OpenERP 5.0 relies on ReportLab 2.3 to create PDF reports. However, ReportLab 2.3 does not support RTL languages, and thus OpenERP does not support them as well. The following steps fix the issue in both ReportLab 2.3 and OpenERP 5.0.

To enable RTL language support in ReportLab 2.3, we need to modify some of its code, and build two supporting libraries, namely FriBiDi2 and PyFribidi2.

The following steps were tested on Ubuntu 9.04 (Jaunty). First, remove older libraries and install libraries necessary for building the new sources.

sudo apt-get purge libfribidi0 python-pyfribidi && sudo apt-get install autoconf gcc

Second, download FriBiDi2 (the latest at the time of writing is fribidi-0.19.2), extract it, open a shell in the extracted directory and execute the following command:

./configure && make && sudo make install

Next, download PyFribidi2 (the latest at the time of writing is pyfribidi2-0.8.0), extract it, open a shell in the extracted directory and execute the following command:

sed -i -r 's/\bfribidi2(\b|_)/fribidi\1/g' configure.in configure setup.py

[ `python -c 'import sys; print sys.version[:3]'` == '2.6' ]
&& sed -i -r 's/\bsite-packages\b/dist-packages/g' configure configure.in

./configure && make && sudo make install

Finally, verify the installation by opening a shell, changing to the home directory, starting the Python interpreter and importing pyfribidi2:

cd
python
>>> import pyfribidi2

Note: If you get an error about libfribidi.so.0 not being found, try to reboot your machine. If the error still persists, you may need to edit ”/etc/ld.so.conf”, adding the following line at the end of the file: ”/usr/local/lib” (without the quotes). Then run `sudo ldconfig`.

Three files need to be modified in ReportLab’s code. The default installation of ReportLab in Ubuntu with Python 2.6 places it in the directory ”/usr/lib/python2.6/dist-packages/reportlab/”. We will need to modify the files there using an editor that is run in super user mode (e.g. `sudo vim`). I will refer to the files below using the relative path in ReportLab’s director.

We need to modify the search path that ReportLab uses to search for fonts to include the default path on Ubuntu.

In the variable T1SearchPath, add the following value as the last element in the tuple (approximately at line 67):

'/usr/share/fonts/type1',

Do the same for TTFSearchPath, adding the following value as the last element in the tuple (approximately at line 85):

'/usr/share/fonts/truetype',

And do the same for CMapSearchPath, adding the following value as the last element in the tuple (approximately at line 113):

'/usr/share/fonts/cmap',

We need to import PyFribidi2 and call it before text is output to the PDF.

Add the following lines after the import statements (approximately at line 19):

# try to import pyfribidi
try:
    import pyfribidi2 as pyfribidi
    log2vis = pyfribidi.log2vis
    DIR_ON = pyfribidi.ON
    DIR_LTR = pyfribidi.LTR
    DIR_RTL = pyfribidi.RTL
except:
    import warnings
    warnings.warn('pyfribidi is not installed - RTL not supported')
    log2vis = None
    DIR_ON = DIR_LTR = DIR_RTL = None

Add a parameter to specify direction in PDFTextObject.init (approximately at line 135), so that the method declaration becomes as follows:

    def __init__(self, canvas, x=0,y=0, direction = 'LTR'):
        self.direction = direction

Finally, add a call to PyFribidi2 at the top of PDFTextObject._formatText (approximately at line 314), so that it becomes like this:

    def _formatText(self, text):
        "Generates PDF text output operator(s)"
        # Use pyfribidi to write the text in the correct visual order.
        directions = { 'LTR': DIR_LTR, 'RTL': DIR_RTL }
        text = log2vis(text, directions.get(self.direction, DIR_ON))

Right before the comment ”#now the font for the rest of the paragraph”, add the following lines (approximately at line 1337):

                # set the paragraph direction
                if self.style.wordWrap == 'RTL':
                    tx.direction = 'RTL'

The fixes below have been adapted from the ReportsUnicode wiki on OpenObject’s website. However, the code in the wiki does not work correctly, so I have fixed a few things in it (and will hopefully fix the wiki soon). Only two files need to be changed: ”[bin/]report/render/rml2pdf/init.py” and ”[bin/]report/render/rml2pdf/trml2pdf.py”.

