mirror of https://github.com/python/cpython
Do a little bit more work on a table-driven basis. A little cleanup.
Seems to work at the moment.
This commit is contained in:
parent
20ee206500
commit
74a11e59a2
139
Doc/partparse.py
139
Doc/partparse.py
|
@ -836,9 +836,10 @@ class Wobj:
|
|||
self.data = self.data + data
|
||||
|
||||
# ignore these commands
|
||||
ignoredcommands = ('bcode', 'ecode', 'hline', 'small', '/')
|
||||
ignoredcommands = ('hline', 'small', '/', 'tableofcontents', 'Large')
|
||||
# map commands like these to themselves as plaintext
|
||||
wordsselves = ('UNIX', 'ABC', 'C', 'ASCII', 'EOF', 'LaTeX', 'POSIX')
|
||||
wordsselves = ('UNIX', 'ABC', 'C', 'ASCII', 'EOF', 'LaTeX', 'POSIX', 'TeX',
|
||||
'SliTeX')
|
||||
# \{ --> {, \} --> }, etc
|
||||
themselves = ('{', '}', ',', '.', '@', ' ', '\n') + wordsselves
|
||||
# these ones also themselves (see argargs macro in myformat.sty)
|
||||
|
@ -851,10 +852,6 @@ markcmds = {'code': ('', ''), 'var': 1, 'emph': ('_', '_'),
|
|||
# recognise patter {\FONTCHANGE-CMD TEXT} to \MAPPED-FC-CMD{TEXT}
|
||||
fontchanges = {'rm': 'r', 'it': 'i', 'em': 'emph', 'bf': 'b', 'tt': 't'}
|
||||
|
||||
# transparent for these commands
|
||||
for_texi = ('emph', 'var', 'strong', 'code', 'kbd', 'key', 'dfn', 'samp',
|
||||
'file', 'r', 'i', 't')
|
||||
|
||||
|
||||
# try to remove macros and return flat text
|
||||
def flattext(buf, pp):
|
||||
|
@ -874,11 +871,10 @@ def flattext(buf, pp):
|
|||
pass
|
||||
elif ch.chtype == chunk_type[CSNAME]:
|
||||
s_buf_data = s(buf, ch.data)
|
||||
if s_buf_data in themselves or hist.inargs and s_buf_data in inargsselves:
|
||||
if convertible_csname(s_buf_data):
|
||||
ch.chtype, ch.data, nix = conversion(s_buf_data)
|
||||
if hist.inargs and s_buf_data in inargsselves:
|
||||
ch.chtype = chunk_type[PLAIN]
|
||||
elif s_buf_data == 'e':
|
||||
ch.chtype = chunk_type[PLAIN]
|
||||
ch.data = '\\'
|
||||
elif len(s_buf_data) == 1 \
|
||||
and s_buf_data in onlylatexspecial:
|
||||
ch.chtype = chunk_type[PLAIN]
|
||||
|
@ -993,6 +989,7 @@ out = Struct()
|
|||
def startchange():
|
||||
global hist, out
|
||||
|
||||
hist.chaptertype = "chapter"
|
||||
hist.inenv = []
|
||||
hist.nodenames = []
|
||||
hist.cindex = []
|
||||
|
@ -1013,17 +1010,42 @@ itemizesymbols = ['bullet', 'minus', 'dots']
|
|||
# same for enumerate
|
||||
enumeratesymbols = ['1', 'A', 'a']
|
||||
|
||||
# Map of things that convert one-to-one. Each entry is a 3-tuple:
|
||||
#
|
||||
# new_chtype, new_data, nix_trailing_empty_group
|
||||
#
|
||||
d = {}
|
||||
# add stuff that converts from one name to another:
|
||||
for name in ('url', 'module', 'function', 'cfunction',
|
||||
'keyword', 'method', 'exception', 'constant',
|
||||
'email', 'class', 'member', 'cdata', 'ctype',
|
||||
'member'):
|
||||
d[name] = 'code'
|
||||
d['program'] = 'strong'
|
||||
d['sectcode'] = 'code'
|
||||
'member', 'sectcode', 'verb'):
|
||||
d[name] = chunk_type[CSNAME], 'code', 0
|
