moment/src/python/matrix_client.py

914 lines
30 KiB
Python
Raw Normal View History

import asyncio
2019-11-05 06:18:01 +11:00
import functools
import html
import io
import logging as log
import platform
2019-11-10 00:52:16 +11:00
import re
2019-10-31 02:10:40 +11:00
import traceback
from contextlib import suppress
2019-11-10 05:20:53 +11:00
from dataclasses import dataclass, field
2019-07-04 11:20:49 +10:00
from datetime import datetime
2019-08-28 17:54:53 +10:00
from functools import partial
2019-07-16 06:14:08 +10:00
from pathlib import Path
2019-11-05 06:18:01 +11:00
from typing import (
2019-11-06 00:19:48 +11:00
Any, BinaryIO, DefaultDict, Dict, Optional, Set, Tuple, Type, Union,
2019-11-05 06:18:01 +11:00
)
2019-11-10 00:52:16 +11:00
from urllib.parse import urlparse
from uuid import uuid4
import cairosvg
2019-10-29 06:27:36 +11:00
from PIL import Image as PILImage
from pymediainfo import MediaInfo
2019-11-05 06:18:01 +11:00
import nio
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
from . import __about__, utils
from .html_filter import HTML_FILTER
from .models.items import (
Account, Event, Member, Room, TypeSpecifier, Upload, UploadStatus,
)
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
from .models.model_store import ModelStore
from .pyotherside_events import AlertRequested
2019-11-05 06:18:01 +11:00
CryptDict = Dict[str, Any]
2019-11-09 06:32:12 +11:00
@dataclass
class MatrixError(Exception):
m_code: str = "M_UNKNOWN"
@classmethod
def from_nio(cls, response: nio.ErrorResponse) -> "MatrixError":
for subcls in cls.__subclasses__():
if subcls.m_code == response.status_code:
return subcls()
return cls(response.status_code)
@dataclass
class MatrixNotFound(MatrixError):
m_code: str = "M_NOT_FOUND"
@dataclass
class MatrixForbidden(MatrixError):
m_code: str = "M_FORBIDDEN"
2019-11-09 06:32:12 +11:00
2019-11-10 05:20:53 +11:00
@dataclass
class UserNotFound(Exception):
user_id: str = field()
@dataclass
class InvalidUserInContext(Exception):
user_id: str = field()
@dataclass
class UploadError(Exception):
http_code: Optional[int] = None
@dataclass
class UploadForbidden(UploadError):
http_code: Optional[int] = 403
@dataclass
class UploadTooLarge(UploadError):
http_code: Optional[int] = 413
2019-07-16 06:14:08 +10:00
@dataclass
class UneededThumbnail(Exception):
pass
@dataclass
class UnthumbnailableError(Exception):
exception: Optional[Exception] = None
2019-07-16 06:14:08 +10:00
class MatrixClient(nio.AsyncClient):
def __init__(self,
backend,
user: str,
homeserver: str = "https://matrix.org",
device_id: Optional[str] = None) -> None:
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
store = Path(backend.app.appdirs.user_data_dir) / "encryption"
store.mkdir(parents=True, exist_ok=True)
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
# TODO: ensure homeserver starts by a scheme://
2019-07-09 03:08:46 +10:00
# TODO: pass a ClientConfig with a pickle key
super().__init__(
homeserver = homeserver,
user = user,
device_id = device_id,
store_path = store,
config = nio.AsyncClientConfig(
max_timeout_retry_wait_time = 10,
),
2019-07-09 03:08:46 +10:00
)
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
from .backend import Backend
self.backend: Backend = backend
self.models: ModelStore = self.backend.models
2019-08-28 05:00:50 +10:00
self.sync_task: Optional[asyncio.Future] = None
self.first_sync_done: asyncio.Event = asyncio.Event()
self.first_sync_date: Optional[datetime] = None
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
self.send_locks: DefaultDict[str, asyncio.Lock] = \
DefaultDict(asyncio.Lock) # {room_id: lock}
self.past_tokens: Dict[str, str] = {} # {room_id: token}
self.fully_loaded_rooms: Set[str] = set() # {room_id}
self.loaded_once_rooms: Set[str] = set() # {room_id}
