, footnotes):
|
|
super().__init__(pattern)
|
|
self.footnotes = footnotes
|
|
|
|
def handleMatch(self, m, data):
|
|
id = m.group(1)
|
|
if id in self.footnotes.footnotes.keys():
|
|
sup = etree.Element("sup")
|
|
a = etree.SubElement(sup, "a")
|
|
sup.set('id', self.footnotes.makeFootnoteRefId(id, found=True))
|
|
a.set('href', '#' + self.footnotes.makeFootnoteId(id))
|
|
a.set('class', 'footnote-ref')
|
|
a.text = str(list(self.footnotes.footnotes.keys()).index(id) + 1)
|
|
return sup, m.start(0), m.end(0)
|
|
else:
|
|
return None, None, None
|
|
|
|
|
|
class FootnotePostTreeprocessor(Treeprocessor):
|
|
""" Amend footnote div with duplicates. """
|
|
|
|
def __init__(self, footnotes):
|
|
self.footnotes = footnotes
|
|
|
|
def add_duplicates(self, li, duplicates):
|
|
""" Adjust current li and add the duplicates: fnref2, fnref3, etc. """
|
|
for link in li.iter('a'):
|
|
# Find the link that needs to be duplicated.
|
|
if link.attrib.get('class', '') == 'footnote-backref':
|
|
ref, rest = link.attrib['href'].split(self.footnotes.get_separator(), 1)
|
|
# Duplicate link the number of times we need to
|
|
# and point the to the appropriate references.
|
|
links = []
|
|
for index in range(2, duplicates + 1):
|
|
sib_link = copy.deepcopy(link)
|
|
sib_link.attrib['href'] = '%s%d%s%s' % (ref, index, self.footnotes.get_separator(), rest)
|
|
links.append(sib_link)
|
|
self.offset += 1
|
|
# Add all the new duplicate links.
|
|
el = list(li)[-1]
|
|
for link in links:
|
|
el.append(link)
|
|
break
|
|
|
|
def get_num_duplicates(self, li):
|
|
""" Get the number of duplicate refs of the footnote. """
|
|
fn, rest = li.attrib.get('id', '').split(self.footnotes.get_separator(), 1)
|
|
link_id = '{}ref{}{}'.format(fn, self.footnotes.get_separator(), rest)
|
|
return self.footnotes.found_refs.get(link_id, 0)
|
|
|
|
def handle_duplicates(self, parent):
|
|
""" Find duplicate footnotes and format and add the duplicates. """
|
|
for li in list(parent):
|
|
# Check number of duplicates footnotes and insert
|
|
# additional links if needed.
|
|
count = self.get_num_duplicates(li)
|
|
if count > 1:
|
|
self.add_duplicates(li, count)
|
|
|
|
def run(self, root):
|
|
""" Crawl the footnote div and add missing duplicate footnotes. """
|
|
self.offset = 0
|
|
for div in root.iter('div'):
|
|
if div.attrib.get('class', '') == 'footnote':
|
|
# Footnotes shoul be under the first orderd list under
|
|
# the footnote div. So once we find it, quit.
|
|
for ol in div.iter('ol'):
|
|
self.handle_duplicates(ol)
|
|
break
|
|
|
|
|
|
class FootnoteTreeprocessor(Treeprocessor):
|
|
""" Build and append footnote div to end of document. """
|
|
|
|
def __init__(self, footnotes):
|
|
self.footnotes = footnotes
|
|
|
|
def run(self, root):
|
|
footnotesDiv = self.footnotes.makeFootnotesDiv(root)
|
|
if footnotesDiv is not None:
|
|
result = self.footnotes.findFootnotesPlaceholder(root)
|
|
if result:
|
|
child, parent, isText = result
|
|
ind = list(parent).index(child)
|
|
if isText:
|
|
parent.remove(child)
|
|
parent.insert(ind, footnotesDiv)
|
|
else:
|
|
parent.insert(ind + 1, footnotesDiv)
|
|
child.tail = None
|
|
else:
|
|
root.append(footnotesDiv)
|
|
|
|
|
|
class FootnotePostprocessor(Postprocessor):
|
|
""" Replace placeholders with html entities. """
|
|
def __init__(self, footnotes):
|
|
self.footnotes = footnotes
|
|
|
|
def run(self, text):
|
|
text = text.replace(
|
|
FN_BACKLINK_TEXT, self.footnotes.getConfig("BACKLINK_TEXT")
|
|
)
|
|
return text.replace(NBSP_PLACEHOLDER, " ")
|
|
|
|
|
|
def makeExtension(**kwargs): # pragma: no cover
|
|
""" Return an instance of the FootnoteExtension """
|
|
return FootnoteExtension(**kwargs)
|