||||
for name in ('emph', 'var', 'strong', 'code', 'kbd', 'key',
|
||||
'dfn', 'samp', 'file', 'r', 'i', 't'):
|
||||
d[name] = chunk_type[CSNAME], name, 0
|
||||
d['program'] = chunk_type[CSNAME], 'strong', 0
|
||||
d['\\'] = chunk_type[CSNAME], '*', 0
|
||||
# add stuff that converts to text:
|
||||
for name in themselves:
|
||||
d[name] = chunk_type[PLAIN], name, 0
|
||||
for name in wordsselves:
|
||||
d[name] = chunk_type[PLAIN], name, 1
|
||||
for name in ',[]()':
|
||||
d[name] = chunk_type[PLAIN], name, 0
|
||||
# a lot of these are LaTeX2e additions
|
||||
for name, value in [('quotedblbase', ',,'), ('quotesinglbase', ','),
|
||||
('textquotedbl', '"'), ('LaTeXe', 'LaTeX2e'),
|
||||
('e', '\\'), ('textquotedblleft', "``"),
|
||||
('textquotedblright', "''"), ('textquoteleft', "`"),
|
||||
('textquoteright', "'"), ('textbackslash', '\\'),
|
||||
('textbar', '|'), ('textless', '<'),
|
||||
('textgreater', '>'), ('textasciicircum', '^'),
|
||||
('Cpp', 'C++'), ('copyright', '')]:
|
||||
d[name] = chunk_type[PLAIN], value, 1
|
||||
convertible_csname = d.has_key
|
||||
conversion = d.get
|
||||
del d, name
|
||||
del d, name, value
|
||||
|
||||
##
|
||||
## \begin{ {func,data,exc}desc }{name}...
|
||||
|
@ -1261,6 +1283,7 @@ def changeit(buf, pp):
|
|||
while 1:
|
||||
# sanity check: length should always equal len(pp)
|
||||
if len(pp) != length:
|
||||
print i, pp[i]
|
||||
raise 'FATAL', 'inconsistent length. thought ' + `length` + ', but should really be ' + `len(pp)`
|
||||
if i >= length:
|
||||
break
|
||||
|
@ -1404,7 +1427,7 @@ def changeit(buf, pp):
|
|||
raise error, 'Sorry, expected plain text argument'
|
||||
hist.itemargmacro = s(buf, pp[i].data)
|
||||
if convertible_csname(hist.itemargmacro):
|
||||
hist.itemargmacro = conversion(hist.itemargmacro)
|
||||
hist.itemargmacro = conversion(hist.itemargmacro)[1]
|
||||
del pp[i:newi]
|
||||
length = length - (newi-i)
|
||||
|
||||
|
@ -1466,7 +1489,7 @@ def changeit(buf, pp):
|
|||
length = length + len(chunks) - 1
|
||||
i = i + len(chunks) - 1
|
||||
|
||||
elif envname in ('sloppypar', 'flushleft'):
|
||||
elif envname in ('sloppypar', 'flushleft', 'document'):
|
||||
pass
|
||||
|
||||
else:
|
||||
|
@ -1523,7 +1546,7 @@ def changeit(buf, pp):
|
|||
chunk(PLAIN, ch.where, "deffn")])]
|
||||
i, length = i+2, length+2
|
||||
|
||||
elif envname in ('seealso', 'sloppypar', 'flushleft'):
|
||||
elif envname in ('seealso', 'sloppypar', 'flushleft', 'document'):
|
||||
pass
|
||||
|
||||
else:
|
||||
|
@ -1541,9 +1564,11 @@ def changeit(buf, pp):
|
|||
pp[i:i+1]=cp + [
|
||||
chunk(PLAIN, ch.where, ']')]
|
||||
length = length+len(cp)
|
||||
|
||||
elif s_buf_data in ignoredcommands:
|
||||
del pp[i-1]
|
||||
i, length = i-1, length-1
|
||||
|
||||
elif s_buf_data == '@' and \
|
||||
i != length and \
|
||||
pp[i].chtype == chunk_type[PLAIN] and \
|
||||
|
@ -1551,10 +1576,22 @@ def changeit(buf, pp):
|
|||
# \@. --> \. --> @.
|
||||
ch.data = '.'
