performance improvement for large emails

[edward.git] / edward

#! /usr/bin/env python3
# -*- coding: utf-8 -*-
"""*********************************************************************
* Edward is free software: you can redistribute it and/or modify       *
* it under the terms of the GNU Affero Public License as published by  *
* the Free Software Foundation, either version 3 of the License, or    *
* (at your option) any later version.                                  *
*                                                                      *
* Edward is distributed in the hope that it will be useful,            *
* but WITHOUT ANY WARRANTY; without even the implied warranty of       *
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the        *
* GNU Affero Public License for more details.                          *
*                                                                      *
* You should have received a copy of the GNU Affero Public License     *
* along with Edward.  If not, see <http://www.gnu.org/licenses/>.      *
*                                                                      *
* Copyright (C) 2014-2015 Andrew Engelbrecht                (AGPLv3+)  *
* Copyright (C) 2014      Josh Drake                        (AGPLv3+)  *
* Copyright (C) 2014      Lisa Marie Maginnis               (AGPLv3+)  *
* Copyright (C) 2009-2015 Tails developers <tails@boum.org> ( GPLv3+)  *
* Copyright (C) 2009      W. Trevor King <wking@drexel.edu> ( GPLv2+)  *
*                                                                      *
* Special thanks to Josh Drake for writing the original edward bot! :) *
*                                                                      *
************************************************************************

Code sourced from these projects:

  * http://agpl.fsf.org/emailselfdefense.fsf.org/edward/PREVIOUS-20150530/edward.tar.gz
  * https://git-tails.immerda.ch/whisperback/tree/whisperBack/encryption.py?h=feature/python3
  * http://www.physics.drexel.edu/~wking/code/python/send_pgp_mime
"""

import re
import io
import os
import sys
import enum
import gpgme
import smtplib
import importlib

import email.parser
import email.message
import email.encoders

from email.mime.text            import MIMEText
from email.mime.multipart       import MIMEMultipart
from email.mime.application     import MIMEApplication
from email.mime.nonmultipart    import MIMENonMultipart

import edward_config

langs = ["de", "el", "en", "es", "fr", "it", "ja", "pt-br", "ro", "ru", "tr"]

"""This list contains the abbreviated names of reply languages available to
edward."""

class TxtType (enum.Enum):
    text        = 0
    message     = 1
    pubkey      = 2
    detachedsig = 3
    signature   = 4


match_pairs =  [(TxtType.message,
                '-----BEGIN PGP MESSAGE-----.*?-----END PGP MESSAGE-----'),
                (TxtType.pubkey,
                '-----BEGIN PGP PUBLIC KEY BLOCK-----.*?-----END PGP PUBLIC KEY BLOCK-----'),
                (TxtType.detachedsig,
                '-----BEGIN PGP SIGNATURE-----.*?-----END PGP SIGNATURE-----')]

"""This list of tuples matches query names with re.search() queries used
to find GPG data for edward to process."""


class EddyMsg (object):
    """
    The EddyMsg class represents relevant parts of a mime message.

    The represented message can be single-part or multi-part.

    'multipart' is set to True if there are multiple mime parts.

    'subparts' points to a list of mime sub-parts if it is a multi-part
    message. Otherwise it points to an empty list.

    'payload_bytes' is a binary representation of the mime part before header
    removal and message decoding.

    'payload_pieces' is a list of objects containing strings that when strung
    together form the fully-decoded string representation of the mime part.

    The 'filename', 'content_type', 'content_disposition' and
    'description_list' come from the mime part parameters.

    """

    multipart               = False
    subparts                = []

    payload_bytes           = None
    payload_pieces          = []

    filename                = None
    content_type            = None
    content_disposition     = None
    description_list        = None


class PayloadPiece (object):
    """
    PayloadPiece represents a complte or sub-section of a mime part.

    Instances of this class are often strung together within one or more arrays
    pointed to by each instance of the EddyMsg class.

    'piece_type' refers to an enum whose value describes the content of
    'string'.  Examples include TxtType.pubkey, for public keys, and
    TxtType.message, for encrypted data (or armored signatures until they are
    known to be such.) Some of the names derive from the header and footer of
    each of these ascii-encoded gpg blocks.

    'string' contains some string of text, such as non-GPG text, an encrypted
    block of text, a signature, or a public key.

    'gpg_data' points to any instances of GPGData that have been created based
    on the contents of 'string'.
    """

    piece_type              = None
    string                  = None
    gpg_data                = None


class GPGData (object):
    """
    GPGData holds info from decryption, sig. verification, and/or pub. keys.

    Instances of this class contain decrypted information, signature
    fingerprints and/or fingerprints of processed and imported public keys.

    'decrypted' is set to True if 'plainobj' was created from encrypted data.

    'plainobj' points to any decrypted, or signed part of, a GPG signature. It
    is intended to be an instance of the EddyMsg class.

    'sigs' is a list of fingerprints of keys used to sign the data in plainobj.

    'sigkey_missing' is set to True if edward doesn't have the key it needs to
    verify the signature on a block of text.

    'key_cannot_encrypt' is set to True if pubkeys or sigs' keys in the payload
    piece are not capable of encryption, are revoked or expired, for instance.

    'keys' is a list of fingerprints of keys obtained in public key blocks.
    """

    decrypted               = False

    plainobj                = None
    sigs                    = []
    sigkey_missing          = False
    key_cannot_encrypt      = False
    keys                    = []


class ReplyInfo (object):
    """
    ReplyInfo contains details that edward uses in generating its reply.

    Instances of this class contain information about whether a message was
    successfully encrypted or signed, and whether a public key was attached, or
    retrievable, from the local GPG store. It stores the fingerprints of
    potential encryption key candidates and the message (if any at all) to
    quote in edward's reply.