self.cleared_events_rooms: Set[str] = set() # {room_id}
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
self.local_echoes_uuid: Set[str] = set()
self.resolved_echoes: Dict[str, str] = {} # {event_id: echo_uuid}
self.skipped_events: DefaultDict[str, int] = DefaultDict(lambda: 0)
from .media_cache import MediaCache
cache_dir = Path(self.backend.app.appdirs.user_cache_dir)
self.media_cache = MediaCache(self, cache_dir)
from .nio_callbacks import NioCallbacks
self.nio_callbacks = NioCallbacks(self)
def __repr__(self) -> str:
return "%s(user_id=%r, homeserver=%r, device_id=%r)" % (
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
type(self).__name__, self.user_id, self.homeserver, self.device_id,
)
@property
def default_device_name(self) -> str:
os_ = f" on {platform.system()}".rstrip()
os_ = f"{os_} {platform.release()}".rstrip() if os_ != " on" else ""
return f"{__about__.__pretty_name__}{os_}"
async def login(self, password: str, device_name: str = "") -> None:
response = await super().login(
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
password, device_name or self.default_device_name,
)
if isinstance(response, nio.LoginError):
2019-08-17 06:30:18 +10:00
raise RuntimeError(response)
asyncio.ensure_future(self.start())
async def resume(self, user_id: str, token: str, device_id: str) -> None:
response = nio.LoginResponse(user_id, device_id, token)
await self.receive_response(response)
2019-10-31 02:10:40 +11:00
asyncio.ensure_future(self.start())
async def logout(self) -> None:
if self.sync_task:
self.sync_task.cancel()
with suppress(asyncio.CancelledError):
await self.sync_task
await super().logout()
await self.close()
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
async def start(self) -> None:
def on_profile_response(future) -> None:
resp = future.result()
if isinstance(resp, nio.ProfileGetResponse):
account = self.models[Account][self.user_id]
account.profile_updated = datetime.now()
account.display_name = resp.displayname or ""
account.avatar_url = resp.avatar_url or ""
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
ft = asyncio.ensure_future(self.backend.get_profile(self.user_id))
ft.add_done_callback(on_profile_response)
2019-10-31 02:10:40 +11:00
while True:
try:
await self.sync_forever(timeout=10_000)
except Exception:
trace = traceback.format_exc().rstrip()
log.error("Exception during sync, will restart:\n%s", trace)
await asyncio.sleep(2)
2019-07-05 16:45:30 +10:00
@property
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
def all_rooms(self) -> Dict[str, nio.MatrixRoom]:
2019-07-05 16:45:30 +10:00
return {**self.invited_rooms, **self.rooms}
async def send_text(self, room_id: str, text: str) -> None:
2019-07-20 10:55:52 +10:00
escape = False
if text.startswith("//") or text.startswith(r"\/"):
escape = True
text = text[1:]
if text.startswith("/me ") and not escape:
2019-10-29 04:49:55 +11:00
event_type = nio.RoomMessageEmote
2019-07-20 10:55:52 +10:00
text = text[len("/me "): ]
content = {"body": text, "msgtype": "m.emote"}
2019-07-22 07:41:43 +10:00
to_html = HTML_FILTER.from_markdown_inline(text, outgoing=True)
echo_body = HTML_FILTER.from_markdown_inline(text)
2019-07-20 10:55:52 +10:00
else:
2019-10-29 04:49:55 +11:00
event_type = nio.RoomMessageText
2019-07-20 10:55:52 +10:00
content = {"body": text, "msgtype": "m.text"}
2019-07-22 07:41:43 +10:00
to_html = HTML_FILTER.from_markdown(text, outgoing=True)
echo_body = HTML_FILTER.from_markdown(text)
2019-07-22 07:41:43 +10:00
if to_html not in (html.escape(text), f"<p>{html.escape(text)}</p>"):
content["format"] = "org.matrix.custom.html"
content["formatted_body"] = to_html
2019-07-04 11:20:49 +10:00
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
uuid = str(uuid4())