|
||||
del pp[i]
|
||||
length = length-1
|
||||
length = length - 1
|
||||
|
||||
elif convertible_csname(s_buf_data):
|
||||
ch.chtype, ch.data, nix = conversion(s_buf_data)
|
||||
try:
|
||||
if nix and pp[i].chtype == chunk_type[GROUP] \
|
||||
and len(pp[i].data) == 0:
|
||||
del pp[i]
|
||||
length = length - 1
|
||||
except IndexError:
|
||||
pass
|
||||
|
||||
elif s_buf_data == '\\':
|
||||
# \\ --> \* --> @*
|
||||
ch.data = '*'
|
||||
|
||||
elif len(s_buf_data) == 1 and \
|
||||
s_buf_data in onlylatexspecial:
|
||||
ch.chtype = chunk_type[PLAIN]
|
||||
|
@ -1567,6 +1604,11 @@ def changeit(buf, pp):
|
|||
del pp[i]
|
||||
length = length-1
|
||||
|
||||
elif s_buf_data == "appendix":
|
||||
hist.chaptertype = "appendix"
|
||||
del pp[i-1]
|
||||
i, length = i-1, length-1
|
||||
|
||||
elif hist.inargs and s_buf_data in inargsselves:
|
||||
# This is the special processing of the
|
||||
# arguments of the \begin{funcdesc}... or
|
||||
|
@ -1597,6 +1639,13 @@ def changeit(buf, pp):
|
|||
i = i - 1
|
||||
length = length - 3
|
||||
|
||||
elif s_buf_data == 'renewcommand':
|
||||
print "ignoring redefinition of \\" \
|
||||
+ s(buf, pp[i].data[0].data)
|
||||
del pp[i-1:i+2]
|
||||
i = i - 1
|
||||
length = length - 3
|
||||
|
||||
elif s_buf_data == 'mbox':
|
||||
stuff = pp[i].data
|
||||
pp[i-1:i+1] = stuff
|
||||
|
@ -1672,6 +1721,7 @@ def changeit(buf, pp):
|
|||
or pp[i].data != []:
|
||||
pp.insert(i, chunk(GROUP, ch.where, []))
|
||||
i, length = i+1, length+1
|
||||
|
||||
elif s_buf_data in themselves:
|
||||
# \UNIX --> &UNIX;
|
||||
ch.chtype = chunk_type[PLAIN]
|
||||
|
@ -1680,8 +1730,14 @@ def changeit(buf, pp):
|
|||
and pp[i].data == []:
|
||||
del pp[i]
|
||||
length = length-1
|
||||
elif s_buf_data in for_texi:
|
||||
pass
|
||||
|
||||
## elif s_buf_data == 'copyright':
|
||||
## if (pp[i].chtype == chunk_type[GROUP]
|
||||
## and not pp[i].data):
|
||||
## del pp[i]
|
||||
## length = length - 1
|
||||
## del pp[i-1]
|
||||
## i, length = i-1, length-1
|
||||
|
||||
elif s_buf_data == 'manpage':
|
||||
ch.data = 'emph'
|
||||
|
@ -1689,10 +1745,6 @@ def changeit(buf, pp):
|
|||
pp[i+1].data = "(%s)" % sect
|
||||
pp[i+1].chtype = chunk_type[PLAIN]
|
||||
|
||||
elif s_buf_data == 'e':
|
||||
# "\e" --> "\"
|
||||
ch.data = '\\'
|
||||
ch.chtype = chunk_type[PLAIN]
|
||||
elif s_buf_data in ('lineiii', 'lineii'):
|
||||
# This is the most tricky one
|
||||
# \lineiii{a1}{a2}[{a3}] -->
|
||||
|
@ -1726,11 +1778,14 @@ def changeit(buf, pp):
|
|||
if length != len(pp):
|
||||
raise 'IN LINEIII IS THE ERR', `i`
|
||||
|
||||
elif s_buf_data in ('chapter', 'section', 'subsection', 'subsubsection'):
|
||||
elif s_buf_data in ('chapter', 'section',
|
||||
'subsection', 'subsubsection'):
|
||||
#\xxxsection{A} ---->
|
||||
