pfwbged.ocrloader/ocrloader.py

#! /usr/bin/env python

import ConfigParser
import urllib
import datetime
import email
import email.parser
import imaplib
import os
import os.path
import requests
import string
import subprocess
import sys
import tempfile
import time
import shutil
import socket

import optparse

parser = optparse.OptionParser()
parser.add_option('--config', dest='config', default='ocrloader.ini')
(options, args) = parser.parse_args()

config_filepath = options.config


import logging
import logging.handlers
logging.basicConfig(level=logging.DEBUG,
                format='%(asctime)s %(message)s')
logging.getLogger().addHandler(logging.StreamHandler())
logging.getLogger().addHandler(logging.handlers.RotatingFileHandler(
                        filename='/var/log/pfwbged.ocrloader.log',
                        maxBytes=1000*1000
        ))

def process(cfg, filename, payload, enable_ocr=True):
    ocr_filename = os.path.join(cfg.get('ocrized_directory'), filename)
    if not enable_ocr:
        if not os.path.exists(ocr_filename):
            logging.debug('  skipping OCR phase')
            fd = file(ocr_filename, 'w')
            fd.write(payload)
            fd.close()

    if not os.path.exists(ocr_filename) and enable_ocr:
        fd, tmpfilename = tempfile.mkstemp(suffix='.pdf', prefix='ocrloader-')
        os.write(fd, payload)
        os.close(fd)
        logging.debug('  running OCR on file (%s / %s)' % (tmpfilename, ocr_filename))
        cmd = ['/opt/ABBYYOCR9/abbyyocr9', '-pi',
                         '-rl', 'French', '-if', tmpfilename,
                         '-f', 'PDF', '-pem', 'ImageOnText', '-pfpr', '150',
                         '-pfq', '100', '-of', ocr_filename]
        logging.debug('  %s' % ' '.join(cmd))
        subprocess.call(cmd)
        if os.stat(ocr_filename)[6] == 0:
            os.unlink(ocr_filename)
        if not os.path.exists(ocr_filename):
            logging.error('failed to OCR %s', filename)
            file('/tmp/' + filename, 'w').write(payload) # keep it for inspection
            return False

    if os.stat(ocr_filename)[6] == 0:
        os.unlink(ocr_filename)

    if cfg.get('store_path'):
        logging.debug('  storing file locally')
        shutil.copy(ocr_filename, os.path.join(cfg.get('store_path'), filename))
        return True
    else:
        logging.debug('  uploading file')
        now = datetime.datetime.now()
        title = string.Template(cfg.get('title')).substitute(
                {'date': now.strftime('%d/%m/%Y'),
                 'time': now.strftime('%H:%M:%S')})
        t = subprocess.call(['curl', '-v', '--insecure', '-X', 'POST',
                  '-F', 'form.widgets.file=@%s;filename=%s;type=application/pdf' % (ocr_filename, filename),
                  '-F', 'form.buttons.import=Import',
                  '-F', 'form.widgets.portal_type=%s' % cfg.get('default_type'),
                  '-F', 'form.widgets.location=%s' % cfg.get('default_directory'),
                  '-F', 'form.widgets.owner=%s' % cfg.get('user'),
                  '-F', 'form.widgets.treating_groups=%s' % (
                      cfg.get('treating_groups') or cfg.get('treating_group') or ''),
                  '-F', 'form.widgets.recipient_groups=%s' % (cfg.get('recipient_groups') or ''),
                  '-F', 'form.widgets.title=%s' % title,
                  '-F', 'form.widgets.notification_recipients=%s' % cfg.get('notification_recipients', ''),
                  '-F', 'form.widgets.keywords=%s' % cfg.get('keywords', ''),
                  '-F', 'form.widgets.transitions_to_apply=%s' % cfg.get('transitions_to_apply', ''),
                  '-u', '%s:%s' % (cfg.get('ged_username'), cfg.get('ged_password')),
                  '%s/@@fileimport' % cfg.get('ged_base_url'),])
        return (t == 0)