2019-10-29 04:49:55 +11:00
2019-10-29 06:27:36 +11:00
await self._local_echo(room_id, uuid, event_type, content=echo_body)
await self._send_message(room_id, uuid, content)
async def send_file(self, room_id: str, path: Union[Path, str]) -> None:
path = Path(path)
size = path.resolve().stat().st_size
encrypt = room_id in self.encrypted_rooms
upload_item = Upload(str(path), total_size=size)
self.models[Upload, room_id][upload_item.uuid] = upload_item
url, mime, crypt_dict = await self.upload_file(
path, upload_item, encrypt=encrypt,
)
await self.media_cache.create_media(url, path.read_bytes())
kind = (mime or "").split("/")[0]
thumb_url: str = ""
thumb_info: Dict[str, Any] = {}
2019-10-29 06:27:36 +11:00
content: dict = {
"body": path.name,
"info": {
"mimetype": mime,
"size": size,
2019-10-29 06:27:36 +11:00
},
}
if encrypt:
content["file"] = {"url": url, **crypt_dict}
else:
content["url"] = url
2019-10-29 06:27:36 +11:00
if kind == "image":
is_svg = mime == "image/svg+xml"
event_type = \
nio.RoomEncryptedImage if encrypt else nio.RoomMessageImage
2019-10-29 06:27:36 +11:00
content["msgtype"] = "m.image"
content["info"]["w"], content["info"]["h"] = (
utils.svg_dimensions(str(path)) if is_svg else
2019-10-29 06:27:36 +11:00
PILImage.open(path).size
)
2019-10-29 06:27:36 +11:00
try:
thumb_data, thumb_url, thumb_info, thumb_crypt_dict = \
await self.upload_thumbnail(
path, upload_item, is_svg=is_svg, encrypt=encrypt,
)
except (UneededThumbnail, UnthumbnailableError):
pass
else:
await self.media_cache.create_thumbnail(
thumb_url,
thumb_data,
content["info"]["w"],
content["info"]["h"],
)
if encrypt:
content["info"]["thumbnail_file"] = {
"url": thumb_url,
**thumb_crypt_dict,
}
else:
content["info"]["thumbnail_url"] = thumb_url
content["info"]["thumbnail_info"] = thumb_info
2019-10-29 06:27:36 +11:00
elif kind == "audio":
event_type = \
nio.RoomEncryptedAudio if encrypt else nio.RoomMessageAudio
2019-10-29 06:27:36 +11:00
content["msgtype"] = "m.audio"
content["info"]["duration"] = getattr(
MediaInfo.parse(path).tracks[0], "duration", 0,
) or 0
elif kind == "video":
event_type = \
nio.RoomEncryptedVideo if encrypt else nio.RoomMessageVideo
2019-10-29 06:27:36 +11:00
content["msgtype"] = "m.video"
tracks = MediaInfo.parse(path).tracks
content["info"]["duration"] = \
getattr(tracks[0], "duration", 0) or 0
content["info"]["w"] = max(
getattr(t, "width", 0) or 0 for t in tracks
)
content["info"]["h"] = max(
getattr(t, "height", 0) or 0 for t in tracks
)
else:
event_type = \
nio.RoomEncryptedFile if encrypt else nio.RoomMessageFile
2019-10-29 06:27:36 +11:00
content["msgtype"] = "m.file"
content["filename"] = path.name
del self.models[Upload, room_id][upload_item.uuid]
2019-10-29 06:27:36 +11:00
uuid = str(uuid4())
await self._local_echo(
room_id, uuid, event_type,
inline_content = path.name,
media_url = url,
media_title = path.name,
media_width = content["info"].get("w", 0),
media_height = content["info"].get("h", 0),
media_duration = content["info"].get("duration", 0),
media_size = content["info"]["size"],
media_mime = content["info"]["mimetype"],
thumbnail_url = thumb_url,
thumbnail_width = thumb_info.get("w", 0),
thumbnail_height = thumb_info.get("h", 0),
2019-10-29 06:27:36 +11:00
)
2019-10-29 04:49:55 +11:00
await self._send_message(room_id, uuid, content)
async def _local_echo(
2019-10-29 06:27:36 +11:00
self, room_id: str, uuid: str,
2019-10-29 04:49:55 +11:00
event_type: Type[nio.Event], **event_fields,
) -> None:
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
our_info = self.models[Member, room_id][self.user_id]
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
2019-10-29 04:49:55 +11:00
event = Event(
2019-08-29 03:29:49 +10:00
source = None,
client_id = f"echo-{uuid}",
event_id = "",
