# This file is part of lyx2lyx # -*- coding: iso-8859-1 -*- # Copyright (C) 2003 José Matos # # This program is free software; you can redistribute it and/or # modify it under the terms of the GNU General Public License # as published by the Free Software Foundation; either version 2 # of the License, or (at your option) any later version. # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # You should have received a copy of the GNU General Public License # along with this program; if not, write to the Free Software # Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. import re from parser_tools import find_token, find_tokens, find_end_of_inset, find_end_of from sys import stderr from string import replace, split, find, replace, strip, join def add_end_layout(lines): i = find_token(lines, '\\layout', 0) if i == -1: return i = i + 1 struct_stack = ["\\layout"] while 1: i = find_tokens(lines, ["\\begin_inset", "\\end_inset", "\\layout", "\\begin_deeper", "\\end_deeper", "\\the_end"], i) token = split(lines[i])[0] if token == "\\begin_inset": struct_stack.append(token) i = i + 1 continue if token == "\\end_inset": tail = struct_stack.pop() if tail == "\\layout": lines.insert(i,"\\end_layout") i = i + 1 #Check if it is the correct tag struct_stack.pop() i = i + 1 continue if token == "\\layout": tail = struct_stack.pop() if tail == token: lines.insert(i,"\\end_layout") i = i + 2 else: struct_stack.append(tail) i = i + 1 struct_stack.append(token) continue if token == "\\begin_deeper" or token == "\\end_deeper": lines.insert(i,"\\end_layout") i = i + 2 continue #case \end_document lines.insert(i, "\\end_layout") return def layout2begin_layout(lines): i = 0 while 1: i = find_token(lines, '\\layout', i) if i == -1: return lines[i] = replace(lines[i], '\\layout', '\\begin_layout') i = i + 1 def valignment_middle(lines, start, end): for i in range(start, end): if re.search('^<(column|cell) .*valignment="center".*>$', lines[i]): lines[i] = replace(lines[i], 'valignment="center"', 'valignment="middle"') def table_valignment_middle(lines): i = 0 while 1: i = find_token(lines, '\\begin_inset Tabular', i) if i == -1: return j = find_end_of_inset(lines, i + 1) if j == -1: #this should not happen valignment_middle(lines, i + 1, len(lines)) return valignment_middle(lines, i + 1, j) i = j + 1 def end_document(lines): i = find_token(lines, "\\the_end", 0) if i == -1: lines.append("\\end_document") return lines[i] = "\\end_document" ## # Convert line and page breaks # Old: #\layout Standard #\line_top \line_bottom \pagebreak_top \pagebreak_bottom \added_space_top xxx \added_space_bottom yyy #0 # # New: #\begin layout Standard # #\newpage # #\lyxline #\begin_inset VSpace xxx #\end_inset # #\end_layout #\begin_layout Standard # #0 #\end_layout #\begin_layout Standard # #\begin_inset VSpace xxx #\end_inset #\lyxline # #\newpage # #\end_layout def convert_breaks(lines): i = 0 while 1: i = find_token(lines, "\\begin_layout", i) if i == -1: return i = i + 1 line_top = find(lines[i],"\\line_top") line_bot = find(lines[i],"\\line_bottom") pb_top = find(lines[i],"\\pagebreak_top") pb_bot = find(lines[i],"\\pagebreak_bottom") vspace_top = find(lines[i],"\\added_space_top") vspace_bot = find(lines[i],"\\added_space_bottom") if line_top == -1 and line_bot == -1 and pb_bot == -1 and pb_top == -1 and vspace_top == -1 and vspace_bot == -1: continue for tag in "\\line_top", "\\line_bottom", "\\pagebreak_top", "\\pagebreak_bottom": lines[i] = replace(lines[i], tag, "") if vspace_top != -1: # the position could be change because of the removal of other # paragraph properties above vspace_top = find(lines[i],"\\added_space_top") tmp_list = split(lines[i][vspace_top:]) vspace_top_value = tmp_list[1] lines[i] = lines[i][:vspace_top] + join(tmp_list[2:]) if vspace_bot != -1: # the position could be change because of the removal of other # paragraph properties above vspace_bot = find(lines[i],"\\added_space_bottom") tmp_list = split(lines[i][vspace_bot:]) vspace_bot_value = tmp_list[1] lines[i] = lines[i][:vspace_bot] + join(tmp_list[2:]) lines[i] = strip(lines[i]) i = i + 1 # Create an empty paragraph for line and page break that belong # above the paragraph if pb_top !=-1 or line_top != -1 or vspace_bot != -1: paragraph_above = ['','\\begin_layout Standard','',''] if pb_top != -1: paragraph_above.extend(['\\newpage ','']) if vspace_top != -1: paragraph_above.extend(['\\begin_inset VSpace ' + vspace_top_value,'\\end_inset ','','']) if line_top != -1: paragraph_above.extend(['\\lyxline ','']) paragraph_above.extend(['\\end_layout','']) #inset new paragraph above the current paragraph lines[i-2:i-2] = paragraph_above i = i + len(paragraph_above) # Ensure that nested style are converted later. k = find_end_of(lines, i, "\\begin_layout", "\\end_layout") if k == -1: return if pb_top !=-1 or line_top != -1 or vspace_bot != -1: paragraph_bellow = ['','\\begin_layout Standard','',''] if line_bot != -1: paragraph_bellow.extend(['\\lyxline ','']) if vspace_bot != -1: paragraph_bellow.extend(['\\begin_inset VSpace ' + vspace_bot_value,'\\end_inset ','','']) if pb_bot != -1: paragraph_bellow.extend(['\\newpage ','']) paragraph_bellow.extend(['\\end_layout','']) #inset new paragraph above the current paragraph lines[k + 1: k + 1] = paragraph_bellow def convert(header, body): add_end_layout(body) layout2begin_layout(body) end_document(body) table_valignment_middle(body) convert_breaks(body) if __name__ == "__main__": pass