|
| 1 | +# coding: utf-8 |
| 2 | +""" |
| 3 | +Public-Inbox stats about mailing lists threads |
| 4 | +
|
| 5 | +Config example:: |
| 6 | +
|
| 7 | + [inbox] |
| 8 | + type = public-inbox |
| 9 | + url = https://lore.kernel.org |
| 10 | +""" |
| 11 | + |
| 12 | +import copy |
| 13 | +import datetime |
| 14 | +import email.utils |
| 15 | +import gzip |
| 16 | +import mailbox |
| 17 | +import tempfile |
| 18 | +import typing |
| 19 | +import urllib.parse |
| 20 | + |
| 21 | +import requests |
| 22 | + |
| 23 | +from did.base import Config, Date, ReportError, User |
| 24 | +from did.stats import Stats, StatsGroup |
| 25 | +from did.utils import item, log |
| 26 | + |
| 27 | + |
| 28 | +class Message(object): |
| 29 | + def __init__(self, msg: mailbox.mboxMessage) -> None: |
| 30 | + self.msg = msg |
| 31 | + |
| 32 | + def __msg_id(self, keyid: str) -> str: |
| 33 | + msgid = self.msg[keyid] |
| 34 | + if msgid is None: |
| 35 | + return None |
| 36 | + |
| 37 | + return msgid.lstrip("<").rstrip(">") |
| 38 | + |
| 39 | + def id(self) -> str: |
| 40 | + return self.__msg_id("Message-Id") |
| 41 | + |
| 42 | + def parent_id(self) -> str: |
| 43 | + return self.__msg_id("In-Reply-To") |
| 44 | + |
| 45 | + def subject(self) -> str: |
| 46 | + subject = self.msg["Subject"] |
| 47 | + |
| 48 | + subject = " ".join(subject.splitlines()) |
| 49 | + subject = " ".join(subject.split()) |
| 50 | + |
| 51 | + return subject |
| 52 | + |
| 53 | + def date(self) -> datetime.datetime: |
| 54 | + return email.utils.parsedate_to_datetime(self.msg["Date"]) |
| 55 | + |
| 56 | + def is_thread_root(self) -> bool: |
| 57 | + return self.parent_id() is None |
| 58 | + |
| 59 | + def is_from_user(self, user: str) -> bool: |
| 60 | + msg_from = email.utils.parseaddr(self.msg["From"])[1] |
| 61 | + |
| 62 | + return email.utils.parseaddr(user)[1] == msg_from |
| 63 | + |
| 64 | + def is_between_dates(self, since: Date, until: Date) -> bool: |
| 65 | + msg_date = self.date().date() |
| 66 | + |
| 67 | + return msg_date >= since.date and msg_date <= until.date |
| 68 | + |
| 69 | + |
| 70 | +def _unique_messages(mbox: mailbox.mbox) -> typing.Iterable[Message]: |
| 71 | + msgs = dict() |
| 72 | + for msg in mbox.values(): |
| 73 | + msg = Message(msg) |
| 74 | + id = msg.id() |
| 75 | + |
| 76 | + if id not in msgs: |
| 77 | + msgs[id] = msg |
| 78 | + yield msg |
| 79 | + |
| 80 | + |
| 81 | +class PublicInbox(object): |
| 82 | + def __init__(self, parent, user: User, url: str) -> None: |
| 83 | + self.parent = parent |
| 84 | + self.threads_cache = dict() |
| 85 | + self.messages_cache = dict() |
| 86 | + self.url = url |
| 87 | + self.user = user |
| 88 | + |
| 89 | + def __get_url(self, path: str) -> str: |
| 90 | + return urllib.parse.urljoin(self.url, path) |
| 91 | + |
| 92 | + def _get_message_url(self, msg: Message) -> str: |
| 93 | + return self.__get_url("/r/%s/" % msg.id()) |
| 94 | + |
| 95 | + def _print_msg(self, options, msg: Message) -> None: |
| 96 | + if options.format == 'markdown': |
| 97 | + item("[{0}]({1})".format(msg.subject(), self._get_message_url(msg)), |
| 98 | + level=1, options=options) |
| 99 | + |
| 100 | + else: |
| 101 | + item(msg.subject(), level=1, options=options) |
| 102 | + |
| 103 | + if options.verbose: |
| 104 | + opt = copy.deepcopy(options) |
| 105 | + opt.width = 0 |
| 106 | + item(self._get_message_url(msg), level=2, options=opt) |
