2024-11-06 23:34:51 -05:00
|
|
|
import re
|
2024-11-07 11:39:19 -05:00
|
|
|
import enum
|
2024-11-08 11:17:59 -05:00
|
|
|
import datetime
|
2024-11-07 17:10:04 -05:00
|
|
|
|
|
|
|
from dateparser.search import search_dates
|
2024-11-07 17:30:13 -05:00
|
|
|
from email.header import decode_header
|
2024-11-06 23:34:51 -05:00
|
|
|
|
|
|
|
class MBoxReaderError(Exception):
|
|
|
|
pass
|
|
|
|
|
|
|
|
class MBoxReaderBuffer():
|
|
|
|
def __init__(self):
|
|
|
|
self.lines = [None, None, None, None]
|
|
|
|
self.index = None
|
|
|
|
|
|
|
|
def add(self, line: str):
|
|
|
|
if self.lines[0] is None:
|
|
|
|
self.lines[0] = line
|
|
|
|
self.index = 0
|
|
|
|
elif self.lines[1] is None:
|
|
|
|
self.lines[1] = line
|
|
|
|
self.index = 1
|
|
|
|
elif self.lines[2] is None:
|
|
|
|
self.lines[2] = line
|
|
|
|
self.index = 2
|
|
|
|
elif self.lines[3] is None:
|
|
|
|
self.lines[3] = line
|
|
|
|
self.index = 3
|
|
|
|
else:
|
|
|
|
self.lines[0] = self.lines[1]
|
|
|
|
self.lines[1] = self.lines[2]
|
|
|
|
self.lines[2] = self.lines[3]
|
|
|
|
self.lines[3] = line
|
|
|
|
self.index = 3
|
|
|
|
|
|
|
|
if self.index is None:
|
|
|
|
self.index = 0
|
|
|
|
elif self.index < 3:
|
|
|
|
self.index += 1
|
|
|
|
|
|
|
|
def is_empty_line(self, line):
|
|
|
|
return self.lines[line] == '\n'
|
|
|
|
|
|
|
|
def is_from_line(self, line):
|
|
|
|
return self.lines[line][0:5] == 'From '
|
|
|
|
|
|
|
|
def is_header_line(self, line):
|
2024-11-08 11:17:59 -05:00
|
|
|
return re.match('^([A-Za-z0-9\\-]+): (.*)$', self.lines[line]) is not None
|
2024-11-06 23:34:51 -05:00
|
|
|
|
|
|
|
def is_start(self):
|
|
|
|
if self.lines[0] is None or self.lines[1] is None:
|
|
|
|
return
|
|
|
|
|
|
|
|
if self.is_from_line(0) and self.is_header_line(1):
|
|
|
|
pass
|
|
|
|
elif self.is_empty_line(0) \
|
|
|
|
and self.is_empty_line(1) \
|
|
|
|
and self.is_from_line(2) \
|
|
|
|
and self.is_header_line(3):
|
|
|
|
return 3
|
|
|
|
|
2024-11-07 11:39:19 -05:00
|
|
|
class MBoxMessageState(enum.Enum):
|
|
|
|
EMPTY = 0
|
|
|
|
HEADER = 1
|
|
|
|
BODY = 2
|
|
|
|
|
2024-11-07 17:30:13 -05:00
|
|
|
def decode(text: str):
|
|
|
|
decoded = decode_header(text)[0]
|
|
|
|
|
|
|
|
if decoded[1] is None:
|
2024-11-08 11:17:59 -05:00
|
|
|
return str(decoded[0])
|
2024-11-07 17:30:13 -05:00
|
|
|
|
2024-11-08 11:17:59 -05:00
|
|
|
try:
|
|
|
|
return str(decoded[0], decoded[1])
|
|
|
|
except:
|
|
|
|
return str(decoded[0])
|
2024-11-07 17:30:13 -05:00
|
|
|
|
2024-11-06 23:34:51 -05:00
|
|
|
class MBoxMessage():
|
2024-11-07 16:15:54 -05:00
|
|
|
__slots__ = 'state', 'headers', 'line', 'content', 'body', 'key',
|
2024-11-06 23:34:51 -05:00
|
|
|
|
|
|
|
def __init__(self):
|
2024-11-07 11:39:19 -05:00
|
|
|
self.state = MBoxMessageState.EMPTY
|
2024-11-06 23:34:51 -05:00
|
|
|
self.headers = dict()
|
2024-11-07 11:39:19 -05:00
|
|
|
self.line = None
|
2024-11-07 16:15:54 -05:00
|
|
|
self.content = ''
|
|
|
|
self.body = None
|
2024-11-06 23:34:51 -05:00
|
|
|
self.key = None
|
|
|
|
|
|
|
|
def add(self, line: str):
|
2024-11-07 16:15:54 -05:00
|
|
|
if self.line is not None:
|
|
|
|
self.content += self.line
|
|
|
|
|
2024-11-07 11:39:19 -05:00
|
|
|
if self.state is MBoxMessageState.EMPTY:
|
|
|
|
self.state = MBoxMessageState.HEADER
|
|
|
|
|
|
|
|
if self.state is MBoxMessageState.HEADER:
|
|
|
|
if line == '\n' or line == '\r\n':
|
|
|
|
