import xml.etree.cElementTree as ET
import psutil
from spreads.vendor.pathlib import Path
import spreads.util as util
from spreads.config import OptionTemplate
from spreads.plugin import HookPlugin, ProcessHooksMixin
IS_WIN = util.is_os('windows')
CLI_BIN = util.find_in_path('scantailor-cli')
GUI_BIN = util.find_in_path('scantailor')
if not CLI_BIN:
raise util.MissingDependencyException(
"Could not find executable `scantailor-cli`. Please"
" install the"
" appropriate package(s)!")
logger = logging.getLogger('spreadsplug.scantailor')
class ScanTailorPlugin(HookPlugin, ProcessHooksMixin):
__name__ = 'scantailor'
@classmethod
def configuration_template(cls):
conf = {
'autopilot':
OptionTemplate(value=True, docstring="Skip manual correction"),
'rotate':
OptionTemplate(value=False, docstring="Rotate pages"),
def process(self, pages, target_path):
""" Run the most recent image of every page through ScanTailor.
:param pages: Pages to be processed
:type pages: list of :py:class:`spreads.workflow.Page`
:param target_path: Base directory where rotated images are to be
stored
:type target_path: :py:class:`pathlib.Path`
"""
autopilot = self.config['autopilot'].get(bool)
if not autopilot and not util.find_in_path('scantailor'):
raise util.MissingDependencyException(
"Could not find executable `scantailor` in"
" $PATH. Please install the appropriate"
" package(s)!")
# Create temporary files/directories
projectfile = Path(tempfile.mkstemp(suffix='.ScanTailor')[1])
out_dir = Path(tempfile.mkdtemp(prefix='st-out'))
# Map input paths to their pages so we can more easily associate
# the generated output files with their pages later on
in_paths = {}
for page in pages:
fpath = page.get_latest_processed(image_only=True)
if fpath is None:
fpath = page.raw_image
in_paths[unicode(fpath)] = page
logger.info("Generating ScanTailor configuration")
self._generate_configuration(sorted(in_paths.keys()), projectfile,
out_dir)
if not autopilot:
logger.warn("If you are changing output settings (in the last "
"step, you *have* to run the last step from the GUI. "
"Due to a bug in ScanTailor, your settings would "
"otherwise be ignored.")
time.sleep(5)
logger.info("Opening ScanTailor GUI for manual adjustment")
util.get_subprocess([GUI_BIN, unicode(projectfile)])
# Check if the user already generated output files from the GUI
if not sum(1 for x in out_dir.glob('*.tif')) == len(pages):
logger.info("Generating output images from ScanTailor "
"configuration.")
self._generate_output(projectfile, out_dir, len(pages))
# Associate generated output files with our pages
for fname in out_dir.glob('*.tif'):
out_stem = fname.stem
for in_path, page in in_paths.iteritems():
if Path(in_path).stem == out_stem:
target_fname = target_path / fname.name
shutil.copyfile(unicode(fname), unicode(target_fname))
page.processed_images[self.__name__] = target_fname
break
else:
logger.warn(
"Could not find page for output file {0}".format(fname))
# Remove temporary files/directories
shutil.rmtree(unicode(out_dir))
# FIXME: This fails on Windows since there seems to be some non-gcable
# reference to the file around, but I currently cannot figure
# out where, so we just ignore the error...
try:
projectfile.unlink()
except WindowsError as e:
if e.errno == 32:
pass
import shutil
import subprocess
import tempfile
import time
import xml.etree.cElementTree as ET
from itertools import chain
import spreads.util as util
from spreads.config import OptionTemplate
from spreads.plugin import HookPlugin, ProcessHooksMixin
from spreads.vendor.pathlib import Path
BIN = util.find_in_path('tesseract')
if not BIN:
raise util.MissingDependencyException(
"Could not find executable `tesseract`. Please install the appropriate"
" package(s)!")
# Newer versions of Tesseract provide a flag to obtain a list of installed
# OCR languages, for older versions we have to read out the directory
# containing the training data for languages.
try:
AVAILABLE_LANGS = (util.get_subprocess(
[BIN, "--list-langs"],
stderr=subprocess.STDOUT,
stdout=subprocess.PIPE).communicate()[0].split("\n")[1:-1])
# There should be at least a single language
if not AVAILABLE_LANGS:
raise ValueError()
except (subprocess.CalledProcessError, ValueError):
AVAILABLE_LANGS = [