Replace the code in this file by the following code:

import os, stat
from tools.misc import debug
from reportlab import rl_config
from reportlab.pdfbase import pdfmetrics
from reportlab.pdfbase import ttfonts
from reportlab.lib.fonts import addMapping

def rl_isreg(filename, dirname):
    try:
        st = os.stat(os.path.join(dirname, filename))
    except OSError, reason:
        if reason.errno == 2:
            return False
        raise
    return stat.S_ISREG(st.st_mode) or stat.S_ISLNK(st.st_mode)

for dirname in rl_config.TTFSearchPath:
    for root, dirs, files in os.walk(dirname): #@UnusedVariable
        for file in [x for x in files
                     if x.lower().endswith('.ttf') and rl_isreg(x, root)
                     ]:
            filename = os.path.join(root, file)
            try:
                face = ttfonts.TTFontFace(filename)
                face.extractInfo(1)
                pdfmetrics.registerFont(ttfonts.TTFont(face.name, filename, asciiReadable=0))
                addMapping(face.familyName, face.bold, face.italic, face.name)
            except:
                pass

from trml2pdf import parseString, parseNode

Right before the parseNode function, add the following function (approximately at line 777). It replaces a number of standard PDF fonts with alternatives that support Unicode characters. I haven’t tried all of them, but Helvetica –> DejaVuSans works for me.

def changeFonts(data):
    fontmap = {
        'Times-Roman':                   'DejaVuSerif',
        'Times-BoldItalic':              'DejaVuSerif-BoldItalic',
        'Times-Bold':                    'DejaVuSerif-Bold',
        'Times-Italic':                  'DejaVuSerif-Italic',

        'Helvetica':                     'DejaVuSans',
        'Helvetica-BoldItalic':          'DejaVuSans-BoldOblique',
        'Helvetica-Bold':                'DejaVuSans-Bold',
        'Helvetica-Italic':              'DejaVuSans-Oblique',

        'Courier':                       'DejaVuSansMono',
        'Courier-Bold':                  'DejaVuSansMono-Bold',
        'Courier-BoldItalic':            'DejaVuSansMono-BoldOblique',
        'Courier-Italic':                'DejaVuSansMono-Oblique',

        'Helvetica-ExtraLight':          'DejaVuSans-ExtraLight',

        'TimesCondensed-Roman':          'DejaVuSerifCondensed',
        'TimesCondensed-BoldItalic':     'DejaVuSerifCondensed-BoldItalic',
        'TimesCondensed-Bold':           'DejaVuSerifCondensed-Bold',
        'TimesCondensed-Italic':         'DejaVuSerifCondensed-Italic',

        'HelveticaCondensed':            'DejaVuSansCondensed',
        'HelveticaCondensed-BoldItalic': 'DejaVuSansCondensed-BoldOblique',
        'HelveticaCondensed-Bold':       'DejaVuSansCondensed-Bold',
        'HelveticaCondensed-Italic':     'DejaVuSansCondensed-Oblique',
    }
    for old, new in fontmap.iteritems():
        data = data.replace('"'+old+'"', '"'+new+'"')
    return data

Then, put the following line at the very top of each of parseNode and parseString (approximately at lines 812 and 820), before the call eTree.XML(rml):

    rml = changeFonts(rml)

(Tested on OpenERP Server 5.0.5 only)

In the method _rml_flowable._textual, modify the line that reads rc1 += etree.tostring(txt_n, encoding=unicode) (approximately at line 460) to become:

                rc1 += etree.tostring(txt_n, encoding=unicode)

That should be all. OpenERP should be all set and ready for creating Arabic reports.