date = datetime.now(),
sender_id = self.user_id,
sender_name = our_info.display_name,
sender_avatar = our_info.avatar_url,
2019-08-29 03:29:49 +10:00
is_local_echo = True,
local_event_type = event_type,
2019-10-29 06:27:36 +11:00
**event_fields,
2019-07-04 11:20:49 +10:00
)
2019-10-29 04:49:55 +11:00
self.local_echoes_uuid.add(uuid)
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
for user_id in self.models[Account]:
if user_id in self.models[Member, room_id]:
2019-10-29 04:49:55 +11:00
self.models[Event, user_id, room_id][f"echo-{uuid}"] = event
self.models[Event, user_id, room_id].sync_now()
2019-07-04 11:20:49 +10:00
2019-10-29 04:49:55 +11:00
await self.set_room_last_event(room_id, event)
async def _send_message(self, room_id: str, uuid: str, content: dict,
) -> None:
2019-08-17 02:23:34 +10:00
async with self.send_locks[room_id]:
2019-07-09 03:08:46 +10:00
response = await self.room_send(
room_id = room_id,
message_type = "m.room.message",
content = content,
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
tx_id = uuid,
2019-07-09 03:08:46 +10:00
ignore_unverified_devices = True,
)
2019-07-05 12:25:06 +10:00
if isinstance(response, nio.RoomSendError):
log.error("Sending message failed: %s", response)
2019-07-04 11:20:49 +10:00
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
async def load_past_events(self, room_id: str) -> bool:
if room_id in self.fully_loaded_rooms or \
room_id in self.invited_rooms or \
room_id in self.cleared_events_rooms:
2019-07-05 16:45:30 +10:00
return False
2019-08-28 05:00:50 +10:00
await self.first_sync_done.wait()
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
2019-07-05 16:45:30 +10:00
response = await self.room_messages(
room_id = room_id,
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
start = self.past_tokens[room_id],
limit = 100 if room_id in self.loaded_once_rooms else 25,
2019-07-05 16:45:30 +10:00
)
if isinstance(response, nio.RoomMessagesError):
log.error("Loading past messages for room %s failed: %s",
room_id, response)
return True
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
self.loaded_once_rooms.add(room_id)
2019-07-05 16:45:30 +10:00
more_to_load = True
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
self.past_tokens[room_id] = response.end
2019-07-05 16:45:30 +10:00
for event in response.chunk:
if isinstance(event, nio.RoomCreateEvent):
self.fully_loaded_rooms.add(room_id)
more_to_load = False
2019-07-05 16:45:30 +10:00
for cb in self.event_callbacks:
if (cb.filter is None or isinstance(event, cb.filter)):
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
await cb.func(self.all_rooms[room_id], event)
2019-07-05 16:45:30 +10:00
return more_to_load
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
async def load_rooms_without_visible_events(self) -> None:
for room_id in self.models[Room, self.user_id]:
asyncio.ensure_future(
self._load_room_without_visible_events(room_id),
)
async def _load_room_without_visible_events(self, room_id: str) -> None:
events = self.models[Event, self.user_id, room_id]
more = True
while self.skipped_events[room_id] and not events and more:
more = await self.load_past_events(room_id)
2019-11-10 05:20:53 +11:00
async def new_direct_chat(self, invite: str, encrypt: bool = False) -> str:
if invite == self.user_id:
raise InvalidUserInContext(invite)
if isinstance(await self.get_profile(invite), nio.ProfileGetError):
raise UserNotFound(invite)
response = await super().room_create(
invite = [invite],
is_direct = True,
visibility = nio.RoomVisibility.private,
initial_state =
[nio.EnableEncryptionBuilder().as_dict()] if encrypt else [],
)
if isinstance(response, nio.RoomCreateError):
raise MatrixError.from_nio(response)
return response.room_id
async def new_group_chat(