    'replies' points one of the dictionaries of translated replies.

    'target_key' refers to the fingerprint of a key used to sign encrypted
    data. This is the preferred key, if it is set, and if is available.

    'fallback_target_key' referst to the fingerprint of a key used to sign
    unencrypted data; alternatively it may be a public key attached to the
    message.

    'encrypt_to_key' the key object to use when encrypting edward's reply

    'msg_to_quote' refers to the part of a message which edward should quote in
    his reply. This should remain as None if there was no encrypted and singed
    part. This is to avoid making edward a service for decrypting other
    people's messages to edward.

    'decrypt_success' is set to True if edward could decrypt part of the
    message.

    'sig_success' is set to True if edward could to some extent verify the
    signature of a signed part of the message to edward.

    'key_can_encrypt' is set to True if a key which can be encrypted to has
    been found.

    'sig_failure' is set to True if edward could not verify a siganture.

    'pubkey_success' is set to True if edward successfully imported a public
    key.

    'sigkey_missing' is set to True if edward doesn't have the public key
    needed for signature verification.

    'key_cannot_encrypt' is set to True if pubkeys or sig's keys in a payload
    piece of the message are not capable of encryption.

    'have_repy_key' is set to True if edward has a public key to encrypt its
    reply to.
    """

    replies                 = None

    target_key              = None
    fallback_target_key     = None
    encrypt_to_key          = None
    msg_to_quote            = ""

    decrypt_success         = False
    sig_success             = False
    pubkey_success          = False
    key_can_encrypt         = False

    decrypt_failure         = False
    sig_failure             = False
    sigkey_missing          = False
    key_cannot_encrypt      = False

    have_reply_key          = False


def main ():

    """
    This is the main function for edward, a GPG reply bot.

    Edward responds to GPG-encrypted and signed mail, encrypting and signing
    the response if the user's public key is, or was, included in the message.

    Args:
        None

    Returns:
        Nothing

    Pre:
        Mime or plaintext email passing in through standard input.  Portions of
        the email may be encrypted. If the To: address contains the text
        "edward-ja", then the reply will contain a reply written in the
        Japanese language.  There are other languages as well. The default
        language is English.

    Post:
        A reply email will be printed to standard output. The contents of the
        reply email depends on whether the original email was encrypted or not,
        has or doesn't have a signature, whether a public key used in the
        original message is provided or locally stored, and the language
        implied by the To: address in the original email.
    """

    print_reply_only = handle_args()

    gpgme_ctx = get_gpg_context(edward_config.gnupghome,
                              edward_config.sign_with_key)

    email_bytes = sys.stdin.buffer.read()
    email_struct = parse_pgp_mime(email_bytes, gpgme_ctx)

    email_to, email_reply_to, email_subject = email_to_reply_to_subject(email_bytes)
    lang, reply_from = import_lang_pick_address(email_to, edward_config.hostname)

    replyinfo_obj = ReplyInfo()
    replyinfo_obj.replies = lang.replies

    prepare_for_reply(email_struct, replyinfo_obj)
    get_key_from_fp(replyinfo_obj, gpgme_ctx)
    reply_plaintext = write_reply(replyinfo_obj)

    reply_mime = generate_encrypted_mime(reply_plaintext, email_reply_to, \
                                         email_subject, replyinfo_obj.encrypt_to_key,
                                         gpgme_ctx)

    if print_reply_only == True:
        print(reply_mime)
    else:
        send_reply(reply_mime, email_reply_to, reply_from)


def get_gpg_context (gnupghome, sign_with_key_fp):
    """
    This function returns the GPG context needed for encryption and signing.

    The context is needed by other functions which use GPG functionality.

    Args:
        gnupghome: The path to "~/.gnupg/" or its alternative.
        sign_with_key: The fingerprint of the key to sign with

    Returns:
        A gpgme context to be used for GPG functions.

    Post:
        the 'armor' flag is set to True and the list of signing keys contains
        the single specified key
    """

    os.environ['GNUPGHOME'] = gnupghome

    gpgme_ctx = gpgme.Context()
    gpgme_ctx.armor = True

    try:
        sign_with_key = gpgme_ctx.get_key(sign_with_key_fp)
    except gpgme.GpgmeError:
        error("unable to load signing key. is the gnupghome "
                + "and signing key properly set in the edward_config.py?")
        exit(1)

    gpgme_ctx.signers = [sign_with_key]

    return gpgme_ctx


def parse_pgp_mime (email_bytes, gpgme_ctx):
    """Parses the email for mime payloads and decrypts/verfies signatures.

    This function creates a representation of a mime or plaintext email with
    the EddyMsg class. It then splits each mime payload into one or more pieces
    which may be plain text or GPG data. It then decrypts encrypted parts and
    does some very basic signature verification on those parts.

    Args:
        email_bytes: an email message in byte string format
        gpgme_ctx:  a gpgme context

    Returns:
        A message as an instance of EddyMsg

    Post:
        the returned EddyMsg instance has split, decrypted, verified and pubkey
        imported payloads
    """

    email_struct = email.parser.BytesParser().parsebytes(email_bytes)

    eddymsg_obj = parse_mime(email_struct)
    split_payloads(eddymsg_obj)
    gpg_on_payloads(eddymsg_obj, gpgme_ctx)

    return eddymsg_obj


def parse_mime(msg_struct):
    """Translates python's email.parser format into an EddyMsg format

    If the message is multi-part, then a recursive object is created, where
    each sub-part is also a EddyMsg instance.

