my.zulip: extract Server/Sender objects, experiment with normalised and denormalised objects

my/zulip/organization.py

@@ -20,48 +20,98 @@ def inputs() -> Sequence[Path]:
 
 
 from datetime import datetime
+
+
 @dataclass(frozen=True)
-class Message:
+class Server:
+    id: int
+    string_id: str
+    name: str
+
+
+@dataclass(frozen=True)
+class Sender:
+    id: int
+    # todo make optional?
+    full_name: str
+    email: str
+
+
+# from the data, seems that subjects are completely implicit and determined by name?
+# streams have ids (can extract from realm/zerver_stream), but unclear how to correlate messages/topics to streams?
+
+@dataclass(frozen=True)
+class _Message:
+    # todo hmm not sure what would be a good field order..
     id: int
     sent: datetime
+    # TODO hmm kinda unclear whether it uses UTC or not??
+    # https://github.com/zulip/zulip/blob/0c2e4eec200d986a9a020f3e9a651d27216e0e85/zerver/models.py#L3071-L3076
+    # it keeps it tz aware.. but not sure what happens after?
+    # https://github.com/zulip/zulip/blob/1dfddffc8dac744fd6a6fbfd937018074c8bb166/zproject/computed_settings.py#L151
     subject: str
-    sender: str
+    sender_id: int
+    server_id: int
     content: str  # TODO hmm, it keeps markdown, not sure how/whether it's worth to prettify at all?
     # TODO recipient??
     # todo keep raw item instead? not sure
 
 
-# TODO hmm kinda unclear whether it uses UTC or not??
+@dataclass(frozen=True)
-# https://github.com/zulip/zulip/blob/0c2e4eec200d986a9a020f3e9a651d27216e0e85/zerver/models.py#L3071-L3076
+class Message:
-# it keeps it tz aware.. but not sure what happens after?
+    id: int
-# https://github.com/zulip/zulip/blob/1dfddffc8dac744fd6a6fbfd937018074c8bb166/zproject/computed_settings.py#L151
+    sent: datetime
+    subject: str
+    sender: Sender
+    server: Server
+    content: str
 
 
+from typing import Union
 from itertools import count
 import json
 from ..core.error import Res
 from ..core.kompress import kopen, kexists
-# TODO check that it also works with unpacked dirs???
+# TODO cache it
-def messages() -> Iterator[Res[Message]]:
+def _entities() -> Iterator[Res[Union[Server, Sender, _Message]]]:
     # TODO hmm -- not sure if max lexicographically will actually be latest?
     last = max(inputs())
     no_suffix = last.name.split('.')[0]
 
+    # TODO check that it also works with unpacked dirs???
     with kopen(last, f'{no_suffix}/realm.json') as f:
         rj = json.load(f)
-    id2user: Dict[int, str] = {}
-    for j in rj['zerver_userprofile']:
-        id2user[j['id']] = j['full_name']
-    for j in rj['zerver_userprofile_crossrealm']:  # e.g. zulip bot
-        id2user[j['id']] = j['email']
 
-    def _parse_message(j: Json) -> Message:
+    [sj] = rj['zerver_realm']
+    server = Server(
+        id=sj['id'],
+        string_id=sj['string_id'],
+        name=sj['name'],
+    )
+    yield server
+
+    for j in rj['zerver_userprofile']:
+        yield Sender(
+            id=j['id'],
+            full_name=j['full_name'],
+            email=j['email'],
+        )
+
+    for j in rj['zerver_userprofile_crossrealm']:  # e.g. zulip bot
+        yield Sender(
+            id=j['id'],
+            full_name=j['email'], # doesn't seem to have anything
+            email=j['email'],
+        )
+
+    def _parse_message(j: Json) -> _Message:
         ds = j['date_sent']
-        return Message(
+        return _Message(
             id        = j['id'],
             sent      = datetime.fromtimestamp(ds),
             subject   = j['subject'],
-            sender  = id2user[j['sender']],
+            sender_id = j['sender'],
+            server_id = server.id,
             content   = j['content'],
         )
 
@@ -78,3 +128,30 @@ def messages() -> Iterator[Res[Message]]:
                 yield _parse_message(j)
             except Exception as e:
                 yield e
+
+
+def messages() -> Iterator[Res[Message]]:
+    id2sender: Dict[int, Sender] = {}
+    id2server: Dict[int, Server] = {}
+    for x in _entities():
+        if isinstance(x, Exception):
+            yield x
+            continue
+        if isinstance(x, Server):
+            id2server[x.id] = x
+            continue
+        if isinstance(x, Sender):
+            id2sender[x.id] = x
+            continue
+        if isinstance(x, _Message):
+            # TODO a bit copypasty... wonder if possible to mixin or something instead
+            yield Message(
+                id=x.id,
+                sent=x.sent,
+                subject=x.subject,
+                sender=id2sender[x.sender_id],
+                server=id2server[x.server_id],
+                content=x.content,
+            )
+            continue
+        assert False # should be unreachable