diff --git a/arclet/entari/__init__.py b/arclet/entari/__init__.py
index 5ddf75c..88f2f97 100644
--- a/arclet/entari/__init__.py
+++ b/arclet/entari/__init__.py
@@ -55,6 +55,8 @@
from .event import BaseEvent as BaseEvent
from .event import attr as attr
from .event import register_internal_event as register_internal_event
+from .event.api import SendRequest as SendRequest
+from .event.api import SendResponse as SendResponse
from .event.base import MessageCreatedEvent as MessageCreatedEvent
from .event.base import MessageEvent as MessageEvent
from .event.base import Reply as Reply
@@ -63,8 +65,6 @@
from .event.lifespan import Cleanup as Cleanup
from .event.lifespan import Ready as Ready
from .event.lifespan import Startup as Startup
-from .event.send import SendRequest as SendRequest
-from .event.send import SendResponse as SendResponse
from .filter import filter_ as filter_
from .localdata import local_data as local_data
from .message import MessageChain as MessageChain
@@ -91,4 +91,4 @@
WS = WebsocketsInfo
WH = WebhookInfo
-__version__ = "0.18.5"
+__version__ = "0.19.0rc2"
diff --git a/arclet/entari/builtins/auto_reload.py b/arclet/entari/builtins/auto_reload.py
index 7f99b03..406043e 100644
--- a/arclet/entari/builtins/auto_reload.py
+++ b/arclet/entari/builtins/auto_reload.py
@@ -1,6 +1,11 @@
+import ast
import asyncio
+import importlib
+import sys
from dataclasses import asdict
from pathlib import Path
+from traceback import format_exception_only
+from types import ModuleType
from arclet.letoderea import post, publish
from launart import Launart, Service, any_completed
@@ -16,7 +21,18 @@
from arclet.entari.config import BasicConfModel, EntariConfig, model_field
from arclet.entari.event.config import ConfigReload
from arclet.entari.logger import log
-from arclet.entari.plugin import PluginRole, find_plugin, find_plugin_by_file, unload_plugin_async
+from arclet.entari.plugin import (
+ Plugin,
+ PluginRole,
+ find_plugin,
+ find_plugin_by_file,
+ plugin_service,
+ reload_plugin,
+ reload_subplugin,
+ unload_plugin_async,
+)
+from arclet.entari.plugin.swap import classify, swap_functions
+from arclet.entari.utils import escape_tag
# declare_static()
loguru_logger.disable("watchfiles.main")
@@ -67,6 +83,29 @@ class Config(BasicConfModel):
logger = log.wrapper("[AutoReload]").opt(colors=True)
+def module_name_from_path(path: Path) -> str | None:
+ path = Path(path).resolve()
+
+ for entry in map(Path, sys.path):
+ try:
+ relative = path.relative_to(entry.resolve())
+ except ValueError:
+ continue
+
+ if relative.suffix == ".py":
+ relative = relative.with_suffix("")
+
+ parts = list(relative.parts)
+
+ # __init__.py 对应包本身
+ if parts[-1] == "__init__":
+ parts.pop()
+
+ return ".".join(parts)
+
+ return None
+
+
class Watcher(Service):
id = "entari.plugin.auto_reload/watcher"
@@ -81,36 +120,129 @@ def stages(self) -> set[Phase]:
def __init__(self, config: Config):
self.config = config
self.fail: dict[str, tuple[str, dict]] = {}
+ self._locks: dict[str, asyncio.Lock] = {}
super().__init__()
+ def _lock_for(self, plugin_id: str) -> asyncio.Lock:
+ if plugin_id not in self._locks:
+ self._locks[plugin_id] = asyncio.Lock()
+ return self._locks[plugin_id]
+
+ async def _reload(self, pid: str, cfg: dict) -> bool:
+ async with self._lock_for(pid):
+ if pid in plugin_service._subplugined:
+ return await reload_subplugin(pid, cfg)
+ return await reload_plugin(pid, cfg)
+
+ async def _reload_upstream(self, module_name: str) -> list[str]:
+ """刷新非插件上游模块,并重载依赖它的插件
+
+ 必须先刷新 sys.modules 中的内容, 否则依赖插件重载时 import 链命中 sys.modules 缓存,拿到的仍是旧模块。
+
+ Returns:
+ 实际重载的插件 id 列表(供同批次去重)。
+ """
+ if module_name in plugin_service.plugins or module_name in plugin_service._subplugined:
+ return []
+ mod = sys.modules.get(module_name)
+ if mod is None or not isinstance(mod, ModuleType):
+ return []
+ dependents = plugin_service.dependents_of(module_name, ensure=True)
+ if not dependents:
+ return []
+ plugins = ", ".join(sorted(dependents))
+ logger.debug(f"Reloading upstream module {module_name!r}, affected plugins: {plugins}")
+ try:
+ importlib.reload(mod)
+ except Exception as e:
+ logger.error(f"Failed to reload upstream module {module_name!r}: {e!r}")
+ return []
+ reloaded: list[str] = []
+ for dep_id in plugin_service.topo_dependents(set(dependents)):
+ async with self._lock_for(dep_id):
+ if await reload_plugin(dep_id):
+ reloaded.append(dep_id)
+ else:
+ logger.error(f"Failed to reload plugin {dep_id!r} after upstream module reload")
+ return reloaded
+
async def watch(self):
async for event in awatch(
*self.config.watch_dirs, debounce=self.config.debounce, step=self.config.step, watch_filter=PythonFilter()
):
+ pending: dict[str, tuple[str, Plugin]] = {}
+ failed: list[str] = []
+ upstream: set[str] = set()
for change in event:
if plugin := find_plugin_by_file(change[1]):
if plugin.is_static:
logger.info(f"Plugin {plugin.id!r} is static, ignored.")
continue
- logger.info(f"Detected change in {plugin.id!r}, reloading...")