# @node A, , ,
|
||||
# @xxxsection A
|
||||
## also: remove commas and quotes
|
||||
if s_buf_data == "chapter":
|
||||
ch.data = hist.chaptertype
|
||||
ch.chtype = chunk_type[CSLINE]
|
||||
length, newi = getnextarg(length, buf, pp, i)
|
||||
afternodenamecmd = next_command_p(length, buf,
|
||||
|
@ -1915,7 +1970,7 @@ def changeit(buf, pp):
|
|||
pp.insert(i, chunk(GROUP, ch.where, ingroupch))
|
||||
length, i = length+1, i+1
|
||||
|
||||
elif s_buf_data in ('stindex', 'kwindex'):
|
||||
elif s_buf_data == 'stindex':
|
||||
# XXX must actually go to newindex st
|
||||
what = (s_buf_data[:2] == "st") and "statement" or "keyword"
|
||||
wh = ch.where
|
||||
|
@ -2091,12 +2146,13 @@ def changeit(buf, pp):
|
|||
ch.chtype = PLAIN
|
||||
ch.data = " "
|
||||
|
||||
elif s_buf_data in ('usepackage', 'input'):
|
||||
del pp[i-1:i+1]
|
||||
i, length = i-1, length-2
|
||||
|
||||
elif s_buf_data in ('noindent', 'indexsubitem', 'footnote'):
|
||||
pass
|
||||
|
||||
elif convertible_csname(s_buf_data):
|
||||
ch.data = conversion(s_buf_data)
|
||||
|
||||
elif s_buf_data == 'label':
|
||||
name = s(buf, pp[i].data[0].data)
|
||||
del pp[i-1:i+1]
|
||||
|
@ -2110,11 +2166,6 @@ def changeit(buf, pp):
|
|||
del pp[i]
|
||||
length = length - 1
|
||||
|
||||
elif s_buf_data == 'Large':
|
||||
del pp[i-1]
|
||||
i = i - 1
|
||||
length = length - 1
|
||||
|
||||
elif s_buf_data == 'ref':
|
||||
name = s(buf, pp[i].data[0].data)
|
||||
if label_nodes.has_key(name):
|
||||
|
@ -2154,12 +2205,12 @@ def dumpit(buf, wm, pp):
|
|||
|
||||
if ch.chtype == chunk_type[CSNAME]:
|
||||
s_buf_data = s(buf, ch.data)
|
||||
if s_buf_data == 'e':
|
||||
wm('\\')
|
||||
continue
|
||||
if s_buf_data == '$':
|
||||
wm('$')
|
||||
continue
|
||||
## if s_buf_data == 'e':
|
||||
## wm('\\')
|
||||
## continue
|
||||
## if s_buf_data == '$':
|
||||
## wm('$')
|
||||
## continue
|
||||
wm('@' + s_buf_data)
|
||||
if s_buf_data == 'node' and \
|
||||
pp[i].chtype == chunk_type[PLAIN] and \
|
||||
|
@ -2232,13 +2283,11 @@ def dumpit(buf, wm, pp):
|
|||
wm('\n')
|
||||
|
||||
elif ch.chtype == chunk_type[COMMENT]:
|
||||
## print 'COMMENT: previous chunk =', pp[i-2]
|
||||
## if pp[i-2].chtype == chunk_type[PLAIN]:
|
||||
## print 'PLAINTEXT =', `s(buf, pp[i-2].data)`
|
||||
if s(buf, ch.data) and \
|
||||
regex.match('^[ \t]*$', s(buf, ch.data)) < 0:
|
||||
if i >= 2 \
|
||||
and pp[i-2].chtype not in (chunk_type[ENDLINE], chunk_type[DENDLINE]) \
|
||||
and pp[i-2].chtype not in (chunk_type[ENDLINE],
|
||||
chunk_type[DENDLINE]) \
|
||||
and not (pp[i-2].chtype == chunk_type[PLAIN]
|
||||
and regex.match('\\(.\\|\n\\)*[ \t]*\n$', s(buf, pp[i-2].data)) >= 0):
|
||||
wm('\n')
|
||||
|
|
|
@ -836,9 +836,10 @@ class Wobj:
|
|||
self.data = self.data + data
|
||||
|
||||
