def doc_from_bytes(docid, rdkey, b):
msg = message_from_string(b)
doc = {}
mp = doc['multipart'] = msg.is_multipart()
headers = doc['headers'] = {}
# Given we have no opportunity to introduce an object which can ignore
# the case of headers, we lowercase the keys
for hn in msg.keys():
vals = msg.get_all(hn)
if vals:
# first do any charset etc conversion...
vals = [_safe_convert_header(v) for v in vals]
if hn.lower() == 'references':
# email.utils.unquote will do bad things to references headers (stripping
# initial and trailing <>'s, so we don't want to use it for the
# references header-- but other fields seem ok. We split the references
# into a list here because why not.
headers[hn.lower()] = [extract_message_ids(vals[0])]
else:
headers[hn.lower()] = [unquote(v) for v in vals]
# a sanity check and to help debug an obscure bug which seemed to
# cause the wrong 'source' doc being passed!
if __debug__ and rdkey[0]=='email' and hn.lower()=='message-id':
from raindrop.proto.imap import get_rdkey_for_email
assert tuple(rdkey)==get_rdkey_for_email(vals[0]), (rdkey, docid, vals)
# XXX - technically msg objects are recursive; handling that requires
# more thought. For now, assume they are flat.
# We must return non-text parts in attachments, so just return
# *everything* in attachments.
attachments = doc['_attachments'] = {}
if mp:
# a multi-part message - flatten it here by walking the list, but
# only looking at the 'leaf' nodes.
# attachments have lost their order; this object helps keep the
# other and is a convenient place to stash other headers coming
# with this part.
mi = doc['multipart_info'] = []
i = 1
for attach in msg.walk():
if not attach.is_multipart():
name = sanitize_attach_name(attach.get_filename())
if not name:
name = "subpart-%d" % i
i += 1
attachments[name] = attach_from_msg((docid, name), attach)
# Put together info about the attachment.
ah = {}
for hn, hv in attach.items():
ah[hn.lower()] = _safe_convert_header(hv)
# content-type is redundant, but may be helpful...
ct = attachments[name]['content_type']
info = {'name': name, 'headers': ah, 'content_type': ct}
mi.append(info)
else:
attachments['body'] = attach_from_msg((docid, 'body'), msg)
return doc
def handler(doc):
# a 'rfc822' stores 'headers' as a dict, with each entry being a list.
# We only care about headers which rfc5322 must appear 0 or 1 times, so
# flatten the header values here...
headers = dict((k, v[0]) for (k, v) in doc['headers'].iteritems())
self_header_message_id = headers.get('message-id')
# check something hasn't got confused...
assert get_rdkey_for_email(self_header_message_id) == tuple(doc['rd_key']), doc
if 'references' in headers:
header_message_ids = headers['references']
elif 'in-reply-to' in headers:
header_message_ids = [headers['in-reply-to']]
else:
header_message_ids = []
# save off the list of referenced messages (XXX - but this isn't used?)
references = header_message_ids[:]
# see if the self-message already exists...
header_message_ids.append(self_header_message_id)
uniq_header_message_ids = set(header_message_ids)
logger.debug("header_message_ids: %s ", header_message_ids)
logger.debug("references: %s", '\n\t'.join(references))
# Open a view trying to locate an existing conversation for any of these
# headers.
keys = [['rd.core.content', 'key-schema_id', [['email', mid], 'rd.msg.conversation']]
for mid in uniq_header_message_ids]
result = open_view(keys=keys, reduce=False,
include_docs=True)
# build a map of the keys we actually got back.
rows = [r for r in result['rows'] if 'error' not in r]
if rows:
assert 'doc' in rows[0], rows
convo_id = rows[0]['doc']['conversation_id']
logger.debug("FOUND CONVERSATION header_message_id %s with conversation_id %s",
self_header_message_id, convo_id)
seen_ids = set(r['value']['rd_key'][1] for r in rows)
else:
logger.debug("CREATING conversation_id %s", header_message_ids[0])
convo_id = header_message_ids[0]
seen_ids = None
items = {'conversation_id': convo_id}
# create convo records for any messages which don't yet exist -
# presumably that includes me too!
for hid in uniq_header_message_ids:
if seen_ids is None or hid not in seen_ids:
rdkey = ['email', hid]
logger.debug('emitting convo schema referenced message %r', rdkey)
emit_schema('rd.msg.conversation', items, rd_key=rdkey)
# make sure current doc gets emitted in case it was
# not part of the uniq_header_message_ids
if doc['rd_key'][1] not in uniq_header_message_ids:
logger.debug('emitting convo schema for my document %(rd_key)r', doc)
emit_schema('rd.msg.conversation', items, rd_key=doc['rd_key'])
def gen_corpus_docs(self, corpus_name, item_spec="*"):
cwd = os.getcwd()
corpus_dir = self.get_corpus_dir(corpus_name)
num = 0
# We try and make life simple for people by auto-determining the
# 'schema' for some well-known file types (eg, .rfc822.txt)
pattern = "%s/%s.*" % (corpus_dir, item_spec)
base_names = set()
for filename in glob.iglob(pattern):
try:
path, name = os.path.split(filename)