    Args:
        msg_struct: an email parsed with email.parser.BytesParser(), which can be
            multi-part

    Returns:
        an instance of EddyMsg, potentially a recursive one.
    """

    eddymsg_obj = get_subpart_data(msg_struct)

    if msg_struct.is_multipart() == True:
        payloads = msg_struct.get_payload()

        eddymsg_obj.multipart = True
        eddymsg_obj.subparts = list(map(parse_mime, payloads))

    return eddymsg_obj


def scan_and_split (payload_piece, match_name, pattern):
    """This splits the payloads of an EddyMsg object into GPG and text parts.

    An EddyMsg object's payload_pieces starts off as a list containing a single
    PayloadPiece object. This function returns a list of these objects which
    have been split into GPG data and regular text, if such splits need to be/
    can be made.

    Args:
        payload_piece: a single payload or a split part of a payload
        match_name: the type of data to try to spit out from the payload piece
        pattern: the search pattern to be used for finding that type of data

    Returns:
        a list of objects of the PayloadPiece class, in the order that the
        string part of payload_piece originally was, broken up according to
        matches specified by 'pattern'.
    """

    # don't try to re-split pieces containing gpg data
    if payload_piece.piece_type != TxtType.text:
        return [payload_piece]

    flags = re.DOTALL | re.MULTILINE
    matches = re.search(pattern, payload_piece.string, flags=flags)

    if matches == None:
        pieces = [payload_piece]

    else:

        beginning               = PayloadPiece()
        beginning.string        = payload_piece.string[:matches.start()]
        beginning.piece_type    = payload_piece.piece_type

        match                   = PayloadPiece()
        match.string            = payload_piece.string[matches.start():matches.end()]
        match.piece_type        = match_name

        rest                    = PayloadPiece()
        rest.string             = payload_piece.string[matches.end():]
        rest.piece_type         = payload_piece.piece_type

        more_pieces = scan_and_split(rest, match_name, pattern)
        pieces = [beginning, match ] + more_pieces

    return pieces


def get_subpart_data (part):
    """This function grabs information from a mime part.

    It copies needed data from an email.parser.BytesParser() object over to an
    EddyMsg object.

    Args:
        part: an email.parser.BytesParser() object

    Returns:
        an EddyMsg() object
    """

    obj = EddyMsg()

    mime_decoded_bytes      = part.get_payload(decode=True)
    charset                 = part.get_content_charset()

    # your guess is as good as a-myy-ee-ine...
    if charset == None:
        charset = 'utf-8'

    payload_string          = part.as_string()
    if payload_string != None:
        # convert each isolated carriage return or line feed to carriage return + line feed
        payload_string_crlf = re.sub(r'\n', '\r\n', re.sub(r'\r', '\n', re.sub(r'\r\n', '\n', payload_string)))
        obj.payload_bytes   = payload_string_crlf.encode(charset)

    obj.filename            = part.get_filename()
    obj.content_type        = part.get_content_type()
    obj.content_disposition = part['content-disposition']
    obj.description_list    = part['content-description']

    if mime_decoded_bytes != None:
        try:
            payload = PayloadPiece()
            payload.string = mime_decoded_bytes.decode(charset)
            payload.piece_type = TxtType.text

            obj.payload_pieces = [payload]
        except UnicodeDecodeError:
            pass

    return obj


def do_to_eddys_pieces (function_to_do, eddymsg_obj, data):
    """A function which maps another function onto a message's subparts.

    This is a higer-order function which recursively performs a specified
    function on each subpart of a multi-part message. Each single-part sub-part
    has the function applied to it. This function also works if the part passed
    in is single-part.

    Args:
        function_to_do: function to perform on sub-parts
        eddymsg_obj: a single part or multi-part EddyMsg object
        data: a second argument to pass to 'function_to_do'

    Returns:
        Nothing

    Post:
        The passed-in EddyMsg object is transformed recursively on its
        sub-parts according to 'function_to_do'.
    """

    if eddymsg_obj.multipart == True:
        for sub in eddymsg_obj.subparts:
            do_to_eddys_pieces(function_to_do, sub, data)
    else:
        function_to_do(eddymsg_obj, data)


def split_payloads (eddymsg_obj):
    """Splits all (sub-)payloads of a message into GPG data and regular text.

    Recursively performs payload splitting on all sub-parts of an EddyMsg
    object into the various GPG data types, such as GPG messages, public key
    blocks and signed text.

    Args:
        eddymsg_obj: an instance of EddyMsg

    Returns:
        Nothing

    Pre:
        The EddyMsg object has payloads that are unsplit (by may be split)..

    Post:
        The EddyMsg object's payloads are all split into GPG and non-GPG parts.
    """

    for match_pair in match_pairs:
        do_to_eddys_pieces(split_payload_pieces, eddymsg_obj, match_pair)


def split_payload_pieces (eddymsg_obj, match_pair):
    """A helper function for split_payloads(); works on PayloadPiece objects.

    This function splits up PayloadPiece objects into multipe PayloadPiece
    objects and replaces the EddyMsg object's previous list of payload pieces
    with the new split up one.

    Args:
        eddymsg_obj: a single-part EddyMsg object.
        match_pair: a tuple from the match_pairs list, which specifies a match
            name and a match pattern.

    Returns:
        Nothing

    Pre:
        The payload piece(s) of an EddyMsg object may be already split or
        unsplit.

    Post:
        The EddyMsg object's payload piece(s) are split into a list of pieces
        if matches of the match_pair are found.
    """

    (match_name, pattern) = match_pair

    new_pieces_list = []
    for piece in eddymsg_obj.payload_pieces:
        new_pieces_list += scan_and_split(piece, match_name, pattern)

    eddymsg_obj.payload_pieces = new_pieces_list


def gpg_on_payloads (eddymsg_obj, gpgme_ctx, prev_parts=[]):
    """Performs GPG operations on the GPG parts of the message

    This function decrypts text, verifies signatures, and imports public keys
    included in an email.