# ignore these commands
|
||||
ignoredcommands = ('bcode', 'ecode', 'hline', 'small', '/')
|
||||
ignoredcommands = ('hline', 'small', '/', 'tableofcontents', 'Large')
|
||||
# map commands like these to themselves as plaintext
|
||||
wordsselves = ('UNIX', 'ABC', 'C', 'ASCII', 'EOF', 'LaTeX', 'POSIX')
|
||||
wordsselves = ('UNIX', 'ABC', 'C', 'ASCII', 'EOF', 'LaTeX', 'POSIX', 'TeX',
|
||||
'SliTeX')
|
||||
# \{ --> {, \} --> }, etc
|
||||
themselves = ('{', '}', ',', '.', '@', ' ', '\n') + wordsselves
|
||||
# these ones also themselves (see argargs macro in myformat.sty)
|
||||
|
@ -851,10 +852,6 @@ markcmds = {'code': ('', ''), 'var': 1, 'emph': ('_', '_'),
|
|||
# recognise patter {\FONTCHANGE-CMD TEXT} to \MAPPED-FC-CMD{TEXT}
|
||||
fontchanges = {'rm': 'r', 'it': 'i', 'em': 'emph', 'bf': 'b', 'tt': 't'}
|
||||
|
||||
# transparent for these commands
|
||||
for_texi = ('emph', 'var', 'strong', 'code', 'kbd', 'key', 'dfn', 'samp',
|
||||
'file', 'r', 'i', 't')
|
||||
|
||||
|
||||
# try to remove macros and return flat text
|
||||
def flattext(buf, pp):
|
||||
|
@ -874,11 +871,10 @@ def flattext(buf, pp):
|
|||
pass
|
||||
elif ch.chtype == chunk_type[CSNAME]:
|
||||
s_buf_data = s(buf, ch.data)
|
||||
if s_buf_data in themselves or hist.inargs and s_buf_data in inargsselves:
|
||||
if convertible_csname(s_buf_data):
|
||||
ch.chtype, ch.data, nix = conversion(s_buf_data)
|
||||
if hist.inargs and s_buf_data in inargsselves:
|
||||
ch.chtype = chunk_type[PLAIN]
|
||||
elif s_buf_data == 'e':
|
||||
ch.chtype = chunk_type[PLAIN]
|
||||
ch.data = '\\'
|
||||
elif len(s_buf_data) == 1 \
|
||||
and s_buf_data in onlylatexspecial:
|
||||
ch.chtype = chunk_type[PLAIN]
|
||||
|
@ -993,6 +989,7 @@ out = Struct()
|
|||
def startchange():
|
||||
global hist, out
|
||||
|
||||
hist.chaptertype = "chapter"
|
||||
hist.inenv = []
|
||||
hist.nodenames = []
|
||||
hist.cindex = []
|
||||
|
@ -1013,17 +1010,42 @@ itemizesymbols = ['bullet', 'minus', 'dots']
|
|||
# same for enumerate
|
||||
enumeratesymbols = ['1', 'A', 'a']
|
||||
|
||||
# Map of things that convert one-to-one. Each entry is a 3-tuple:
|
||||
#
|
||||
# new_chtype, new_data, nix_trailing_empty_group
|
||||
#
|
||||
d = {}
|
||||
# add stuff that converts from one name to another:
|
||||
for name in ('url', 'module', 'function', 'cfunction',
|
||||
'keyword', 'method', 'exception', 'constant',
|
||||
'email', 'class', 'member', 'cdata', 'ctype',
|
||||
'member'):
|
||||
d[name] = 'code'
|
||||
d['program'] = 'strong'
|
||||
d['sectcode'] = 'code'
|
||||
'member', 'sectcode', 'verb'):
|
||||
d[name] = chunk_type[CSNAME], 'code', 0
|
||||
for name in ('emph', 'var', 'strong', 'code', 'kbd', 'key',
|
||||
'dfn', 'samp', 'file', 'r', 'i', 't'):
|
||||
d[name] = chunk_type[CSNAME], name, 0
|
||||