2019-11-09 06:32:12 +11:00
self,
name: Optional[str] = None,
topic: Optional[str] = None,
public: bool = False,
encrypt: bool = False,
2019-11-09 06:32:12 +11:00
federate: bool = True,
) -> str:
response = await super().room_create(
name = name or None,
topic = topic or None,
2019-11-09 06:32:12 +11:00
federate = federate,
visibility =
nio.RoomVisibility.public if public else
nio.RoomVisibility.private,
initial_state =
[nio.EnableEncryptionBuilder().as_dict()] if encrypt else [],
2019-11-09 06:32:12 +11:00
)
if isinstance(response, nio.RoomCreateError):
raise MatrixError.from_nio(response)
2019-11-09 06:32:12 +11:00
return response.room_id
2019-11-10 00:52:16 +11:00
async def room_join(self, alias_or_id_or_url: str) -> str:
string = alias_or_id_or_url.strip()
if re.match(r"^https?://", string):
for part in urlparse(string).fragment.split("/"):
if re.match(r"^[#!].+:.+", part):
2019-11-10 00:52:16 +11:00
string = part
break
else:
raise ValueError(f"No alias or room id found in url {string}")
if not re.match(r"^[#!].+:.+", string):
raise ValueError("Not an alias or room id")
2019-11-10 00:52:16 +11:00
response = await super().join(string)
if isinstance(response, nio.JoinError):
raise MatrixError.from_nio(response)
2019-11-10 00:52:16 +11:00
return response.room_id
2019-11-09 06:32:12 +11:00
2019-07-08 12:19:17 +10:00
async def room_forget(self, room_id: str) -> None:
await super().room_leave(room_id)
2019-07-08 12:19:17 +10:00
await super().room_forget(room_id)
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
self.models[Room, self.user_id].pop(room_id, None)
self.models.pop((Event, self.user_id, room_id), None)
self.models.pop((Member, room_id), None)
2019-07-08 12:19:17 +10:00
2019-11-05 06:18:01 +11:00
async def encrypt_attachment(self, data: bytes) -> Tuple[bytes, CryptDict]:
func = functools.partial(
nio.crypto.attachments.encrypt_attachment,
data,
)
# Run in a separate thread
return await asyncio.get_event_loop().run_in_executor(None, func)
async def upload_thumbnail(
self,
path: Union[Path, str],
item: Optional[Upload] = None,
is_svg: bool = False,
encrypt: bool = False,
) -> Tuple[bytes, str, Dict[str, Any], CryptDict]:
png_modes = ("1", "L", "P", "RGBA")
try:
if is_svg:
svg_width, svg_height = utils.svg_dimensions(str(path))
thumb = PILImage.open(io.BytesIO(
cairosvg.svg2png(
url = str(path),
parent_width = svg_width,
parent_height = svg_height,
),
))
else:
thumb = PILImage.open(path)
small = thumb.width <= 800 and thumb.height <= 600
is_jpg_png = thumb.format in ("JPEG", "PNG")
jpgable_png = thumb.format == "PNG" and thumb.mode not in png_modes
if small and is_jpg_png and not jpgable_png and not is_svg:
raise UneededThumbnail()
if item:
item.status = UploadStatus.CreatingThumbnail
2019-10-30 20:34:32 +11:00
if not small:
thumb.thumbnail((800, 600), PILImage.LANCZOS)
2019-07-16 06:14:08 +10:00
with io.BytesIO() as out:
if thumb.mode in png_modes:
thumb.save(out, "PNG", optimize=True)
mime = "image/png"
else:
thumb.convert("RGB").save(out, "JPEG", optimize=True)
mime = "image/jpeg"
data = out.getvalue()
if encrypt:
if item:
item.status = UploadStatus.EncryptingThumbnail
2019-11-05 06:18:01 +11:00
data, crypt_dict = await self.encrypt_attachment(data)
upload_mime = "application/octet-stream"
else:
crypt_dict, upload_mime = {}, mime
if item:
item.status = UploadStatus.UploadingThumbnail
return (
data,
await self.upload(data, upload_mime, Path(path).name),
{
"w": thumb.width,
"h": thumb.height,
"mimetype": mime,
"size": len(data),
},
crypt_dict,
)
except OSError as err:
2019-10-30 20:36:54 +11:00
log.warning("Error when creating thumbnail: %s", err)
raise UnthumbnailableError(err)
async def upload_file(
self,
path: Union[Path, str],
item: Optional[Upload] = None,
encrypt: bool = False,