| 107 | + |
| 108 | + def __get_mbox_from_content(self, content: bytes) -> mailbox.mbox: |
| 109 | + content = gzip.decompress(content) |
| 110 | + |
| 111 | + with tempfile.NamedTemporaryFile() as tmp: |
| 112 | + tmp.write(content) |
| 113 | + tmp.seek(0) |
| 114 | + |
| 115 | + return mailbox.mbox(tmp.name) |
| 116 | + |
| 117 | + def __get_msgs_from_mbox(self, mbox: mailbox.mbox) -> list[Message]: |
| 118 | + msgs = list() |
| 119 | + |
| 120 | + for msg in _unique_messages(mbox): |
| 121 | + msg_id = msg.id() |
| 122 | + |
| 123 | + log.debug("Found message %s." % msg_id) |
| 124 | + msgs.append(msg) |
| 125 | + |
| 126 | + if msg_id not in self.messages_cache: |
| 127 | + log.debug("Message %s is new, adding to the cache." % msg_id) |
| 128 | + self.messages_cache[msg_id] = msg |
| 129 | + |
| 130 | + return msgs |
| 131 | + |
| 132 | + def __fetch_thread_root(self, msg: Message) -> Message: |
| 133 | + msg_id = msg.id() |
| 134 | + url = self.__get_url("/all/%s/t.mbox.gz" % msg_id) |
| 135 | + |
| 136 | + log.debug("Fetching message %s thread (%s)" % (msg_id, url)) |
| 137 | + resp = requests.get(url) |
| 138 | + mbox = self.__get_mbox_from_content(resp.content) |
| 139 | + for msg in self.__get_msgs_from_mbox(mbox): |
| 140 | + if msg.is_thread_root(): |
| 141 | + log.debug("Found message %s thread root: %s." % (msg_id, msg.id())) |
| 142 | + return msg |
| 143 | + |
| 144 | + def __get_thread_root(self, msg: Message) -> Message: |
| 145 | + log.debug("Looking for thread root of message %s" % msg.id()) |
| 146 | + if msg.is_thread_root(): |
| 147 | + log.debug("Message is thread root already. Returning.") |
| 148 | + return msg |
| 149 | + |
| 150 | + parent_id = msg.parent_id() |
| 151 | + if parent_id not in self.messages_cache: |
| 152 | + root = self.__fetch_thread_root(msg) |
| 153 | + log.debug("Found root message %s for message %s" % (root.id(), msg.id())) |
| 154 | + return root |
| 155 | + |
| 156 | + while True: |
| 157 | + log.debug("Parent is %s" % parent_id) |
| 158 | + assert parent_id in self.messages_cache |
| 159 | + parent = self.messages_cache[parent_id] |
| 160 | + if parent.is_thread_root(): |
| 161 | + log.debug("Parent is the thread root, returning.") |
| 162 | + return parent |
| 163 | + |
| 164 | + parent_id = parent.parent_id() |
| 165 | + if parent_id not in self.messages_cache: |
| 166 | + root = self.__fetch_thread_root(msg) |
| 167 | + log.debug( |
| 168 | + "Found root message %s for message %s" % |
| 169 | + (root.id(), msg.id())) |
| 170 | + return root |
| 171 | + |
| 172 | + def __fetch_all_threads(self, since: Date, until: Date) -> list[Message]: |
| 173 | + since_str = since.date.isoformat() |
| 174 | + until_str = until.date.isoformat() |
| 175 | + |
| 176 | + log.info("Fetching all mails on server %s from %s between %s and %s" % |
| 177 | + (self.url, self.user, since_str, until_str)) |
| 178 | + resp = requests.post( |
| 179 | + self.__get_url("/all/"), |
| 180 | + headers={"Content-Length": "0"}, |
| 181 | + params={ |
| 182 | + "q": "(f:%s AND d:%s..%s)" |
| 183 | + % (self.user.email, since_str, until_str), |
| 184 | + "x": "m", |
| 185 | + }, |
| 186 | + ) |
| 187 | + |
| 188 | + if not resp.ok: |
| 189 | + return [] |
| 190 | + |
| 191 | + mbox = self.__get_mbox_from_content(resp.content) |
| 192 | + return self.__get_msgs_from_mbox(mbox) |
| 193 | + |
| 194 | + def get_all_threads(self, since: Date, until: Date): |