    Args:
        eddymsg_obj: an EddyMsg object with its payload_pieces split into GPG
            and non-GPG sections by split_payloads()
        gpgme_ctx: a gpgme context

        prev_parts: a list of mime parts that occur before the eddymsg_obj
            part, under the same multi-part mime part. This is used for
            verifying detached signatures. For the root mime part, this should
            be an empty list, which is the default value if this paramater is
            omitted.

    Return:
        Nothing

    Pre:
        eddymsg_obj should have its payloads split into gpg and non-gpg pieces.

    Post:
        Decryption, verification and key imports occur. the gpg_data members of
        PayloadPiece objects get filled in with GPGData objects with some of
        their attributes set.
    """

    if eddymsg_obj.multipart == True:
        prev_parts=[]
        for sub in eddymsg_obj.subparts:
            gpg_on_payloads (sub, gpgme_ctx, prev_parts)
            prev_parts += [sub]

        return

    for piece in eddymsg_obj.payload_pieces:

        if piece.piece_type == TxtType.text:
            # don't transform the plaintext.
            pass

        elif piece.piece_type == TxtType.message:
            piece.gpg_data = GPGData()

            (plaintext_b, sigs, sigkey_missing, key_cannot_encrypt) = decrypt_block(piece.string, gpgme_ctx)

            piece.gpg_data.sigkey_missing = sigkey_missing
            piece.gpg_data.key_cannot_encrypt = key_cannot_encrypt

            if plaintext_b:
                piece.gpg_data.decrypted = True
                piece.gpg_data.sigs = sigs
                # recurse!
                piece.gpg_data.plainobj = parse_pgp_mime(plaintext_b, gpgme_ctx)
                continue

            # if not encrypted, check to see if this is an armored signature.
            (plaintext_b, sigs,  sigkey_missing, key_cannot_encrypt) = verify_sig_message(piece.string, gpgme_ctx)

            piece.gpg_data.sigkey_missing = sigkey_missing
            piece.gpg_data.key_cannot_encrypt = key_cannot_encrypt

            if plaintext_b:
                piece.piece_type = TxtType.signature
                piece.gpg_data.sigs = sigs
                # recurse!
                piece.gpg_data.plainobj = parse_pgp_mime(plaintext_b, gpgme_ctx)

        elif piece.piece_type == TxtType.pubkey:
            piece.gpg_data = GPGData()

            (key_fps, key_cannot_encrypt) = add_gpg_key(piece.string, gpgme_ctx)

            piece.gpg_data.key_cannot_encrypt = key_cannot_encrypt

            if key_fps != []:
                piece.gpg_data.keys = key_fps

        elif piece.piece_type == TxtType.detachedsig:
            piece.gpg_data = GPGData()

            for prev in prev_parts:
                (sig_fps, sigkey_missing, key_cannot_encrypt) = verify_detached_signature(piece.string, prev.payload_bytes, gpgme_ctx)

                piece.gpg_data.sigkey_missing = sigkey_missing
                piece.gpg_data.key_cannot_encrypt = key_cannot_encrypt

                if sig_fps != []:
                    piece.gpg_data.sigs = sig_fps
                    piece.gpg_data.plainobj = prev
                    break

        else:
            pass


def prepare_for_reply (eddymsg_obj, replyinfo_obj):
    """Updates replyinfo_obj with info on the message's GPG success/failures

    This function marks replyinfo_obj with information about whether encrypted
    text in eddymsg_obj was successfully decrypted, signatures were verified
    and whether a public key was found or not.

    Args:
        eddymsg_obj: a message in the EddyMsg format
        replyinfo_obj: an instance of ReplyInfo

    Returns:
        Nothing

    Pre:
        eddymsg_obj has had its gpg_data created by gpg_on_payloads

    Post:
        replyinfo_obj has been updated with info about decryption/sig
        verififcation status, etc. However the desired key isn't imported until
        later, so the success or failure of that updates the values set here.
    """

    do_to_eddys_pieces(prepare_for_reply_pieces, eddymsg_obj, replyinfo_obj)

def prepare_for_reply_pieces (eddymsg_obj, replyinfo_obj):
    """A helper function for prepare_for_reply

    It updates replyinfo_obj with GPG success/failure information, when
    supplied a single-part EddyMsg object.

    Args:
        eddymsg_obj: a single-part message in the EddyMsg format
        replyinfo_obj: an object which holds information about the message's
            GPG status

    Returns:
        Nothing

    Pre:
        eddymsg_obj is a single-part message. (it may be a part of a multi-part
        message.) It has had its gpg_data created by gpg_on_payloads if it has
        gpg data.

    Post:
        replyinfo_obj has been updated with gpg success/failure information
    """

    for piece in eddymsg_obj.payload_pieces:
        if piece.piece_type == TxtType.text:
            # don't quote the plaintext part.
            pass

        elif piece.piece_type == TxtType.message:
            prepare_for_reply_message(piece, replyinfo_obj)

        elif piece.piece_type == TxtType.pubkey:
            prepare_for_reply_pubkey(piece, replyinfo_obj)

        elif (piece.piece_type == TxtType.detachedsig) \
            or (piece.piece_type == TxtType.signature):
                    prepare_for_reply_sig(piece, replyinfo_obj)


def prepare_for_reply_message (piece, replyinfo_obj):
    """Helper function for prepare_for_reply()

    This function is called when the piece_type of a payload piece is
    TxtType.message, or GPG Message block. This should be encrypted text. If
    the encryted block is correclty signed, a sig will be attached to
    .target_key unless there is already one there.