# don't use splitext - we want the *first* dot.
first, _ = filename.split(".", 1)
base = os.path.join(path, first)
except ValueError:
base = filename
base_names.add(base)
for basename in base_names:
if basename.endswith('README') or basename.endswith('raindrop'):
continue
# .json files get first go - they may 'override' what we would
# otherwise deduce.
elif os.path.exists(basename + ".json"):
filename = basename + ".json"
with open(filename) as f:
try:
ob = json.load(f)
except ValueError, why:
self.fail("%r has invalid json: %r" % (filename, why))
for name, data in ob.get('_attachments', {}).iteritems():
fname = os.path.join(corpus_dir, data['filename'])
with open(fname, 'rb') as attach_f:
enc_data = base64.encodestring(attach_f.read()).replace('\n', '')
data['data'] = enc_data
elif os.path.exists(basename + ".rfc822.txt"):
# plain rfc822.txt file.
with open(basename + ".rfc822.txt", 'rb') as f:
data = f.read()
msg_id = message_from_string(data)['message-id']
ob = {
'rd_schema_id': 'rd.msg.rfc822',
'rd_key': get_rdkey_for_email(msg_id),
'rd_ext_id': 'proto.imap',
'rd_source': None,
'_attachments' : {
'rfc822': {
'content_type': 'message',
'data': base64.encodestring(data).replace('\n', ''),
}
}
}
def handler(doc):
# a 'rfc822' stores 'headers' as a dict, with each entry being a list.
# We only care about headers which rfc5322 must appear 0 or 1 times, so
# flatten the header values here...
headers = dict((k, v[0]) for (k, v) in doc['headers'].iteritems())
self_header_message_id = headers.get('message-id')
# check something hasn't got confused...
assert get_rdkey_for_email(self_header_message_id) == tuple(doc['rd_key']), doc
references = set()
if 'references' in headers:
# 'references' is a bit special though - the provider of the source
# schema has already split them!
for ref in headers['references']:
references.add(ref)
if 'in-reply-to' in headers:
references.add(headers['in-reply-to'])
# the self-message...
references.add(self_header_message_id)
logger.debug("references: %s", '\n\t'.join(references))
keys = (get_rdkey_for_email(mid) for mid in references)
find_and_emit_conversation(keys)
def rfc822_to_schema_item(self, fp):
data = fp.read() # we need to use the data twice...
msg_id = message_from_string(data)['message-id']
si = {
'rd_schema_id': 'rd.msg.rfc822',
'rd_key': get_rdkey_for_email(msg_id),
'rd_ext_id': 'proto.imap',
'rd_source': None,
'attachments' : {
'rfc822': {
'content_type': 'message',
'data': data,
}
},
'items' : {},
}
return si
def failed_recipient(doc):
# failed_recipient is the email address(es) that was the intended target
# which for whatever reason doesn't exist
# we want to grab this because:
# 1. We could look through the address book for similar identities
# 2. In the UI we could offer a message "You tried to email X but the address appears incorrect"
# XXX however currently we do nothing with it
failed_recipient = doc['headers']['x-failed-recipients'][0]
logger.debug("found DSN failed recipient message intended for %s", failed_recipient)
body = open_schema_attachment(doc, 'body')
# Scan Message Body for the original message id
# e.g. Message-ID: <*****@*****.**>
match = re.search('Message-ID:\s*<(.+)>.*', body)
if (match):
logger.debug("found Message-ID header in DSN message '%s'", match.group(1))
id = match.group(1)
# make an rd_key for the referenced message.
rdkey_orig = get_rdkey_for_email(id)
# and say this message and the original are related.
find_and_emit_conversation([doc['rd_key'], rdkey_orig])
else:
logger.info("No match found for DSN messsage")
return
def handler(doc):
# This is dealing with the 'imap folder state cache doc' - it stores
# all meta-data about all items in a folder; so one document holds the
# state for many messages. We first need to determine which are
# different...
rdkeys = []
imap_flags = []
folder_name = doc['rd_key'][1][1]
for item in doc['infos']:
msg_id = item['ENVELOPE'][-1]
rdkey = get_rdkey_for_email(msg_id)
rdkeys.append(rdkey)
imap_flags.append((rdkey, item['FLAGS']))
result = open_view('raindrop!content!all', 'msg-seen-flag', keys=rdkeys)
# turn the result into a dict keyed by rdkey
couch_values = {}
for row in result['rows']:
couch_values[hashable_key(row['key'])] = row['value']