self.state = MBoxMessageState.BODY
|
2024-11-06 23:34:51 -05:00
|
|
|
elif line[0] == ' ' or line[0] == '\t':
|
2024-11-07 17:30:13 -05:00
|
|
|
self.headers[self.key] += ' ' + decode(line.strip())
|
2024-11-06 23:34:51 -05:00
|
|
|
else:
|
|
|
|
match = re.match('^([^:]+): (.*)$', line)
|
|
|
|
|
|
|
|
if match:
|
2024-11-07 16:36:57 -05:00
|
|
|
self.key = match[1].lower()
|
2024-11-06 23:34:51 -05:00
|
|
|
|
2024-11-07 17:30:13 -05:00
|
|
|
self.headers[self.key] = decode(match[2].rstrip())
|
2024-11-07 11:39:19 -05:00
|
|
|
elif self.state is MBoxMessageState.BODY:
|
2024-11-07 16:15:54 -05:00
|
|
|
if self.body is None:
|
|
|
|
self.body = ''
|
2024-11-07 11:39:19 -05:00
|
|
|
else:
|
|
|
|
self.body += self.line
|
2024-11-06 23:34:51 -05:00
|
|
|
|
2024-11-07 16:15:54 -05:00
|
|
|
self.line = line
|
2024-11-06 23:34:51 -05:00
|
|
|
|
2024-11-07 16:36:57 -05:00
|
|
|
def header(self, key: str):
|
2024-11-07 17:10:04 -05:00
|
|
|
return self.headers.get(key.lower())
|
2024-11-07 16:36:57 -05:00
|
|
|
|
|
|
|
def date(self):
|
2024-11-08 11:17:59 -05:00
|
|
|
try:
|
|
|
|
return search_dates(self.headers['date'])[0][1]
|
|
|
|
except:
|
|
|
|
return datetime.datetime.fromtimestamp(0)
|
|
|
|
|
|
|
|
def sender(self):
|
|
|
|
return self.headers.get('from', 'Unknown')
|
|
|
|
|
|
|
|
def subject(self):
|
|
|
|
return self.headers.get('subject', '(no subject)')
|
2024-11-07 16:36:57 -05:00
|
|
|
|
2024-11-06 23:34:51 -05:00
|
|
|
def is_first_line(self):
|
2024-11-07 16:15:54 -05:00
|
|
|
return len(self.headers) == 1 and (self.body == '' or self.body is None)
|
2024-11-06 23:34:51 -05:00
|
|
|
|
2024-11-07 15:44:20 -05:00
|
|
|
@staticmethod
|
|
|
|
def each_line(text: str):
|
|
|
|
start = 0
|
|
|
|
end = len(text)
|
|
|
|
|
|
|
|
while True:
|
|
|
|
try:
|
|
|
|
index = text.index('\n', start, end)
|
|
|
|
|
|
|
|
yield text[start:index+1]
|
|
|
|
|
|
|
|
start = index + 1
|
|
|
|
|
|
|
|
if start == end:
|
|
|
|
break
|
|
|
|
except ValueError:
|
|
|
|
yield text[start:end]
|
|
|
|
break
|
|
|
|
|
|
|
|
@staticmethod
|
|
|
|
def parse(text: str):
|
|
|
|
message = MBoxMessage()
|
|
|
|
|
|
|
|
for line in MBoxMessage.each_line(text):
|
|
|
|
message.add(line)
|
|
|
|
|
|
|
|
return message
|
|
|
|
|
2024-11-06 23:34:51 -05:00
|
|
|
class MBoxReader():
|
|
|
|
__slots__ = 'path', 'fh', 'line', 'buf', 'message',
|
|
|
|
|
|
|
|
def __init__(self, path: str):
|
|
|
|
self.path = path
|
|
|
|
self.fh = open(path, 'r', newline='')
|
|
|
|
self.line = 0
|
|
|
|
self.buf = MBoxReaderBuffer()
|
|
|
|
self.message = None
|
|
|
|
|
|
|
|
def get_message(self):
|
|
|
|
while True:
|
|
|
|
line = self.fh.readline()
|
|
|
|
|
2024-11-07 11:39:19 -05:00
|
|
|
if line is None or line == '':
|
2024-11-06 23:34:51 -05:00
|
|
|
ret = self.message
|
|
|
|
|
|
|
|
self.message = None
|
|
|
|
|
|
|
|
return ret
|
|
|
|
|
|
|
|
self.line += 1
|
|
|
|
|
|
|
|
self.buf.add(line)
|
|
|
|
|
|
|
|
if self.buf.is_start():
|
|
|
|
if self.message is None:
|
|
|
|
self.message = MBoxMessage()
|
|
|
|
else:
|
|
|
|
ret = self.message
|
|
|
|
|
|
|
|
self.message = MBoxMessage()
|
|
|
|
self.message.add(line)
|
|
|
|
|
|
|
|
return ret
|
|
|
|
|
|
|
|
if self.message:
|
|
|
|
self.message.add(line)
|
|
|
|
|
|
|
|
def messages(self):
|
|
|
|
while True:
|
|
|
|
message = self.get_message()
|
|
|
|
|
|
|
|
if message is None:
|
|
|
|
break
|
|
|
|
|
|
|
|
yield message
|