#    try:
#        r = requests.post(cfg.get('ged_base_url') + '/@@fileimport',
#            auth=(cfg.get('ged_username'), cfg.get('ged_password')),
#            verify=False,
#            proxies={'https': 'http://172.23.3.30:3128'},
#            files={'form.widgets.file': (filename, file(ocr_filename))},
#            data={'form.buttons.import': 'Import',
#                  'form.widgets.portal_type': cfg.get('default_type'),
#                  'form.widgets.location': cfg.get('default_directory'),
#                  'form.widgets.owner': cfg.get('user'),
#                 })
#    except Exception, e:
#        print e
#        return False
#
#    if r.status_code != requests.codes.ok:
#        file('/tmp/error.html', 'w').write(r.text)
#    return (r.status_code == requests.codes.ok)


while True:
    cfg = ConfigParser.ConfigParser()
    cfg.read(config_filepath)
    for section in cfg.sections():
        logging.debug('processing %s', section)
        ged_base_url = cfg.get(section, 'ged_base_url')

        if section.startswith('/'):
            # handle dropped files
            for basedir, dirnames, filenames in os.walk(section):
                for filename in filenames:
                    if filename.endswith('.uploaded'):
                        continue
                    filepath = os.path.realpath(os.path.join(basedir, filename))
                    if not filepath.startswith(basedir):
                        # check the real path as an attacker could create a
                        # symlink to whatever directory and cause total
                        # destruction of it. (as well as the upload of its
                        # contents to the GED...).
                        logging.warn('wrong base dir for %s', filepath)
                        continue
                    payload = file(filepath).read()
                    logging.debug('  uploading file %s', filepath)
                    enable_ocr = (not filename.startswith('no-ocr-') and filename.endswith('.pdf'))
                    if not process(dict(cfg.items(section)), filename, payload, enable_ocr):
                        logging.error('  error processing %s', filepath)
                    else:
                        if cfg.get(section, 'file_success_action') == 'delete':
                            os.unlink(filepath)
                        else:
                            os.rename(filepath, filepath + '.uploaded')
            continue

        # handle imap mailboxes
        imap_server = cfg.get(section, 'imap_server')
        ssl = cfg.getboolean(section, 'ssl')

        try:
            if ssl:
                M = imaplib.IMAP4_SSL(host=imap_server)
            else:
                M = imaplib.IMAP4(host_imap_server)
        except:
            logging.error('failed to connect to imap server')
            time.sleep(30)
            continue
        try:
            M.login(section, cfg.get(section, 'password'))
        except imaplib.IMAP4.error:
            continue
        try:
            M.select()
        except socket.error:
            logging.error('failure talking to imap server')
            continue
        typ, data = M.search(None, '(NOT SEEN)')
        for num in data[0].split():
            try:
                typ, data = M.fetch(num, '(RFC822)')
            except socket.error:
                logging.error('failure talking to imap server')
                break
            msg = email.parser.Parser().parsestr(data[0][1])
            enable_ocr = True
            if 'disable_ocr' in msg['Subject']:
                enable_ocr = False
            for part in msg.walk():
                if part.get_content_type() == 'application/pdf':
                    filename = part.get_filename()
                    logging.info('  handling %s', filename)
                    payload = part.get_payload(decode=True)
                    if not process(dict(cfg.items(section)), filename, payload, enable_ocr):
                        logging.error('  error -> marking as unseen')
                        try:
                            M.store(num, '-FLAGS', r'\Seen')
                        except socket.error:
                            logging.error('failure talking to imap server')
                            pass
                    break
        try:
            M.close()
            M.logout()
        except socket.error:
            logging.error('failure talking to imap server')
            pass

    logging.debug('waiting a bit %s', time.strftime('%Y-%m-%d %H:%M:%S'))
    time.sleep(30)