    Args:
        piece: a PayloadPiece object.
        replyinfo_obj: object which gets updated with decryption status, etc.

    Returns:
        Nothing

    Pre:
        the piece.payload_piece value should be TxtType.message.

    Post:
        replyinfo_obj gets updated with decryption status, signing status, a
        potential signing key, posession status of the public key for the
        signature and encryption capability status if that key is missing.
    """

    if piece.gpg_data.plainobj == None:
        replyinfo_obj.decrypt_failure = True
        return

    replyinfo_obj.decrypt_success = True

    # we already have a key (and a message)
    if replyinfo_obj.target_key != None:
        return

    if piece.gpg_data.sigs == []:
        if piece.gpg_data.sigkey_missing == True:
            replyinfo_obj.sigkey_missing = True

        if piece.gpg_data.key_cannot_encrypt == True:
            replyinfo_obj.key_cannot_encrypt = True

        # only include a signed message in the reply.
        get_signed_part = True

    else:
        replyinfo_obj.target_key = piece.gpg_data.sigs[0]
        replyinfo_obj.sig_success = True
        get_signed_part = False

    flatten_decrypted_payloads(piece.gpg_data.plainobj, replyinfo_obj, get_signed_part)

    # to catch public keys in encrypted blocks
    prepare_for_reply(piece.gpg_data.plainobj, replyinfo_obj)


def prepare_for_reply_pubkey (piece, replyinfo_obj):
    """Helper function for prepare_for_reply(). Marks pubkey import status.

    Marks replyinfo_obj with pub key import status.

    Args:
        piece: a PayloadPiece object
        replyinfo_obj: a ReplyInfo object

    Pre:
        piece.piece_type should be set to TxtType.pubkey .

    Post:
        replyinfo_obj has its fields updated.
    """

    if piece.gpg_data.keys == []:
        if piece.gpg_data.key_cannot_encrypt == True:
            replyinfo_obj.key_cannot_encrypt = True
    else:
        replyinfo_obj.pubkey_success = True

        # prefer public key as a fallback for the encrypted reply
        replyinfo_obj.fallback_target_key = piece.gpg_data.keys[0]


def prepare_for_reply_sig (piece, replyinfo_obj):
    """Helper function for prepare_for_reply(). Marks sig verification status.

    Marks replyinfo_obj with signature verification status.

    Args:
        piece: a PayloadPiece object
        replyinfo_obj: a ReplyInfo object

    Pre:
        piece.piece_type should be set to TxtType.signature, or
        TxtType.detachedsig .

    Post:
        replyinfo_obj has its fields updated.
    """

    if piece.gpg_data.sigs == []:
        replyinfo_obj.sig_failure = True

        if piece.gpg_data.sigkey_missing == True:
            replyinfo_obj.sigkey_missing = True

        if piece.gpg_data.key_cannot_encrypt == True:
            replyinfo_obj.key_cannot_encrypt = True

    else:
        replyinfo_obj.sig_success = True

        if replyinfo_obj.fallback_target_key == None:
            replyinfo_obj.fallback_target_key = piece.gpg_data.sigs[0]

    if (piece.piece_type == TxtType.signature):
        # to catch public keys in signature blocks
        prepare_for_reply(piece.gpg_data.plainobj, replyinfo_obj)


def flatten_decrypted_payloads (eddymsg_obj, replyinfo_obj, get_signed_part):
    """For creating a string representation of a signed, encrypted part.

    When given a decrypted payload, it will add either the plaintext or signed
    plaintext to the reply message, depeding on 'get_signed_part'. This is
    useful for ensuring that the reply message only comes from a signed and
    ecrypted GPG message. It also sets the target_key for encrypting the reply
    if it's told to get signed text only.

    Args:
        eddymsg_obj: the message in EddyMsg format created by decrypting GPG
            text
        replyinfo_obj: a ReplyInfo object for holding the message to quote and
            the target_key to encrypt to.
        get_signed_part: True if we should only include text that contains a
            further signature. If False, then include plain text.

    Returns:
        Nothing

    Pre:
        The EddyMsg instance passed in should be a piece.gpg_data.plainobj
        which represents decrypted text. It may or may not be signed on that
        level.

    Post:
        the ReplyInfo instance may have a new 'target_key' set and its
        'msg_to_quote' will be updated with (possibly signed) plaintext, if any
        could be found.
    """

    if eddymsg_obj == None:
        return

    # recurse on multi-part mime
    if eddymsg_obj.multipart == True:
        for sub in eddymsg_obj.subparts:
            flatten_decrypted_payloads(sub, replyinfo_obj, get_signed_part)

    for piece in eddymsg_obj.payload_pieces:
        if (get_signed_part):
            if ((piece.piece_type == TxtType.detachedsig) \
                    or (piece.piece_type == TxtType.signature)) \
                    and (piece.gpg_data != None) \
                    and (piece.gpg_data.plainobj != None):
                        flatten_decrypted_payloads(piece.gpg_data.plainobj, replyinfo_obj, False)
                        replyinfo_obj.target_key = piece.gpg_data.sigs[0]
                        break
        else:
            if (eddymsg_obj.content_disposition == None \
                    or not eddymsg_obj.content_disposition.startswith("attachment")) \
                    and piece.piece_type == TxtType.text:
                replyinfo_obj.msg_to_quote += piece.string


def get_key_from_fp (replyinfo_obj, gpgme_ctx):
    """Obtains a public key object from a key fingerprint

    If the .target_key is not set, then we use .fallback_target_key, if
    available.