| 195 | + if (since, until) not in self.threads_cache: |
| 196 | + self.threads_cache[(since, until)] = self.__fetch_all_threads(since, until) |
| 197 | + |
| 198 | + assert (since, until) in self.threads_cache |
| 199 | + |
| 200 | + found = list() |
| 201 | + for msg in self.threads_cache[(since, until)]: |
| 202 | + msg_id = msg.id() |
| 203 | + if msg_id in found: |
| 204 | + continue |
| 205 | + |
| 206 | + if not msg.is_thread_root(): |
| 207 | + root = self.__get_thread_root(msg) |
| 208 | + root_id = root.id() |
| 209 | + if root_id in found: |
| 210 | + log.debug("Root message already encountered... Skip.") |
| 211 | + continue |
| 212 | + |
| 213 | + found.append(root_id) |
| 214 | + yield root |
| 215 | + else: |
| 216 | + found.append(msg_id) |
| 217 | + yield msg |
| 218 | + |
| 219 | + |
| 220 | +class ThreadsStarted(Stats): |
| 221 | + """ Mail threads started """ |
| 222 | + |
| 223 | + def fetch(self): |
| 224 | + log.info( |
| 225 | + "Searching for new threads on {0} started by {1}".format( |
| 226 | + self.parent.url, |
| 227 | + self.user, |
| 228 | + ) |
| 229 | + ) |
| 230 | + |
| 231 | + self.stats = [ |
| 232 | + msg |
| 233 | + for msg in self.parent.public_inbox.get_all_threads( |
| 234 | + self.options.since, self.options.until) |
| 235 | + if msg.is_from_user(self.user.email) |
| 236 | + and msg.is_between_dates(self.options.since, self.options.until) |
| 237 | + ] |
| 238 | + |
| 239 | + def show(self): |
| 240 | + if not self._error and not self.stats: |
| 241 | + return |
| 242 | + |
| 243 | + self.header() |
| 244 | + for msg in self.stats: |
| 245 | + self.parent.public_inbox._print_msg(self.options, msg) |
| 246 | + |
| 247 | + |
| 248 | +class ThreadsInvolved(Stats): |
| 249 | + """ Mail threads involved in """ |
| 250 | + |
| 251 | + def fetch(self): |
| 252 | + log.info( |
| 253 | + "Searching for mail threads on {0} where {1} was involved".format( |
| 254 | + self.parent.url, |
| 255 | + self.user, |
| 256 | + ) |
| 257 | + ) |
| 258 | + |
| 259 | + self.stats = [ |
| 260 | + msg |
| 261 | + for msg in self.parent.public_inbox.get_all_threads( |
| 262 | + self.options.since, self.options.until) |
| 263 | + if not msg.is_from_user(self.user.email) |
| 264 | + or not msg.is_between_dates(self.options.since, self.options.until) |
| 265 | + ] |
| 266 | + |
| 267 | + def show(self): |
| 268 | + if not self._error and not self.stats: |
| 269 | + return |
| 270 | + |
| 271 | + self.header() |
| 272 | + for msg in self.stats: |
| 273 | + self.parent.public_inbox._print_msg(self.options, msg) |
| 274 | + |
| 275 | + |
| 276 | +class PublicInboxStats(StatsGroup): |
| 277 | + """ Public-Inbox Mailing List Archive """ |
| 278 | + |
| 279 | + order = 750 |
| 280 | + |
| 281 | + def __init__(self, option, name=None, parent=None, user=None): |
| 282 | + StatsGroup.__init__(self, option, name, parent, user) |
| 283 | + |
| 284 | + config = dict(Config().section(option)) |
| 285 | + try: |
| 286 | + self.url = config["url"] |
| 287 | + except KeyError: |
| 288 | + raise ReportError("No url in the [{0}] section.".format(option)) |
| 289 | + |
| 290 | + self.public_inbox = PublicInbox(self.parent, self.user, self.url) |
| 291 | + self.stats = [ |
| 292 | + ThreadsStarted(option=option + "-started", parent=self), |
| 293 | + ThreadsInvolved(option=option + "-involved", parent=self), |
| 294 | + ] |
0 commit comments