Add option to ignore missing message ID

master
Sven Slootweg 12 years ago
parent 132dfc9b7e
commit 633abafb7b

203
parse

@ -15,6 +15,9 @@ parser.add_argument('-a', dest='attachment_dir', action='store', default='attach
parser.add_argument('-f', '--forced', dest='forced', action='store_true', parser.add_argument('-f', '--forced', dest='forced', action='store_true',
help='force insertion into database, even if entries already exist') help='force insertion into database, even if entries already exist')
parser.add_argument('-I', '--ignore-invalid', dest='ignore_invalid', action='store_true',
help='process invalid e-mail files anyway, for example when missing message-id headers')
args = parser.parse_args() args = parser.parse_args()
options = vars(args) options = vars(args)
@ -100,112 +103,120 @@ for email_file in file_list:
message = email.message_from_file(open(email_file, 'r')) message = email.message_from_file(open(email_file, 'r'))
if message['message-id'] is None: if message['message-id'] is None:
print "%s is not a valid e-mail file." % email_file if options['ignore_invalid'] == True:
else: message_id = ""
if 'subject' not in message or message['subject'] is None: print "WARNING: %s does not contain a valid message-id header. Empty message-id assumed." % email_file
subject = ""
else: else:
subject = message['subject'] print "%s is not a valid e-mail file." % email_file
finished += 1
textbody = "" continue
htmlbody = "" else:
attachment_list = [] message_id = message['message-id']
sha1_hash = hashlib.sha1("%s/%s/%s/%s" % (message['from'], message['to'], message['message-id'], subject)).hexdigest()
if options['forced'] == False: if 'subject' not in message or message['subject'] is None:
cursor.execute("SELECT * FROM emails WHERE `Hash` = ?", (sha1_hash,)) subject = ""
if len(cursor.fetchall()) > 0: else:
print "Skipping %s, already exists in the database according to message hash." % sha1_hash subject = message['subject']
continue
textbody = ""
message_parts = [find_submessages(message)] htmlbody = ""
message_parts = flatten(message_parts) attachment_list = []
sha1_hash = hashlib.sha1("%s/%s/%s/%s" % (message['from'], message['to'], message_id, subject)).hexdigest()
if options['forced'] == False:
cursor.execute("SELECT * FROM emails WHERE `Hash` = ?", (sha1_hash,))
if len(cursor.fetchall()) > 0:
print "Skipping %s, already exists in the database according to message hash." % sha1_hash
continue
for part in message_parts: message_parts = [find_submessages(message)]
if part.get_filename() is None: message_parts = flatten(message_parts)
# Part of the message
if part.get_content_type() == "text/plain": for part in message_parts:
if textbody == "": if part.get_filename() is None:
# Part of the message
if part.get_content_type() == "text/plain":
if textbody == "":
try:
textbody = part.get_payload(decode=True).decode(get_charset(part))
except UnicodeDecodeError:
# This part is probably in windows-1252 encoding
try: try:
textbody = part.get_payload(decode=True).decode(get_charset(part)) textbody = part.get_payload(decode=True).decode('windows-1252')
except UnicodeDecodeError: except UnicodeDecodeError:
# This part is probably in windows-1252 encoding # Ok, we really have no clue how to decode this, we'll just skip it...
try: continue
textbody = part.get_payload(decode=True).decode('windows-1252') except LookupError:
except UnicodeDecodeError: pass
# Ok, we really have no clue how to decode this, we'll just skip it... elif part.get_content_type() == "text/html" or part.get_content_type == "text/xhtml+xml" or part.get_content_type == "application/xhtml+xml":
continue if htmlbody == "":
except LookupError: try:
pass htmlbody = part.get_payload(decode=True).decode(get_charset(part))
elif part.get_content_type() == "text/html" or part.get_content_type == "text/xhtml+xml" or part.get_content_type == "application/xhtml+xml": except UnicodeDecodeError:
if htmlbody == "": # This part is probably in windows-1252 encoding
try: try:
htmlbody = part.get_payload(decode=True).decode(get_charset(part)) htmlbody = part.get_payload(decode=True).decode('windows-1252')
except UnicodeDecodeError: except UnicodeDecodeError:
# This part is probably in windows-1252 encoding # Ok, we really have no clue how to decode this, we'll just skip it...
try: continue
htmlbody = part.get_payload(decode=True).decode('windows-1252') except LookupError:
except UnicodeDecodeError: pass
# Ok, we really have no clue how to decode this, we'll just skip it...
continue
except LookupError:
pass
else:
