twitter: use import_source and proper merging for tweets from different sources

+ use proper datetime_aware for created_at

my/fbmessenger/common.py

@@ -1,3 +1,5 @@
+from my.core import __NOT_HPI_MODULE__
+
 from datetime import datetime
 from typing import Iterator, Optional, TYPE_CHECKING
 
@@ -35,7 +37,9 @@ class Message(Protocol):
 
 from itertools import chain
 from more_itertools import unique_everseen
-from my.core import Res
+from my.core import warn_if_empty, Res
+
+@warn_if_empty
 def _merge_messages(*sources: Iterator[Res[Message]]) -> Iterator[Res[Message]]:
     # todo might be nice to dump some stats for debugging, e.g. how many were overlapping?
     def key(r: Res[Message]):

my/twitter/all.py

@@ -1,22 +1,51 @@
 """
 Unified Twitter data (merged from the archive and periodic updates)
 """
+from typing import Iterator
+from ..core import Res
+from ..core.source import import_source
+from .common import merge_tweets, Tweet
+
 
 # NOTE: you can comment out the sources you don't need
-from . import twint, archive
-
-from .common import merge_tweets
+src_twint   = import_source(module_name=f'my.twitter.twint')
+src_archive = import_source(module_name=f'my.twitter.archive')
 
 
-def tweets():
+@src_twint
+def _tweets_twint() -> Iterator[Res[Tweet]]:
+    from . import twint as src
+    return src.tweets()
+
+@src_archive
+def _tweets_archive() -> Iterator[Res[Tweet]]:
+    from . import archive as src
+    return src.tweets()
+
+
+@src_twint
+def _likes_twint() -> Iterator[Res[Tweet]]:
+    from . import twint as src
+    return src.likes()
+
+@src_archive
+def _likes_archive() -> Iterator[Res[Tweet]]:
+    from . import archive as src
+    return src.likes()
+
+
+def tweets() -> Iterator[Res[Tweet]]:
     yield from merge_tweets(
-        twint  .tweets(),
-        archive.tweets(),
+        _tweets_twint(),
+        _tweets_archive(),
     )
 
 
-def likes():
+def likes() -> Iterator[Res[Tweet]]:
     yield from merge_tweets(
-        twint  .likes(),
-        archive.likes(),
+        _likes_twint(),
+        _likes_archive(),
     )
+
+
+# TODO maybe to avoid all the boilerplate above could use some sort of module Protocol?

my/twitter/archive.py

@@ -18,7 +18,7 @@ except ImportError as e:
 
 
 from dataclasses import dataclass
-from ..core import Paths
+from ..core import Paths, Res, datetime_aware
 
 @dataclass
 class twitter_archive(user_config):
@@ -32,7 +32,7 @@ config = make_config(twitter_archive)
 
 
 from datetime import datetime
-from typing import List, Optional, Iterable, NamedTuple, Sequence
+from typing import List, Optional, NamedTuple, Sequence, Iterator
 from pathlib import Path
 import json
 
@@ -61,7 +61,7 @@ class Tweet(NamedTuple):
         return self.raw['id_str']
 
     @property
-    def created_at(self) -> datetime:
+    def created_at(self) -> datetime_aware:
         dts = self.raw['created_at']
         return datetime.strptime(dts, '%a %b %d %H:%M:%S %z %Y')
 
@@ -159,12 +159,12 @@ class ZipExport:
         [acc] = self.raw('account')
         return acc['username']
 
-    def tweets(self) -> Iterable[Tweet]:
+    def tweets(self) -> Iterator[Tweet]:
         for r in self.raw('tweet'):
             yield Tweet(r, screen_name=self.screen_name())
 
 
-    def likes(self) -> Iterable[Like]:
+    def likes(self) -> Iterator[Like]:
         # TODO ugh. would be nice to unify Tweet/Like interface
         # however, akeout only got tweetId, full text and url
         for r in self.raw('like'):
@@ -172,18 +172,18 @@ class ZipExport:
 
 
 # todo not sure about list and sorting? although can't hurt considering json is not iterative?
-def tweets() -> Iterable[Tweet]:
+def tweets() -> Iterator[Res[Tweet]]:
     for inp in inputs():
         yield from sorted(ZipExport(inp).tweets(), key=lambda t: t.dt)
 
 
-def likes() -> Iterable[Like]:
+def likes() -> Iterator[Res[Like]]:
     for inp in inputs():
         yield from ZipExport(inp).likes()
 