    Args:
        replyinfo_obj: ReplyInfo instance
        gpgme_ctx: the gpgme context

    Return:
        Nothing

    Pre:
        Loading a key requires that we have the public key imported. This
        requires that they email contains the pub key block, or that it was
        previously sent to edward.

    Post:
        If the key can be loaded, then replyinfo_obj.reply_to_key points to the
        public key object.  If the key cannot be loaded, then the replyinfo_obj
        is marked as having no public key available. If the key is not capable
        of encryption, it will not be used, and replyinfo_obj will be marked
        accordingly.
    """

    for key in (replyinfo_obj.target_key, replyinfo_obj.fallback_target_key):
        if key != None:
            try:
                encrypt_to_key = gpgme_ctx.get_key(key)

            except gpgme.GpgmeError:
                continue

            if is_key_usable(encrypt_to_key):
                replyinfo_obj.encrypt_to_key = encrypt_to_key
                replyinfo_obj.have_reply_key = True
                replyinfo_obj.key_can_encrypt = True
                return

            else:
                replyinfo_obj.key_cannot_encrypt = True



def write_reply (replyinfo_obj):
    """Write the reply email body about the GPG successes/failures.

    The reply is about whether decryption, sig verification and key
    import/loading was successful or failed. If text was successfully decrypted
    and verified, then the first instance of such text will be included in
    quoted form.

    Args:
        replyinfo_obj: contains details of GPG processing status

    Returns:
        the plaintext message to be sent to the user

    Pre:
        replyinfo_obj should be populated with info about GPG processing status.
    """

    reply_plain = ""

    if (replyinfo_obj.pubkey_success == True):
        reply_plain += replyinfo_obj.replies['greeting']
        reply_plain += "\n\n"


    if replyinfo_obj.decrypt_success == True:
        debug('decrypt success')
        reply_plain += replyinfo_obj.replies['success_decrypt']

        if (replyinfo_obj.sig_success == True) and (replyinfo_obj.have_reply_key == True):
            debug('message quoted')
            reply_plain += replyinfo_obj.replies['space']
            reply_plain += replyinfo_obj.replies['quote_follows']
            reply_plain += "\n\n"
            quoted_text = email_quote_text(replyinfo_obj.msg_to_quote)
            reply_plain += quoted_text

        reply_plain += "\n\n"

    elif replyinfo_obj.decrypt_failure == True:
        debug('decrypt failure')
        reply_plain += replyinfo_obj.replies['failed_decrypt']
        reply_plain += "\n\n"


    if replyinfo_obj.sig_success == True:
        debug('signature success')
        reply_plain += replyinfo_obj.replies['sig_success']
        reply_plain += "\n\n"

    elif replyinfo_obj.sig_failure == True:
        debug('signature failure')
        reply_plain += replyinfo_obj.replies['sig_failure']
        reply_plain += "\n\n"


    if (replyinfo_obj.pubkey_success == True):
        debug('public key received')
        reply_plain += replyinfo_obj.replies['public_key_received']
        reply_plain += "\n\n"

    elif (replyinfo_obj.sigkey_missing == True):
        debug('no public key')
        reply_plain += replyinfo_obj.replies['no_public_key']
        reply_plain += "\n\n"

    elif (replyinfo_obj.key_can_encrypt == False) \
            and (replyinfo_obj.key_cannot_encrypt == True):
        debug('bad public key')
        reply_plain += replyinfo_obj.replies['no_public_key']
        reply_plain += "\n\n"


    if (reply_plain == ""):
        debug('plaintext message')
        reply_plain += replyinfo_obj.replies['failed_decrypt']
        reply_plain += "\n\n"


    reply_plain += replyinfo_obj.replies['signature']
    reply_plain += "\n\n"

    return reply_plain


def add_gpg_key (key_block, gpgme_ctx):
    """Adds a GPG pubkey to the local keystore

    This adds keys received through email into the key store so they can be
    used later.

    Args:
        key_block: the string form of the ascii-armored public key block
        gpgme_ctx: the gpgme context

    Returns:
        the fingerprint(s) of the imported key(s) which can be used for
        encryption, and a boolean marking whether none of the keys are capable
        of encryption.
    """

    fp = io.BytesIO(key_block.encode('ascii'))

    try:
        result = gpgme_ctx.import_(fp)
        imports = result.imports
    except gpgme.GpgmeError:
        imports = []

    key_fingerprints = []
    key_cannot_encrypt = False

    for import_res in imports:
        fingerprint = import_res[0]

        try:
            key_obj = gpgme_ctx.get_key(fingerprint)
        except:
            key_obj = None

        if key_obj != None and is_key_usable(key_obj):
            key_fingerprints += [fingerprint]
            key_cannot_encrypt = False

            debug("added gpg key: " + fingerprint)

        elif key_fingerprints == []:
            key_cannot_encrypt = True

    return (key_fingerprints, key_cannot_encrypt)


def verify_sig_message (msg_block, gpgme_ctx):
    """Verifies the signature of a signed, ascii-armored block of text.

    It encodes the string into ascii, since binary GPG files are currently
    unsupported, and alternative, the ascii-armored format is encodable into
    ascii.

    Args:
        msg_block: a GPG Message block in string form. It may be encrypted or
            not. If it is encrypted, it will return empty results.
        gpgme_ctx: the gpgme context

    Returns:
        A tuple containing the plaintext bytes of the signed part, the list of
        fingerprints of encryption-capable keys signing the data, a boolean
        marking whether edward is missing all public keys for validating any of
        the signatures, and a boolean marking whether all sigs' keys are
        incapable of encryption. If verification failed, perhaps because the
        message was also encrypted, sensible default values are returned.
    """

    block_b = io.BytesIO(msg_block.encode('ascii'))
    plain_b = io.BytesIO()

    try:
        sigs = gpgme_ctx.verify(block_b, None, plain_b)
    except gpgme.GpgmeError:
        return ("",[],False,False)

    plaintext_b = plain_b.getvalue()

    (fingerprints, sigkey_missing, key_cannot_encrypt) = get_signature_fp(sigs, gpgme_ctx)

    return (plaintext_b, fingerprints, sigkey_missing, key_cannot_encrypt)


def verify_detached_signature (detached_sig, plaintext_bytes, gpgme_ctx):
    """Verifies the signature of a detached signature.