d['program'] = chunk_type[CSNAME], 'strong', 0
|
||||
d['\\'] = chunk_type[CSNAME], '*', 0
|
||||
# add stuff that converts to text:
|
||||
for name in themselves:
|
||||
d[name] = chunk_type[PLAIN], name, 0
|
||||
for name in wordsselves:
|
||||
d[name] = chunk_type[PLAIN], name, 1
|
||||
for name in ',[]()':
|
||||
d[name] = chunk_type[PLAIN], name, 0
|
||||
# a lot of these are LaTeX2e additions
|
||||
for name, value in [('quotedblbase', ',,'), ('quotesinglbase', ','),
|
||||
('textquotedbl', '"'), ('LaTeXe', 'LaTeX2e'),
|
||||
('e', '\\'), ('textquotedblleft', "``"),
|
||||
('textquotedblright', "''"), ('textquoteleft', "`"),
|
||||
('textquoteright', "'"), ('textbackslash', '\\'),
|
||||
('textbar', '|'), ('textless', '<'),
|
||||
('textgreater', '>'), ('textasciicircum', '^'),
|
||||
('Cpp', 'C++'), ('copyright', '')]:
|
||||
d[name] = chunk_type[PLAIN], value, 1
|
||||
convertible_csname = d.has_key
|
||||
conversion = d.get
|
||||
del d, name
|
||||
del d, name, value
|
||||
|
||||
##
|
||||
## \begin{ {func,data,exc}desc }{name}...
|
||||
|
@ -1261,6 +1283,7 @@ def changeit(buf, pp):
|
|||
while 1:
|
||||
# sanity check: length should always equal len(pp)
|
||||
if len(pp) != length:
|
||||
print i, pp[i]
|
||||
raise 'FATAL', 'inconsistent length. thought ' + `length` + ', but should really be ' + `len(pp)`
|
||||
if i >= length:
|
||||
break
|
||||
|
@ -1404,7 +1427,7 @@ def changeit(buf, pp):
|
|||
raise error, 'Sorry, expected plain text argument'
|
||||
hist.itemargmacro = s(buf, pp[i].data)
|
||||
if convertible_csname(hist.itemargmacro):
|
||||
hist.itemargmacro = conversion(hist.itemargmacro)
|
||||
hist.itemargmacro = conversion(hist.itemargmacro)[1]
|
||||
del pp[i:newi]
|
||||
length = length - (newi-i)
|
||||
|
||||
|
@ -1466,7 +1489,7 @@ def changeit(buf, pp):
|
|||
length = length + len(chunks) - 1
|
||||
i = i + len(chunks) - 1
|
||||
|
||||
elif envname in ('sloppypar', 'flushleft'):
|
||||
elif envname in ('sloppypar', 'flushleft', 'document'):
|
||||
pass
|
||||
|
||||
else:
|
||||
|
@ -1523,7 +1546,7 @@ def changeit(buf, pp):
|
|||
chunk(PLAIN, ch.where, "deffn")])]
|
||||
i, length = i+2, length+2
|
||||
|
||||
elif envname in ('seealso', 'sloppypar', 'flushleft'):
|
||||
elif envname in ('seealso', 'sloppypar', 'flushleft', 'document'):
|
||||
pass
|
||||
|
||||
else:
|
||||
|
@ -1541,9 +1564,11 @@ def changeit(buf, pp):
|
|||
pp[i:i+1]=cp + [
|
||||
chunk(PLAIN, ch.where, ']')]
|
||||
length = length+len(cp)
|
||||
|
||||
elif s_buf_data in ignoredcommands:
|
||||
del pp[i-1]
|
||||
i, length = i-1, length-1
|
||||
|
||||
elif s_buf_data == '@' and \
|
||||
i != length and \
|
||||
pp[i].chtype == chunk_type[PLAIN] and \
|
||||
|
@ -1551,10 +1576,22 @@ def changeit(buf, pp):
|
|||
# \@. --> \. --> @.
|
||||
ch.data = '.'