 
-def stats():
-    from ..core import stat
+from ..core import stat, Stats
+def stats() -> Stats:
     return {
         **stat(tweets),
         **stat(likes),

my/twitter/common.py

@@ -1,12 +1,21 @@
+from my.core import __NOT_HPI_MODULE__
+
 from itertools import chain
+from typing import Iterator, Any
 
 from more_itertools import unique_everseen
 
-from ..core import warn_if_empty, __NOT_HPI_MODULE__
 
+# TODO add proper Protocol for Tweet
+Tweet = Any
+
+
+from my.core import warn_if_empty, Res
 @warn_if_empty
-def merge_tweets(*sources):
-    yield from unique_everseen(
-        chain(*sources),
-        key=lambda t: t.id_str,
-    )
+def merge_tweets(*sources: Iterator[Res[Tweet]]) -> Iterator[Res[Tweet]]:
+    def key(r: Res[Tweet]):
+        if isinstance(r, Exception):
+            return str(r)
+        else:
+            return r.id_str
+    yield from unique_everseen(chain(*sources), key=key)

my/twitter/talon.py

@@ -7,11 +7,12 @@ from dataclasses import dataclass
 from datetime import datetime
 from typing import Iterator, Sequence, Optional, Dict
 
+import pytz
 
 from my.config import twitter as user_config
 
 
-from ..core import Paths
+from ..core import Paths, Res, datetime_aware
 @dataclass
 class config(user_config.talon):
     # paths[s]/glob to the exported sqlite databases
@@ -28,8 +29,7 @@ def inputs() -> Sequence[Path]:
 @dataclass(unsafe_hash=True)
 class Tweet:
     id_str: str
-    # TODO figure out if utc
-    created_at: datetime
+    created_at: datetime_aware
     screen_name: str
     text: str
     urls: Sequence[str]
@@ -45,7 +45,6 @@ class _IsFavorire:
 
 
 from typing import Union
-from ..core.error import Res
 from ..core.dataset import connect_readonly
 Entity = Union[_IsTweet, _IsFavorire]
 def _entities() -> Iterator[Res[Entity]]:
@@ -86,9 +85,17 @@ def _process_favorite_tweets(db) -> Iterator[Res[Entity]]:
 def _parse_tweet(row) -> Tweet:
     # TODO row['retweeter] if not empty, would be user's name and means retweet?
     # screen name would be the actual tweet's author
+
+    # ok so looks like it's tz aware..
+    # https://github.com/klinker24/talon-for-twitter-android/blob/c3b0612717ba3ea93c0cae6d907d7d86d640069e/app/src/main/java/com/klinker/android/twitter_l/data/sq_lite/FavoriteTweetsDataSource.java#L95
+    # uses https://docs.oracle.com/javase/7/docs/api/java/util/Date.html#getTime()
+    # and it's created here, so looks like it's properly parsed from the api
+    # https://github.com/Twitter4J/Twitter4J/blob/8376fade8d557896bb9319fb46e39a55b134b166/twitter4j-core/src/internal-json/java/twitter4j/ParseUtil.java#L69-L79
+    created_at = datetime.fromtimestamp(row['time'] / 1000, tz=pytz.utc)
+
     return Tweet(
         id_str=str(row['tweet_id']),
-        created_at=datetime.fromtimestamp(row['time'] / 1000),
+        created_at=created_at,
         screen_name=row['screen_name'],
         text=row['text'],
         # todo hmm text sometimes is trimmed with ellipsis? at least urls

my/twitter/twint.py

@@ -5,6 +5,7 @@ Twitter data (tweets and favorites). Uses [[https://github.com/twintproject/twin
 REQUIRES = ['dataset']
 
 from ..core.common import Paths
+from ..core.error import Res
 from dataclasses import dataclass
 from my.config import twint as user_config
 
@@ -21,10 +22,10 @@ config = make_config(twint)
 
 
 from datetime import datetime
-from typing import NamedTuple, Iterable, List
+from typing import NamedTuple, Iterator, List
 from pathlib import Path
 
-from ..core.common import get_files, LazyLogger, Json
+from ..core.common import get_files, LazyLogger, Json, datetime_aware
 from ..core.time import abbr_to_timezone
 
 log = LazyLogger(__name__)
@@ -42,7 +43,7 @@ class Tweet(NamedTuple):
         return self.row['id_str']
 
     @property
-    def created_at(self) -> datetime:
+    def created_at(self) -> datetime_aware:
         seconds = self.row['created_at'] / 1000
         tz_abbr = self.row['timezone']
         tz = abbr_to_timezone(tz_abbr)
@@ -97,20 +98,20 @@ def _get_db():
     return connect_readonly(db_path)
 
 
-def tweets() -> Iterable[Tweet]:
+def tweets() -> Iterator[Res[Tweet]]:
     db = _get_db()
     res = db.query(_QUERY.format(where='F.tweet_id IS NULL'))
     yield from map(Tweet, res)
 
 
-def likes() -> Iterable[Tweet]:
+def likes() -> Iterator[Res[Tweet]]:
     db = _get_db()
     res = db.query(_QUERY.format(where='F.tweet_id IS NOT NULL'))
     yield from map(Tweet, res)
 
 
-def stats():
-    from ..core import stat
+from ..core import stat, Stats
+def stats() -> Stats:
     return {
         **stat(tweets),
         **stat(likes),