# Technically this is supposed to be part of the message body, but we have no idea what format it is in...
print "WARNING: Unknown message format encountered in %s, message may be incomplete." % sha1_hash
else: else:
# Attachment # Technically this is supposed to be part of the message body, but we have no idea what format it is in...
attachment_data = part.get_payload(decode=True) print "WARNING: Unknown message format encountered in %s, message may be incomplete." % sha1_hash
attachment_sha1 = hashlib.sha1(attachment_data).hexdigest() else:
attachment_filename = part.get_filename() # Attachment
attachment_type = part.get_content_type() attachment_data = part.get_payload(decode=True)
attachment_extension = os.path.splitext(attachment_filename)[1] attachment_sha1 = hashlib.sha1(attachment_data).hexdigest()
attachment_size = len(attachment_data) attachment_filename = part.get_filename()
attachment_destination = "%s/%s%s" % (options['attachment_dir'], attachment_sha1, attachment_extension) attachment_type = part.get_content_type()
attachment_list.append((attachment_filename, attachment_type, attachment_sha1, attachment_size)) attachment_extension = os.path.splitext(attachment_filename)[1]
attachment_size = len(attachment_data)
attachment_file = open(attachment_destination, "w") attachment_destination = "%s/%s%s" % (options['attachment_dir'], attachment_sha1, attachment_extension)
attachment_file.write(attachment_data) attachment_list.append((attachment_filename, attachment_type, attachment_sha1, attachment_size))
attachment_file.close()
try:
timestamp = int(time.mktime(email.utils.parsedate(message['date'])))
except TypeError:
timestamp = 0
print "WARNING: Failed to determine unix timestamp for %s." % sha1_hash
try:
new_row = (getheader(message['message-id']), getheader(message['from']), getheader(message['to']), getheader(subject), timestamp, textbody, htmlbody, sha1_hash)
except UnicodeDecodeError:
print "ERROR: Failed parsing %s, headers could not be decoded." % sha1_hash
continue
except email.errors.HeaderParseError:
print "ERROR: Failed parsing %s, headers could not be parsed." % sha1_hash
continue
cursor.execute("INSERT INTO emails VALUES (?, ?, ?, ?, ?, ?, ?, ?)", new_row)
print "Successfully parsed and inserted e-mail with SHA1 hash %s." % sha1_hash
if len(attachment_list) > 0:
inserted = 0
for attachment in attachment_list: attachment_file = open(attachment_destination, "w")
if options['forced'] == False: attachment_file.write(attachment_data)
cursor.execute("SELECT * FROM attachments WHERE `hash` = ? AND `message_hash` = ?", (attachment[2], sha1_hash)) attachment_file.close()
if len(cursor.fetchall()) > 0:
print "Skipping attachment %s, already exists in the database." % attachment[2] try:
continue timestamp = int(time.mktime(email.utils.parsedate(message['date'])))
except TypeError:
try: timestamp = 0
new_row = (sha1_hash, attachment[0], attachment[1], attachment[2], attachment[3]) print "WARNING: Failed to determine unix timestamp for %s." % sha1_hash
cursor.execute("INSERT INTO attachments VALUES (?, ?, ?, ?, ?)", new_row)
inserted += 1 try:
except sqlite3.ProgrammingError: new_row = (getheader(message_id), getheader(message['from']), getheader(message['to']), getheader(subject), timestamp, textbody, htmlbody, sha1_hash)
print "Inserting of attachment %s failed." % attachment[2] except UnicodeDecodeError:
print "ERROR: Failed parsing %s, headers could not be decoded." % sha1_hash
continue
except email.errors.HeaderParseError:
print "ERROR: Failed parsing %s, headers could not be parsed." % sha1_hash
continue
cursor.execute("INSERT INTO emails VALUES (?, ?, ?, ?, ?, ?, ?, ?)", new_row)
print "Successfully parsed and inserted e-mail with SHA1 hash %s." % sha1_hash
if len(attachment_list) > 0:
inserted = 0
for attachment in attachment_list:
if options['forced'] == False:
cursor.execute("SELECT * FROM attachments WHERE `hash` = ? AND `message_hash` = ?", (attachment[2], sha1_hash))
if len(cursor.fetchall()) > 0:
print "Skipping attachment %s, already exists in the database." % attachment[2]
continue
if inserted > 0: try:
print "Successfully inserted %d attachment(s) for %s." % (inserted, sha1_hash) new_row = (sha1_hash, attachment[0], attachment[1], attachment[2], attachment[3])
cursor.execute("INSERT INTO attachments VALUES (?, ?, ?, ?, ?)", new_row)
inserted += 1
except sqlite3.ProgrammingError:
print "Inserting of attachment %s failed." % attachment[2]
if inserted > 0:
print "Successfully inserted %d attachment(s) for %s." % (inserted, sha1_hash)
finished += 1 finished += 1