|
||||
del pp[i]
|
||||
length = length-1
|
||||
length = length - 1
|
||||
|
||||
elif convertible_csname(s_buf_data):
|
||||
ch.chtype, ch.data, nix = conversion(s_buf_data)
|
||||
try:
|
||||
if nix and pp[i].chtype == chunk_type[GROUP] \
|
||||
and len(pp[i].data) == 0:
|
||||
del pp[i]
|
||||
length = length - 1
|
||||
except IndexError:
|
||||
pass
|
||||
|
||||
elif s_buf_data == '\\':
|
||||
# \\ --> \* --> @*
|
||||
ch.data = '*'
|
||||
|
||||
elif len(s_buf_data) == 1 and \
|
||||
s_buf_data in onlylatexspecial:
|
||||
ch.chtype = chunk_type[PLAIN]
|
||||
|
@ -1567,6 +1604,11 @@ def changeit(buf, pp):
|
|||
del pp[i]
|
||||
length = length-1
|
||||
|
||||
elif s_buf_data == "appendix":
|
||||
hist.chaptertype = "appendix"
|
||||
del pp[i-1]
|
||||
i, length = i-1, length-1
|
||||
|
||||
elif hist.inargs and s_buf_data in inargsselves:
|
||||
# This is the special processing of the
|
||||
# arguments of the \begin{funcdesc}... or
|
||||
|
@ -1597,6 +1639,13 @@ def changeit(buf, pp):
|
|||
i = i - 1
|
||||
length = length - 3
|
||||
|
||||
elif s_buf_data == 'renewcommand':
|
||||
print "ignoring redefinition of \\" \
|
||||
+ s(buf, pp[i].data[0].data)
|
||||
del pp[i-1:i+2]
|
||||
i = i - 1
|
||||
length = length - 3
|
||||
|
||||
elif s_buf_data == 'mbox':
|
||||
stuff = pp[i].data
|
||||
pp[i-1:i+1] = stuff
|
||||
|
@ -1672,6 +1721,7 @@ def changeit(buf, pp):
|
|||
or pp[i].data != []:
|
||||
pp.insert(i, chunk(GROUP, ch.where, []))
|
||||
i, length = i+1, length+1
|
||||
|
||||
elif s_buf_data in themselves:
|
||||
# \UNIX --> &UNIX;
|
||||
ch.chtype = chunk_type[PLAIN]
|
||||
|
@ -1680,8 +1730,14 @@ def changeit(buf, pp):
|
|||
and pp[i].data == []:
|
||||
del pp[i]
|
||||
length = length-1
|
||||
elif s_buf_data in for_texi:
|
||||
pass
|
||||
|
||||
## elif s_buf_data == 'copyright':
|
||||
## if (pp[i].chtype == chunk_type[GROUP]
|
||||
## and not pp[i].data):
|
||||
## del pp[i]
|
||||
## length = length - 1
|
||||
## del pp[i-1]
|
||||
## i, length = i-1, length-1
|
||||
|
||||
elif s_buf_data == 'manpage':
|
||||
ch.data = 'emph'
|
||||
|
@ -1689,10 +1745,6 @@ def changeit(buf, pp):
|
|||
pp[i+1].data = "(%s)" % sect
|
||||
pp[i+1].chtype = chunk_type[PLAIN]
|
||||
|
||||
elif s_buf_data == 'e':
|
||||
# "\e" --> "\"
|
||||
ch.data = '\\'
|
||||
ch.chtype = chunk_type[PLAIN]
|
||||
elif s_buf_data in ('lineiii', 'lineii'):
|
||||
# This is the most tricky one
|
||||
# \lineiii{a1}{a2}[{a3}] -->
|
||||
|
@ -1726,11 +1778,14 @@ def changeit(buf, pp):
|
|||
if length != len(pp):
|
||||
raise 'IN LINEIII IS THE ERR', `i`
|
||||
|
||||
elif s_buf_data in ('chapter', 'section', 'subsection', 'subsubsection'):
|
||||
elif s_buf_data in ('chapter', 'section',
|
||||
'subsection', 'subsubsection'):
|