) -> Tuple[str, str, CryptDict]:
2019-07-16 06:14:08 +10:00
with open(path, "rb") as file:
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
mime = utils.guess_mime(file)
2019-07-16 06:14:08 +10:00
file.seek(0, 0)
2019-11-05 06:18:01 +11:00
data: Union[BinaryIO, bytes]
if encrypt:
if item:
item.status = UploadStatus.Encrypting
2019-11-05 06:18:01 +11:00
data, crypt_dict = await self.encrypt_attachment(file.read())
upload_mime = "application/octet-stream"
else:
data, crypt_dict, upload_mime = file, {}, mime
if item:
item.status = UploadStatus.Uploading
return (
await self.upload(data, upload_mime, Path(path).name),
mime,
crypt_dict,
)
async def upload(self, data, mime: str, filename: Optional[str] = None,
) -> str:
response = await super().upload(data, mime, filename)
2019-07-16 06:14:08 +10:00
if not isinstance(response, nio.ErrorResponse):
return response.content_uri
2019-07-16 06:14:08 +10:00
if response.status_code == 403:
raise UploadForbidden()
2019-07-16 06:14:08 +10:00
if response.status_code == 413:
raise UploadTooLarge()
2019-07-16 06:14:08 +10:00
raise UploadError(response.status_code)
2019-07-16 06:14:08 +10:00
async def set_avatar_from_file(self, path: Union[Path, str]) -> None:
2019-10-29 06:27:36 +11:00
# TODO: check if mime is image
await self.set_avatar((await self.upload_file(path))[0])
2019-07-16 06:14:08 +10:00
2019-08-29 01:42:52 +10:00
async def import_keys(self, infile: str, passphrase: str) -> None:
2019-08-28 17:54:53 +10:00
# Reimplemented until better solutions are worked on in nio
2019-08-29 01:42:52 +10:00
await self.clear_import_error()
2019-08-28 17:54:53 +10:00
loop = asyncio.get_event_loop()
2019-08-29 01:42:52 +10:00
account = self.models[Account][self.user_id]
2019-08-28 17:54:53 +10:00
import_keys = partial(self.olm.import_keys_static, infile, passphrase)
2019-08-29 01:42:52 +10:00
account.importing_key = 0
account.total_keys_to_import = -1 # preparing
2019-08-28 17:54:53 +10:00
try:
sessions = await loop.run_in_executor(None, import_keys)
except nio.EncryptionError as err: # XXX raise
2019-08-29 01:42:52 +10:00
account.import_error = (infile, passphrase, str(err))
return
2019-08-28 17:54:53 +10:00
account.total_keys_to_import = len(sessions)
for session in sessions:
if self.olm.inbound_group_store.add(session):
await loop.run_in_executor(
None, self.store.save_inbound_group_session, session,
)
account.importing_key += 1
account.importing_key = 0
account.total_keys_to_import = 0
await self.retry_decrypting_events()
2019-08-28 17:54:53 +10:00
2019-09-08 09:17:32 +10:00
async def export_keys(self, outfile: str, passphrase: str) -> None:
path = Path(outfile)
path.parent.mkdir(parents=True, exist_ok=True)
# The QML dialog asks the user if he wants to overwrite before this
if path.exists():
path.unlink()
2019-09-08 09:17:32 +10:00
await super().export_keys(outfile, passphrase)
2019-08-29 01:42:52 +10:00
async def clear_import_error(self) -> None:
self.models[Account][self.user_id].import_error = ("", "", "")
async def retry_decrypting_events(self) -> None:
for sync_id, model in self.models.items():
if not (isinstance(sync_id, tuple) and
sync_id[0:2] == (Event, self.user_id)):
continue
_, _, room_id = sync_id
for ev in model.values():
room = self.all_rooms[room_id]
if isinstance(ev.source, nio.MegolmEvent):
try:
decrypted = self.decrypt_event(ev.source)
if not decrypted:
raise nio.EncryptionError()
except nio.EncryptionError:
continue
for cb in self.event_callbacks:
if not cb.filter or isinstance(decrypted, cb.filter):
await asyncio.coroutine(cb.func)(room, decrypted)
async def clear_events(self, room_id: str) -> None:
self.cleared_events_rooms.add(room_id)
model = self.models[Event, self.user_id, room_id]
2019-09-09 01:49:47 +10:00
if model:
model.clear()