    This requires the signature part and the signed part as separate arguments.

    Args:
        detached_sig: the signature part of the detached signature
        plaintext_bytes: the byte form of the message being signed.
        gpgme_ctx: the gpgme context

    Returns:
        A tuple containging a list of encryption capable signing fingerprints
        if the signature verification was sucessful, a boolean marking whether
        edward is missing all public keys for validating any of the signatures,
        and a boolean marking whether all signing keys are incapable of
        encryption.  Otherwise, a tuple containing an empty list and True are
        returned.
    """

    detached_sig_fp = io.BytesIO(detached_sig.encode('ascii'))
    plaintext_fp = io.BytesIO(plaintext_bytes)

    try:
        sigs = gpgme_ctx.verify(detached_sig_fp, plaintext_fp, None)
    except gpgme.GpgmeError:
        return ([],False,False)

    (fingerprints, sigkey_missing, key_cannot_encrypt) = get_signature_fp(sigs, gpgme_ctx)

    return (fingerprints, sigkey_missing, key_cannot_encrypt)


def decrypt_block (msg_block, gpgme_ctx):
    """Decrypts a block of GPG text and verifies any included sigatures.

    Some encypted messages have embeded signatures, so those are verified too.

    Args:
        msg_block: the encrypted(/signed) text
        gpgme_ctx: the gpgme context

    Returns:
        A tuple containing plaintext bytes, encryption-capable signatures (if
        decryption and signature verification were successful, respectively), a
        boolean marking whether edward is missing all public keys for
        validating any of the signatures, and a boolean marking whether all
        signature keys are incapable of encryption.
    """

    block_b = io.BytesIO(msg_block.encode('ascii'))
    plain_b = io.BytesIO()

    try:
        sigs = gpgme_ctx.decrypt_verify(block_b, plain_b)
    except gpgme.GpgmeError:
        return ("",[],False,False)

    plaintext_b = plain_b.getvalue()

    (fingerprints, sigkey_missing, key_cannot_encrypt) = get_signature_fp(sigs, gpgme_ctx)

    return (plaintext_b, fingerprints, sigkey_missing, key_cannot_encrypt)


def get_signature_fp (sigs, gpgme_ctx):
    """Selects valid signatures from output of gpgme signature verifying functions

    get_signature_fp returns a list of valid signature fingerprints if those
    fingerprints are associated with available keys capable of encryption.

    Args:
        sigs: a signature verification result object list
        gpgme_ctx: a gpgme context

    Returns:
        fingerprints: a list of fingerprints
        sigkey_missing: a boolean marking whether public keys are missing for
            all available signatures.
        key_cannot_encrypt: a boolearn marking whether available public keys are
            incapable of encryption.
    """

    sigkey_missing = False
    key_cannot_encrypt = False
    fingerprints = []

    for sig in sigs:
        if (sig.summary == 0) or (sig.summary & gpgme.SIGSUM_VALID != 0) or (sig.summary & gpgme.SIGSUM_GREEN != 0):
            try:
                key_obj = gpgme_ctx.get_key(sig.fpr)
            except:
                if fingerprints == []:
                    sigkey_missing = True
                    continue

            if is_key_usable(key_obj):
                fingerprints += [sig.fpr]
                key_cannot_encrypt = False
                sigkey_missing = False

            elif fingerprints == []:
                key_cannot_encrypt = True

        elif fingerprints == []:
            if (sig.summary & gpgme.SIGSUM_KEY_MISSING != 0):
                sigkey_missing = True

    return (fingerprints, sigkey_missing, key_cannot_encrypt)


def is_key_usable (key_obj):
    """Returns boolean representing key usability regarding encryption

    Tests various feature of key and returns usability

    Args:
        key_obj: a gpgme key object

    Returns:
        A boolean representing key usability
    """
    if key_obj.can_encrypt and not key_obj.invalid and not key_obj.expired \
            and not key_obj.revoked and not key_obj.disabled:
        return True
    else:
        return False


def email_to_reply_to_subject (email_bytes):
    """Returns the email's To:, Reply-To: (or From:), and Subject: fields

    Returns this information from an email.

    Args:
        email_bytes: the byte string form of the email

    Returns:
        the email To:, Reply-To: (or From:), and Subject: fields as strings
    """

    email_struct = email.parser.BytesHeaderParser().parsebytes(email_bytes)

    email_to        = email_struct['To']
    email_from      = email_struct['From']
    email_reply_to  = email_struct['Reply-To']

    email_subject   = email_struct['Subject']

    if email_reply_to == None:
        email_reply_to = email_from

    return email_to, email_reply_to, email_subject


def import_lang_pick_address(email_to, hostname):
    """Imports language file for i18n support; makes reply from address

    The language imported depends on the To: address of the email received by
    edward. an -en ending implies the English language, whereas a -ja ending
    implies Japanese. The list of supported languages is listed in the 'langs'
    list at the beginning of the program. This function also chooses the
    language-dependent address which can be used as the From address in the
    reply email.

