diff --git a/bin/tools/__init__.py b/bin/tools/__init__.py index 59428bcf710..b84026572a7 100644 --- a/bin/tools/__init__.py +++ b/bin/tools/__init__.py @@ -1,7 +1,7 @@ # -*- encoding: utf-8 -*- ############################################################################## # -# OpenERP, Open Source Management Solution +# OpenERP, Open Source Management Solution # Copyright (C) 2004-2008 Tiny SPRL (). All Rights Reserved # $Id$ # @@ -26,6 +26,7 @@ from translate import * from graph import graph from amount_to_text import * from amount_to_text_en import * +from pdf_utils import * # vim:expandtab:smartindent:tabstop=4:softtabstop=4:shiftwidth=4: diff --git a/bin/tools/pdf_utils.py b/bin/tools/pdf_utils.py new file mode 100644 index 00000000000..7699177a1d6 --- /dev/null +++ b/bin/tools/pdf_utils.py @@ -0,0 +1,112 @@ +# -*- encoding: utf-8 -*- +############################################################################## +# +# OpenERP, Open Source Management Solution +# Copyright (C) 2004-2008 Tiny SPRL (). All Rights Reserved +# $Id$ +# +# 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 3 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, see . +# +############################################################################## +""" Copyright (c) 2003-2007 LOGILAB S.A. (Paris, FRANCE). + http://www.logilab.fr/ -- mailto:contact@logilab.fr + +manipulate pdf and fdf files. pdftk recommended. + +Notes regarding pdftk, pdf forms and fdf files (form definition file) +fields names can be extracted with: + pdftk orig.pdf generate_fdf output truc.fdf +to merge fdf and pdf: + pdftk orig.pdf fill_form test.fdf output result.pdf [flatten] +without flatten, one could further edit the resulting form. +with flatten, everything is turned into text. +""" + +import os +import netsvc,logging + + +HEAD="""%FDF-1.2 +%\xE2\xE3\xCF\xD3 +1 0 obj +<< +/FDF +<< +/Fields [ +""" + +TAIL="""] +>> +>> +endobj +trailer + +<< +/Root 1 0 R +>> +%%EOF +""" + +def output_field( f ): + return "\xfe\xff" + "".join( [ "\x00"+c for c in f ] ) + +def extract_keys(lines): + keys = [] + for line in lines: + if line.startswith('/V'): + pass #print 'value',line + elif line.startswith('/T'): + key = line[7:-2] + key = ''.join(key.split('\x00')) + keys.append( key ) + return keys + +def write_field(out, key, value): + out.write("<<\n") + if value: + out.write("/V (%s)\n" %value) + else: + out.write("/V /\n") + out.write("/T (%s)\n" % output_field(key) ) + out.write(">> \n") + +def write_fields(out, fields): + out.write(HEAD) + for key in fields: + value = fields[key] + write_field(out, key, value) +# write_field(out, key+"a", value) # pour copie-carbone sur autres pages + out.write(TAIL) + +def extract_keys_from_pdf(filename): + # what about using 'pdftk filename dump_data_fields' and parsing the output ? + os.system('pdftk %s generate_fdf output /tmp/toto.fdf' % filename) + lines = file('/tmp/toto.fdf').readlines() + return extract_keys(lines) + + +def fill_pdf(infile, outfile, fields): + write_fields(file('/tmp/toto.fdf', 'w'), fields) + os.system('pdftk %s fill_form /tmp/toto.fdf output %s flatten' % (infile, outfile)) + +def testfill_pdf(infile, outfile): + keys = extract_keys_from_pdf(infile) + fields = [] + for key in keys: + fields.append( (key, key, '') ) + fill_pdf(infile, outfile, fields) + + +# vim:expandtab:smartindent:tabstop=4:softtabstop=4:shiftwidth=4: +