model.sync_now()
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
# Functions to register data into models
async def event_is_past(self, ev: Union[nio.Event, Event]) -> bool:
if not self.first_sync_date:
return True
if isinstance(ev, Event):
return ev.date < self.first_sync_date
date = datetime.fromtimestamp(ev.server_timestamp / 1000)
return date < self.first_sync_date
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
async def set_room_last_event(self, room_id: str, item: Event) -> None:
model = self.models[Room, self.user_id]
room = model[room_id]
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
if room.last_event is None:
room.last_event = item.serialized
if item.is_local_echo:
model.sync_now()
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
return
is_profile_ev = item.type_specifier == TypeSpecifier.profile_change
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
# If there were no better events available to show previously
prev_is_profile_ev = \
room.last_event["type_specifier"] == TypeSpecifier.profile_change
# If this is a profile event, only replace the currently shown one if
# it was also a profile event (we had nothing better to show).
if is_profile_ev and not prev_is_profile_ev:
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
return
# If this event is older than the currently shown one, only replace
# it if the previous was a profile event.
if item.date < room.last_event["date"] and not prev_is_profile_ev:
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
return
room.last_event = item.serialized
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
if item.is_local_echo:
model.sync_now()
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
async def register_nio_room(self, room: nio.MatrixRoom, left: bool = False,
) -> None:
# Add room
try:
last_ev = self.models[Room, self.user_id][room.room_id].last_event
except KeyError:
last_ev = None
inviter = getattr(room, "inviter", "") or ""
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
self.models[Room, self.user_id][room.room_id] = Room(
room_id = room.room_id,
display_name = room.display_name,
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
avatar_url = room.gen_avatar_url or "",
topic = HTML_FILTER.filter_inline(room.topic or ""),
inviter_id = inviter,
inviter_name = room.user_name(inviter) if inviter else "",
inviter_avatar =
(room.avatar_url(inviter) or "") if inviter else "",
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
left = left,
last_event = last_ev,
)
# Add the room members to the added room
new_dict = {
user_id: Member(
user_id = user_id,
display_name = room.user_name(user_id) # disambiguated
if member.display_name else "",
avatar_url = member.avatar_url or "",
typing = user_id in room.typing_users,
power_level = member.power_level,
) for user_id, member in room.users.items()
}
self.models[Member, room.room_id].update(new_dict)
async def get_member_name_avatar(self, room_id: str, user_id: str,
) -> Tuple[str, str]:
try:
item = self.models[Member, room_id][user_id]
except KeyError: # e.g. user is not anymore in the room
info = await self.backend.get_profile(user_id)
return (info.displayname or "", info.avatar_url or "") \
if isinstance(info, nio.ProfileGetResponse) else \
("", "")
else:
return (item.display_name, item.avatar_url)
async def register_nio_event(
self, room: nio.MatrixRoom, ev: nio.Event, **fields,
) -> None:
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
await self.register_nio_room(room)
sender_name, sender_avatar = \
await self.get_member_name_avatar(room.room_id, ev.sender)
target_id = getattr(ev, "state_key", "") or ""
target_name, target_avatar = \
await self.get_member_name_avatar(room.room_id, target_id) \
if target_id else ("", "")