    Args:
        email_to: the string containing the email address that the mail was
            sent to.
        hostname: the hostname part of the reply email's from address

    Returns:
        the reference to the imported language module. The only variable in
        this file is the 'replies' dictionary.
    """

    # default
    use_lang = "en"

    if email_to != None:
        for lang in langs:
            if "edward-" + lang in email_to:
                use_lang = lang
                break

    lang_mod_name = "lang." + re.sub('-', '_', use_lang)
    lang_module = importlib.import_module(lang_mod_name)

    reply_from = "edward-" + use_lang + "@" + hostname

    return lang_module, reply_from


def generate_encrypted_mime (plaintext, email_to, email_subject, encrypt_to_key,
                    gpgme_ctx):
    """This function creates the mime email reply. It can encrypt the email.

    If the encrypt_key is included, then the email is encrypted and signed.
    Otherwise it is unencrypted.

    Args:
        plaintext: the plaintext body of the message to create.
        email_to: the email address to reply to
        email_subject: the subject to use in reply
        encrypt_to_key: the key object to use for encrypting the email. (or
            None)
        gpgme_ctx: the gpgme context

    Returns
        A string version of the mime message, possibly encrypted and signed.
    """

    plaintext_mime = MIMEText(plaintext)
    plaintext_mime.set_charset('utf-8')

    if (encrypt_to_key != None):

        encrypted_text = encrypt_sign_message(plaintext_mime.as_string(),
                                              encrypt_to_key,
                                              gpgme_ctx)

        control_mime = MIMEApplication("Version: 1",
                                       _subtype='pgp-encrypted',
                                       _encoder=email.encoders.encode_7or8bit)
        control_mime['Content-Description'] = 'PGP/MIME version identification'
        control_mime.set_charset('us-ascii')

        encoded_mime = MIMEApplication(encrypted_text,
                                       _subtype='octet-stream; name="encrypted.asc"',
                                       _encoder=email.encoders.encode_7or8bit)
        encoded_mime['Content-Description'] = 'OpenPGP encrypted message'
        encoded_mime['Content-Disposition'] = 'inline; filename="encrypted.asc"'
        encoded_mime.set_charset('us-ascii')

        message_mime = MIMEMultipart(_subtype="encrypted", protocol="application/pgp-encrypted")
        message_mime.attach(control_mime)
        message_mime.attach(encoded_mime)
        message_mime['Content-Disposition'] = 'inline'

    else:
        message_mime = plaintext_mime

    message_mime['To'] = email_to
    message_mime['Subject'] = email_subject

    reply = message_mime.as_string()

    return reply


def send_reply(email_txt, reply_to, reply_from):
    """Sends reply email

    Sent to original sender

    Args:
        email_txt: message as a string
        reply_to: recipient of reply
        reply_from: edward's specific email address

    Post:
        Email is sent
    """

    if reply_to == None:
        error("*** ERROR: No one to send email to.")
        exit(1)

    s = smtplib.SMTP('localhost')
    s.sendmail(reply_from, reply_to, email_txt)
    s.quit()


def email_quote_text (text):
    """Quotes input text by inserting "> "s

    This is useful for quoting a text for the reply message. It inserts "> "
    strings at the beginning of lines.

    Args:
        text: plain text to quote

    Returns:
        Quoted text
    """

    quoted_message = re.sub(r'^', r'> ', text, flags=re.MULTILINE)

    return quoted_message


def encrypt_sign_message (plaintext, encrypt_to_key, gpgme_ctx):
    """Encrypts and signs plaintext

    This encrypts and signs a message.

    Args:
        plaintext: text to sign and ecrypt
        encrypt_to_key: the key object to encrypt to
        gpgme_ctx: the gpgme context

    Returns:
        An encrypted and signed string of text
    """

    # the plaintext should be mime encoded in an ascii-compatible form
    plaintext_bytes = io.BytesIO(plaintext.encode('ascii'))
    encrypted_bytes = io.BytesIO()

    gpgme_ctx.encrypt_sign([encrypt_to_key], gpgme.ENCRYPT_ALWAYS_TRUST,
            plaintext_bytes, encrypted_bytes)

    encrypted_txt = encrypted_bytes.getvalue().decode('ascii')
    return encrypted_txt


def error (error_msg):
    """Write an error message to stdout

    The error message includes the program name.

    Args:
        error_msg: the message to print

    Returns:
        Nothing

    Post:
        An error message is printed to stdout
    """

    sys.stderr.write(progname + ": " + str(error_msg) + "\n")


def debug (debug_msg):
    """Writes a debug message to stdout if debug == True

    If the debug option is set in edward_config.py, then the passed message
    gets printed to stdout.

    Args:
        debug_msg: the message to print to stdout

    Returns:
        Nothing

    Post:
        A debug message is printed to stdout
    """

    if edward_config.debug == True:
        error(debug_msg)


def handle_args ():
    """Sets the progname variable and processes optional argument

    If there are more than two arguments then edward complains and quits. An
    single "-p" argument sets the print_reply_only option, which makes edward
    print email replies instead of mailing them.

    Args:
        None

    Returns:
        True if edward should print arguments instead of mailing them,
        otherwise it returns False.

    Post:
        Exits with error 1 if there are more than two arguments, otherwise
        returns the print_reply_only option.
    """

    global progname
    progname = sys.argv[0]

    print_reply_only = False

    if len(sys.argv) > 2:
        print(progname + " usage:  " + progname + " [-p]\n\n" \
                + "        -p      print reply message to stdout, do not mail it\n", \
                file=sys.stderr)
        exit(1)

    elif (len(sys.argv) == 2) and (sys.argv[1] == "-p"):
        print_reply_only = True

    return print_reply_only


if __name__ == "__main__":
    """Executes main if this file is not loaded interactively"""

    main()