def __init__(self, em):
self.currentindex = 0
self.pdftotext = None
self.em = em
self.confdir = rclconfig.RclConfig().getConfDir()
cf_doocr = rclconfig.RclConfig().getConfParam("pdfocr")
cf_attach = rclconfig.RclConfig().getConfParam("pdfattach")
self.pdftotext = rclexecm.which("pdftotext")
if not self.pdftotext:
self.pdftotext = rclexecm.which("poppler/pdftotext")
# Check if we need to escape portions of text where old
# versions of pdftotext output raw HTML special characters.
self.needescape = True
try:
version = subprocess.check_output([self.pdftotext, "-v"],
stderr=subprocess.STDOUT)
major, minor, rev = version.split()[2].split('.')
# Don't know exactly when this changed but it's fixed in
# jessie 0.26.5
if int(major) > 0 or int(minor) >= 26:
self.needescape = False
except:
pass
# See if we'll try to perform OCR. Need the commands and the
# either the presence of a file in the config dir (historical)
# or a set config variable.
self.ocrpossible = False
if cf_doocr or os.path.isfile(os.path.join(self.confdir, "ocrpdf")):
self.tesseract = rclexecm.which("tesseract")
if self.tesseract:
self.pdftoppm = rclexecm.which("pdftoppm")
if self.pdftoppm:
self.ocrpossible = True
self.maybemaketmpdir()
# self.em.rclog("OCRPOSSIBLE: %d" % self.ocrpossible)
# Pdftk is optionally used to extract attachments. This takes
# a hit on perfmance even in the absence of any attachments,
# so it can be disabled in the configuration.
self.attextractdone = False
self.attachlist = []
if cf_attach:
self.pdftk = rclexecm.which("pdftk")
else:
self.pdftk = None
if self.pdftk:
self.maybemaketmpdir()
def __init__(self, logger, infile=sys.stdin):
self.log = logger
config = rclconfig.RclConfig()
dir1 = os.path.join(config.getConfDir(), "examples")
dir2 = os.path.join(config.datadir, "examples")
self.mimemap = conftree.ConfStack('mimemap', [dir1, dir2])
self.infile = infile
self.fields = {}
self.msg = EmailBuilder(self.log, self.mimemap)
def __init__(self, em):
self.currentindex = 0
self.pdftotext = None
self.pdfinfo = None
self.pdftk = None
self.em = em
self.tesseract = None
# Avoid picking up a default version on Windows, we want ours
if not _mswindows:
self.pdftotext = rclexecm.which("pdftotext")
if not self.pdftotext:
self.pdftotext = rclexecm.which("poppler/pdftotext")
if not self.pdftotext:
# No need for anything else. openfile() will return an
# error at once
return
self.config = rclconfig.RclConfig()
self.confdir = self.config.getConfDir()
# The user can set a list of meta tags to be extracted from
# the XMP metadata packet. These are specified as
# (xmltag,rcltag) pairs
self.extrameta = self.config.getConfParam("pdfextrameta")
if self.extrameta:
self.extrametafix = self.config.getConfParam("pdfextrametafix")
self._initextrameta()
# Check if we need to escape portions of text where old
# versions of pdftotext output raw HTML special characters.
self.needescape = True
try:
version = subprocess.check_output([self.pdftotext, "-v"],
stderr=subprocess.STDOUT)
major,minor,rev = version.split()[2].split('.')
# Don't know exactly when this changed but it's fixed in
# jessie 0.26.5
if int(major) > 0 or int(minor) >= 26:
self.needescape = False
except:
pass
# Pdftk is optionally used to extract attachments. This takes
# a hit on performance even in the absence of any attachments,
# so it can be disabled in the configuration.
self.attextractdone = False
self.attachlist = []
cf_attach = self.config.getConfParam("pdfattach")
cf_attach = rclexecm.configparamtrue(cf_attach)
if cf_attach:
self.pdftk = rclexecm.which("pdftk")
if self.pdftk:
self.maybemaketmpdir()
PY3 = sys.version > '3'
if PY3:
def makebytes(data):
if isinstance(data, bytes):
return data
else:
return data.encode("UTF-8")
else:
def makebytes(data):
if isinstance(data, unicode):
return data.encode("UTF-8")
else:
return data
my_config = rclconfig.RclConfig()
############################################
# RclExecM implements the
# communication protocol with the recollindex process. It calls the
# object specific of the document type to actually get the data.
class RclExecM:
noteof = 0
eofnext = 1
eofnow = 2
noerror = 0
subdocerror = 1
fileerror = 2
def __init__(self):