Bläddra i källkod

Reorder table tokenizer methods for forward declaration

Make sure py tokenizer methods only call methods that have been declared
earlier. Not necessary but makes it much easier to maintain/write
the C tokenizer if methods are in the same order.
tags/v0.4
David Winegar 10 år sedan
förälder
incheckning
ddaa3ec92a
1 ändrade filer med 34 tillägg och 34 borttagningar
  1. +34
    -34
      mwparserfromhell/parser/tokenizer.py

+ 34
- 34
mwparserfromhell/parser/tokenizer.py Visa fil

@@ -1002,6 +1002,40 @@ class Tokenizer(object):
self._fail_route()
return self._pop()

def _parse_as_table_style(self, end_token, break_on_table_end=False):
"""Parse until ``end_token`` as style attributes for a table."""
data = _TagOpenData()
data.context = _TagOpenData.CX_ATTR_READY
while True:
this, next = self._read(), self._read(1)
can_exit = (not data.context & (data.CX_NAME) or
data.context & data.CX_NOTE_SPACE)
if this is self.END:
if self._context & contexts.TAG_ATTR:
if data.context & data.CX_QUOTED:
# Unclosed attribute quote: reset, don't die
data.context = data.CX_ATTR_VALUE
self._pop()
self._head = data.reset
continue
self._pop()
self._fail_route()
elif this == end_token and can_exit:
if data.context & (data.CX_ATTR_NAME | data.CX_ATTR_VALUE):
self._push_tag_buffer(data)
if this.isspace():
data.padding_buffer["first"] += this
return (self._pop(), data.padding_buffer["first"])
elif break_on_table_end and this == "|" and next == "}":
if data.context & (data.CX_ATTR_NAME | data.CX_ATTR_VALUE):
self._push_tag_buffer(data)
if this.isspace():
data.padding_buffer["first"] += this
return (self._pop(), data.padding_buffer["first"])
else:
self._handle_tag_data(data, this)
self._head += 1

def _handle_table_start(self):
"""Handle the start of a table."""
self._head += 2
@@ -1095,40 +1129,6 @@ class Tokenizer(object):
# offset displacement done by _parse()
self._head -= 1

def _parse_as_table_style(self, end_token, break_on_table_end=False):
"""Parse until ``end_token`` as style attributes for a table."""
data = _TagOpenData()
data.context = _TagOpenData.CX_ATTR_READY
while True:
this, next = self._read(), self._read(1)
can_exit = (not data.context & (data.CX_NAME) or
data.context & data.CX_NOTE_SPACE)
if this is self.END:
if self._context & contexts.TAG_ATTR:
if data.context & data.CX_QUOTED:
# Unclosed attribute quote: reset, don't die
data.context = data.CX_ATTR_VALUE
self._pop()
self._head = data.reset
continue
self._pop()
self._fail_route()
elif this == end_token and can_exit:
if data.context & (data.CX_ATTR_NAME | data.CX_ATTR_VALUE):
self._push_tag_buffer(data)
if this.isspace():
data.padding_buffer["first"] += this
return (self._pop(), data.padding_buffer["first"])
elif break_on_table_end and this == "|" and next == "}":
if data.context & (data.CX_ATTR_NAME | data.CX_ATTR_VALUE):
self._push_tag_buffer(data)
if this.isspace():
data.padding_buffer["first"] += this
return (self._pop(), data.padding_buffer["first"])
else:
self._handle_tag_data(data, this)
self._head += 1

def _handle_table_cell_end(self, reset_for_style=False):
"""Returns the context, stack, and whether to reset the cell for style
in a tuple."""


Laddar…
Avbryt
Spara