# work out which of these rdkeys actually exist in our db.
existing_rdkeys = set()
keys = [['key-schema_id', [rdkey, 'rd.msg.rfc822']]
for rdkey in rdkeys]
result = open_view(keys=keys, reduce=False)
for row in result['rows']:
existing_rdkeys.add(hashable_key(row['value']['rd_key']))
# find what is different...
nnew = 0
nupdated = 0
# Note it is fairly common to see multiples with the same msg ID in, eg
# a 'drafts' folder, so skip duplicates to avoid conflicts.
seen_keys = set()
for rdkey, flags in imap_flags:
if rdkey in seen_keys:
logger.info('skipping duplicate message in folder %r: %r',
folder_name, rdkey)
continue
if rdkey not in existing_rdkeys:
# this means we haven't actually sucked the message into raindrop
# yet (eg, --max-age may have caused only a subset of the messages
# to be grabbed, although all messages in the folder are returned
# in the input document)
logger.debug('skipping message not yet in folder %r: %r',
folder_name, rdkey)
continue
seen_keys.add(rdkey)
seen_now = "\\Seen" in flags
try:
couch_value = couch_values[rdkey]
except KeyError:
# new message
items = {'seen' : seen_now,
'outgoing_state' : 'incoming',
}
emit_schema('rd.msg.seen', items, rdkey)
nnew += 1
else:
# If the state in couch is anything other than 'incoming'', it
# represents a request to change the state on the server (or the
# process of trying to update the server).
if couch_value.get('outgoing_state') != 'incoming':
logger.info("found outgoing 'seen' state request in doc with key %r", rdkey)
continue
seen_couch = couch_value['seen']
if seen_now != seen_couch:
items = {'seen' : seen_now,
'outgoing_state' : 'incoming',
'_rev' : couch_value['_rev'],
}
emit_schema('rd.msg.seen', items, rdkey)
nupdated += 1
logger.info("folder %r needs %d new and %d updated 'seen' records",
folder_name, nnew, nupdated)
def handler(doc):
# This is dealing with the 'imap folder state cache doc' - it stores
# all meta-data about all items in a folder; so one document holds the
# state for many messages. We first need to determine which are
# different...
keys = []
rdkeys = []
imap_flags = []
folder_name = doc['rd_key'][1][1]
for item in doc['infos']:
msg_id = item['ENVELOPE'][-1]
rdkey = get_rdkey_for_email(msg_id)
rdkeys.append(rdkey)
keys.append(['rd.core.content', 'key-schema_id', [rdkey, 'rd.msg.seen']])
imap_flags.append((rdkey, item['FLAGS']))
result = open_view(keys=keys, reduce=False, include_docs=True)
# turn the result into a dict also keyed by rdkey
couch_docs = {}
for row in result['rows']:
couch_docs[tuple(row['value']['rd_key'])] = row['doc']
# work out which of these rdkeys actually exist in our db.
existing_rdkeys = set()
keys = []
for rdkey in rdkeys:
keys.append(['rd.core.content', 'key-schema_id', [rdkey, 'rd.msg.rfc822']])
result = open_view(keys=keys, reduce=False)
for row in result['rows']:
existing_rdkeys.add(tuple(row['value']['rd_key']))
# find what is different...
nnew = 0
to_up = []
# Note it is fairly common to see multiples with the same msg ID in, eg
# a 'drafts' folder, so skip duplicates to avoid conflicts.
seen_keys = set()
for rdkey, flags in imap_flags:
if rdkey in seen_keys:
logger.info('skipping duplicate message in folder %r: %r',
folder_name, rdkey)
continue
if rdkey not in existing_rdkeys:
logger.debug('skipping message not yet in folder %r: %r',
folder_name, rdkey)
continue
seen_keys.add(rdkey)
seen_now = "\\Seen" in flags
try:
doc = couch_docs[rdkey]
except KeyError:
# new message
items = {'seen' : seen_now,
'outgoing_state' : 'incoming',
}
emit_schema('rd.msg.seen', items, rdkey)
nnew += 1
else:
# If the state in couch is anything other than 'incoming'', it
# represents a request to change the state on the server (or the
# process of trying to update the server).
if doc.get('outgoing_state') != 'incoming':
logger.info("found outgoing 'seen' state request in doc %(_id)r", doc)
continue
seen_couch = doc['seen']
if seen_now != seen_couch:
doc['seen'] = seen_now
to_up.append(doc)
if to_up:
update_documents(to_up)
logger.info("folder %r needs %d new and %d updated 'seen' records",
folder_name, nnew, len(to_up))