@ -7,6 +7,9 @@ parser = argparse.ArgumentParser(description='Parses emails into an SQLite datab
parser.add_argument('-p', dest='pattern', action='store', default='*', parser.add_argument('-p', dest='pattern', action='store', default='*',
help='glob pattern (including path) that has to be matched for a file to be parsed') help='glob pattern (including path) that has to be matched for a file to be parsed')
parser.add_argument('-I', '--ignore-invalid', dest='ignore_invalid', action='store_true',
help='process invalid e-mail files anyway, for example when missing message-id headers')
args = parser.parse_args() args = parser.parse_args()
options = vars(args) options = vars(args)
@ -60,21 +63,29 @@ for email_file in file_list:
# There is no need to check for the format here, since if the filename is not a valid hash, it simply won't match anything. # There is no need to check for the format here, since if the filename is not a valid hash, it simply won't match anything.
message = email.message_from_file(open(email_file, 'r')) message = email.message_from_file(open(email_file, 'r'))
if message['message-id'] is None: if message['message-id'] is None:
print "%s is not a valid e-mail file." % email_file if options['ignore_invalid'] == True:
else: message_id = ""
if 'subject' not in message or message['subject'] is None: print "WARNING: %s does not contain a valid message-id header. Empty message-id assumed." % email_file
subject = ""
else: else:
subject = message['subject'] print "%s is not a valid e-mail file." % email_file
finished += 1
sha1_hash = hashlib.sha1("%s/%s/%s/%s" % (message['from'], message['to'], message['message-id'], subject)).hexdigest() continue
else:
new_path = "%s/%s.eml" % (os.path.dirname(email_file), sha1_hash) message_id = message['message-id']
os.rename(email_file, new_path) if 'subject' not in message or message['subject'] is None:
print "%s -> %s" % (email_file, new_path) subject = ""
else:
subject = message['subject']
sha1_hash = hashlib.sha1("%s/%s/%s/%s" % (message['from'], message['to'], message_id, subject)).hexdigest()
new_path = "%s/%s.eml" % (os.path.dirname(email_file), sha1_hash)
os.rename(email_file, new_path)
print "%s -> %s" % (email_file, new_path)
finished += 1 finished += 1

Loading…
Cancel
Save