# Create Event ModelItem
item = Event(
source = ev,
client_id = ev.event_id,
event_id = ev.event_id,
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
date = datetime.fromtimestamp(ev.server_timestamp / 1000),
sender_id = ev.sender,
sender_name = sender_name,
sender_avatar = sender_avatar,
target_id = target_id,
target_name = target_name,
target_avatar = target_avatar,
2019-09-14 12:34:20 +10:00
**fields,
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
)
# Add the Event to model
if ev.transaction_id in self.local_echoes_uuid:
self.resolved_echoes[ev.event_id] = ev.transaction_id
self.local_echoes_uuid.discard(ev.transaction_id)
item.client_id = f"echo-{ev.transaction_id}"
elif ev.sender in self.backend.clients:
client = self.backend.clients[ev.sender]
# Wait until our other account has no more pending local echoes,
# so that we can know if this event should replace an echo
# from that client by finding its ID in the resolved_echoes dict.
# Server only gives back the transaction ID to the original sender.
while client.local_echoes_uuid: # while there are pending echoes
await asyncio.sleep(0.1)
with suppress(KeyError):
item.client_id = f"echo-{client.resolved_echoes[ev.event_id]}"
elif not await self.event_is_past(ev):
AlertRequested()
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
self.models[Event, self.user_id, room.room_id][item.client_id] = item
Big performance refactoring & various improvements Instead of passing all sorts of events for the JS to handle and manually add to different data models, we now handle everything we can in Python. For any change, the python models send a sync event with their contents (no more than 4 times per second) to JS, and the QSyncable library's JsonListModel takes care of converting it to a QML ListModel and sending the appropriate signals. The SortFilterProxyModel library is not used anymore, the only case where we need to filter/sort something now is when the user interacts with the "Filter rooms" or "Filter members" fields. These cases are handled by a simple JS function. We now keep separated room and timeline models for different accounts, the previous approach of sharing all the data we could between accounts created a lot of complications (local echoes, decrypted messages replacing others, etc). The users's own account profile changes are now hidden in the timeline. On startup, if all events for a room were only own profile changes, more events will be loaded. Any kind of image format supported by Qt is now handled by the pyotherside image provider, instead of just PNG/JPG. SVGs which previously caused errors are supported as well. The typing members bar paddings/margins are fixed. The behavior of the avatar/"upload a profile picture" overlay is fixed. Config files read from disk are now cached (TODO: make them reloadable again). Pylint is not used anymore because of all its annoying false warnings and lack of understanding for dataclasses, it is replaced by flake8 with a custom config and various plugins. Debug mode is now considered on if the program was compiled with the right option, instead of taking an argument from CLI. When on, C++ will set a flag in the Window QML component. The loading screen is now unloaded after the UI is ready, where previously it just stayed in the background invisible and wasted CPU. The overall refactoring and improvements make us now able to handle rooms with thousand of members and no lazy-loading, where previously everything would freeze and simply scrolling up to load past events in any room would block the UI for a few seconds.
2019-08-11 22:01:22 +10:00
await self.set_room_last_event(room.room_id, item)
if item.sender_id == self.user_id:
self.models[Event, self.user_id, room.room_id].sync_now()