||||
#\xxxsection{A} ---->
|
||||
# @node A, , ,
|
||||
# @xxxsection A
|
||||
## also: remove commas and quotes
|
||||
if s_buf_data == "chapter":
|
||||
ch.data = hist.chaptertype
|
||||
ch.chtype = chunk_type[CSLINE]
|
||||
length, newi = getnextarg(length, buf, pp, i)
|
||||
afternodenamecmd = next_command_p(length, buf,
|
||||
|
@ -1915,7 +1970,7 @@ def changeit(buf, pp):
|
|||
pp.insert(i, chunk(GROUP, ch.where, ingroupch))
|
||||
length, i = length+1, i+1
|
||||
|
||||
elif s_buf_data in ('stindex', 'kwindex'):
|
||||
elif s_buf_data == 'stindex':
|
||||
# XXX must actually go to newindex st
|
||||
what = (s_buf_data[:2] == "st") and "statement" or "keyword"
|
||||
wh = ch.where
|
||||
|
@ -2091,12 +2146,13 @@ def changeit(buf, pp):
|
|||
ch.chtype = PLAIN
|
||||
ch.data = " "
|
||||
|
||||
elif s_buf_data in ('usepackage', 'input'):
|
||||
del pp[i-1:i+1]
|
||||
i, length = i-1, length-2
|
||||
|
||||
elif s_buf_data in ('noindent', 'indexsubitem', 'footnote'):
|
||||
pass
|
||||
|
||||
elif convertible_csname(s_buf_data):
|
||||
ch.data = conversion(s_buf_data)
|
||||
|
||||
elif s_buf_data == 'label':
|
||||
name = s(buf, pp[i].data[0].data)
|
||||
del pp[i-1:i+1]
|
||||
|
@ -2110,11 +2166,6 @@ def changeit(buf, pp):
|
|||
del pp[i]
|
||||
length = length - 1
|
||||
|
||||
elif s_buf_data == 'Large':
|
||||
del pp[i-1]
|
||||
i = i - 1
|
||||
length = length - 1
|
||||
|
||||
elif s_buf_data == 'ref':
|
||||
name = s(buf, pp[i].data[0].data)
|
||||
if label_nodes.has_key(name):
|
||||
|
@ -2154,12 +2205,12 @@ def dumpit(buf, wm, pp):
|
|||
|
||||
if ch.chtype == chunk_type[CSNAME]:
|
||||
s_buf_data = s(buf, ch.data)
|
||||
if s_buf_data == 'e':
|
||||
wm('\\')
|
||||
continue
|
||||
if s_buf_data == '$':
|
||||
wm('$')
|
||||
continue
|
||||
## if s_buf_data == 'e':
|
||||
## wm('\\')
|
||||
## continue
|
||||
## if s_buf_data == '$':
|
||||
## wm('$')
|
||||
## continue
|
||||
wm('@' + s_buf_data)
|
||||
if s_buf_data == 'node' and \
|
||||
pp[i].chtype == chunk_type[PLAIN] and \
|
||||
|
@ -2232,13 +2283,11 @@ def dumpit(buf, wm, pp):
|
|||
wm('\n')
|
||||
|
||||
elif ch.chtype == chunk_type[COMMENT]:
|
||||
## print 'COMMENT: previous chunk =', pp[i-2]
|
||||
## if pp[i-2].chtype == chunk_type[PLAIN]:
|
||||
## print 'PLAINTEXT =', `s(buf, pp[i-2].data)`
|
||||
if s(buf, ch.data) and \
|
||||
regex.match('^[ \t]*$', s(buf, ch.data)) < 0:
|
||||
if i >= 2 \
|
||||
and pp[i-2].chtype not in (chunk_type[ENDLINE], chunk_type[DENDLINE]) \
|
||||
and pp[i-2].chtype not in (chunk_type[ENDLINE],
|
||||
chunk_type[DENDLINE]) \
|
||||
and not (pp[i-2].chtype == chunk_type[PLAIN]
|
||||
and regex.match('\\(.\\|\n\\)*[ \t]*\n$', s(buf, pp[i-2].data)) >= 0):
|
||||
wm('\n')
|
||||
|
|
Loading…
Reference in New Issue