component/ocropy Add the ocropy library for OCR

This commit add component needed for Optical Character Recognition:
Ocropy, and the patch that need to be applied to make it work with erp5.

The Ocropy egg on Pypi is empty, so this component have to be built from
source. The default neural network model is downloaded from the author
website and installed in python2.7/share/ocropus.

The patch modify how the setup works to allow the C functions in
native.py to be compiled during setup (removing the need to make zope
use gcc) and provide a way to specify the model path for Ocropy.
The patched setup now need to call some modules from different eggs,
whose path needs to be put into sys.path before it can be run.

More models are available here:
https://github.com/tmbdev/ocropy/wiki/Models

component/ocropy/buildout.cfg

+[buildout]
+extends =
+  ../git/buildout.cfg
+  ../gzip/buildout.cfg
+  ../scipy/buildout.cfg
+  ../lxml-python/buildout.cfg
+  ../matplotlib/buildout.cfg
+  ../numpy/buildout.cfg
+  ../patch/buildout.cfg
+  ../pillow/buildout.cfg
+  ../numpy/buildout.cfg
+
+parts =
+  matplotlib
+  ocropy
+
+[ocropy-eng-traineddata]
+recipe = hexagonit.recipe.download
+destination = ${python2.7:location}/share/ocropus
+filename = en-default.pyrnn.gz
+md5sum = cedd140c7d7650e910f0550ad0f04727
+download-only = true
+url = http://www.tmbdev.net/en-default.pyrnn.gz
+
+[ocropy]
+recipe = slapos.recipe.build
+egg = ocropy
+setup-eggs =
+  ${numpy:egg}
+  ${matplotlib:egg}
+  ${pillow-python:egg}
+url = https://github.com/tmbdev/ocropy/tarball/4efbddca22bb2f0c639af0694e7a1386f2f097b5
+md5sum = 240b8866dd7248816e01af469a328c09
+patch = ${:_profile_base_location_}/ocropy.patch
+python-bin = ${python2.7:location}/bin/python
+script =
+  import os
+  import sys
+  from itertools import chain
+  
+  # Update of the path to include eggs location
+  list_modules = lambda dir: [dir + file for file in os.listdir(dir)]
+  sys.path[0:0] = chain.from_iterable(map(list_modules, ['${buildout:develop-eggs-directory}/', '${buildout:eggs-directory}/']))
+  
+  os.environ['PYTHONPATH'] = ':'.join(sys.path)
+  os.environ['OCROPY_MODEL_PATH'] = "${ocropy-eng-traineddata:location}/${ocropy-eng-traineddata:filename}"
+  workingdir = guessworkdir(self.extract(self.download(%(url)r, %(md5sum)r)))
+  os.chdir(workingdir)
+  self.applyPatchList(self.options.get('patch'), \
+    patch_binary='${patch:location}/bin/patch', \
+    patch_options='-p0', cwd=workingdir)
+  call([self.options.get('python-bin'), os.path.join(workingdir, 'setup.py'),
+        'install'])

component/ocropy/ocropy.patch

+diff --git ocrolib/__init__.py ocrolib/__init__.py
+index 1e0d627..81e85fb 100644
+--- ocrolib/__init__.py
++++ ocrolib/__init__.py
+@@ -1,7 +1,7 @@
+ __all__ = [
+     "binnednn","cairoextras","common","components","dbtables",
+     "fgen","gmmtree","gtkyield","hocr","lang","native",
+-    "mlp","multiclass","default","lineest"
++    "mlp","multiclass","default","lineest", "psegutils"
+ ]
+ 
+ ################################################################
+@@ -9,5 +9,6 @@ __all__ = [
+ ################################################################
+ 
+ import default
++from psegutils import *
+ from common import *
+ from default import traceback as trace
+diff --git ocrolib/common.py ocrolib/common.py
+index 27c0f26..b68c5c8 100644
+--- ocrolib/common.py
++++ ocrolib/common.py
+@@ -14,6 +14,7 @@ import unicodedata
+ import inspect
+ import glob
+ import cPickle
++import gzip
+ from ocrolib.exceptions import (BadClassLabel, BadInput, FileNotFound,
+                                 OcropusException)
+ 
+@@ -428,6 +429,7 @@ def unpickle_find_global(mname,cname):
+         exec "import "+mname
+     return getattr(sys.modules[mname],cname)
+ 
++
+ def load_object(fname,zip=0,nofind=0,verbose=0):
+     """Loads an object from disk. By default, this handles zipped files
+     and searches in the usual places for OCRopus. It also handles some
+@@ -439,8 +441,7 @@ def load_object(fname,zip=0,nofind=0,verbose=0):
+     if zip==0 and fname.endswith(".gz"):
+         zip = 1
+     if zip>0:
+-        # with gzip.GzipFile(fname,"rb") as stream:
+-        with os.popen("gunzip < '%s'"%fname,"rb") as stream:
++        with gzip.GzipFile(fname,"rb") as stream:
+             unpickler = cPickle.Unpickler(stream)
+             unpickler.find_global = unpickle_find_global
+             return unpickler.load()
+diff --git ocrolib/native.py ocrolib/native.py
+index b7a207f..240450b 100644
+--- ocrolib/native.py
++++ ocrolib/native.py
+@@ -44,6 +44,7 @@ class CompileError(Exception):
+ 
+ def compile_and_find(c_string,prefix=".pynative",opt="-g -O4",libs="-lm",
+                      options="-shared -fopenmp -std=c99 -fPIC",verbose=0):
++    prefix = os.path.join(os.path.dirname(__file__), prefix)
+     if not os.path.exists(prefix):
+         os.mkdir(prefix)
+     m = hashlib.md5()
+diff --git setup.py setup.py
+index 2ec5832..151bf55 100644
+--- setup.py
++++ setup.py
+@@ -10,7 +10,9 @@ assert sys.version_info[0]==2 and sys.version_info[1]>=7,\
+ from distutils.core import setup #, Extension, Command
+ #from distutils.command.install_data import install_data
+ 
+-if not os.path.exists("models/en-default.pyrnn.gz"):
++models = os.environ.get('OCROPY_MODEL_PATH', '').split(':') or \
++    [c for c in glob.glob("models/*pyrnn.gz")]
++if not models:
+     print()
+     print("You should download the default model 'en-default.pyrnn.gz'")
+     print("and put it into ./models.")
+@@ -18,16 +20,23 @@ if not os.path.exists("models/en-default.pyrnn.gz"):
+     print("Check https://github.com/tmbdev/ocropy for the location")
+     print("of model files.")
+     print()
++    sys.exit(1)
+ 
+-models = [c for c in glob.glob("models/*pyrnn.gz")]
+ scripts = [c for c in glob.glob("ocropus-*") if "." not in c and "~" not in c]
+ 
++# compile pynative files now and include them in the build
++sys.path.insert(0, os.curdir)
++import ocrolib.nutils
++pynative_files = [os.path.join(*c.split(os.path.sep)[1:]) \
++    for c in glob.glob('ocrolib/.pynative/*')]
++
+ setup(
+     name = 'ocropy',
+     version = 'v1.0',
+     author = "Thomas Breuel",
+     description = "The OCRopy RNN-based Text Line Recognizer",
+     packages = ["ocrolib"],
++    package_data = {'ocrolib': pynative_files},
+     data_files= [('share/ocropus', models)],
+     scripts = scripts,
+     )