- pid = plugin.id
- _conf = plugin.config.copy()
- del plugin
- await unload_plugin_async(pid)
- if plugin := load_plugin(pid, _conf):
- logger.info(f"Reloaded {plugin.id!r}")
- del plugin
- else:
- logger.error(f"Failed to reload {pid!r}")
- self.fail[change[1]] = (pid, _conf)
+ pending.setdefault(plugin.id, (change[1], plugin))
elif change[1] in self.fail:
- logger.info(f"Detected change in {change[1]!r} which failed to reload, retrying...")
- if plugin := load_plugin(*self.fail[change[1]]):
- logger.info(f"Reloaded {plugin.id!r}")
- del plugin
- del self.fail[change[1]]
+ failed.append(change[1])
+ elif module_name := module_name_from_path(Path(change[1])):
+ upstream.add(module_name)
+ reloaded: set[str] = set()
+ for module_name in upstream:
+ reloaded.update(await self._reload_upstream(module_name))
+ for pid, (file_path, plugin) in pending.items():
+ if pid in reloaded:
+ self.fail.pop(file_path, None)
+ continue
+ nodes: ast.Module | None = None
+ if (
+ plugin._inspect
+ and plugin.module.__file__
+ and (path := Path(file_path).resolve()) == Path(plugin.module.__file__).resolve()
+ ):
+ try:
+ nodes = ast.parse(path.read_bytes(), filename=path, type_comments=True)
+ except (OSError, SyntaxError) as e:
+ trace = escape_tag("".join(format_exception_only(e)))
+ logger.error(f"Change in {pid!r} occurred exception, skipped:\n{trace}")
+ continue
else:
- logger.error(f"Failed to reload {self.fail[change[1]][0]!r}")
+ if ast.dump(nodes, include_attributes=False) == plugin._inspect.dump:
+ logger.debug(f"Change in {pid!r} has no semantic difference, skipped.")
+ self.fail.pop(file_path, None)
+ continue
+ logger.info(f"Detected change in {pid!r}, reloading...")
+ if plugin._inspect and nodes:
+ changes = classify(plugin._inspect.nodes, nodes)
+ if changes is not None and swap_functions(plugin, nodes, changes):
+ if changes:
+ logger.info(
+ f"Hot swapped functions in {pid!r}: "
+ f"{', '.join(f'{change.qualname}' for change in changes)} "
+ f"successfully."
+ )
+ else:
+ logger.debug(f"Change in {pid!r} has no function-level diff, skipped.")
+ self.fail.pop(file_path, None)
+ continue
+ logger.debug(f"Hot swap functions in {pid!r} failed, falling back to full reload.")
+ _conf = plugin.config.copy()
+ del plugin
+ if await self._reload(pid, _conf):
+ logger.info(f"Reloaded {pid!r}")
+ self.fail.pop(file_path, None)
+ else:
+ logger.error(f"Failed to reload {pid!r}")
+ self.fail[file_path] = (pid, _conf)
+ pending.clear()
+ for file_path in failed:
+ if file_path not in self.fail:
+ continue
+ pid, _conf = self.fail[file_path]
+ if file_path not in self.fail:
+ continue
+ logger.info(f"Detected change in {file_path!r} which failed to reload, retrying...")
+ if await self._reload(pid, _conf):
+ logger.info(f"Reloaded {pid!r}")
+ del self.fail[file_path]
+ else:
+ logger.error(f"Failed to reload {pid!r}")
async def watch_config(self):
file = EntariConfig.instance.path.resolve()
@@ -196,10 +328,9 @@ async def watch_config(self):
_conf = plg.config.copy()
async def _():
- await unload_plugin_async(pid)
- if plg := load_plugin(plugin_name, new_conf):
- logger.info(f"Reloaded {plg.id!r}")
- del plg
+ if await self._reload(pid, new_conf):
+ logger.info(f"Reloaded {pid!r}")
+ self.fail.pop(plugin_file, None)
else:
logger.error(f"Failed to reload {plugin_name!r}")
self.fail[plugin_file] = (pid, _conf)
diff --git a/arclet/entari/command/__init__.py b/arclet/entari/command/__init__.py
index bb4494b..5a3aa3f 100644
--- a/arclet/entari/command/__init__.py
+++ b/arclet/entari/command/__init__.py
@@ -166,7 +166,15 @@ def wrapper(func: Callable[..., TM]) -> Subscriber[TM]:
self.subscribers[target.id] = target
def _remove(_):
- command_manager.delete(get_cmd(_))
+ _cmd = get_cmd(_)
+ # if returned, it means the subscriber is already staged reload.
+ try:
+ record = command_manager._resolve(_cmd._hash)
+ except KeyError:
+ pass
+ else:
+ if id(_cmd) == id(record):
+ command_manager.delete(_cmd)
self.trie[key].remove(target.id) # type: ignore
if not self.trie[key]:
self.trie.pop(key, None) # type: ignore
@@ -203,7 +211,15 @@ def _remove(_):
self.trie.setdefault(_key, []).append(target.id)
def _remove(_):
- command_manager.delete(get_cmd(_))
+ _cmd = get_cmd(_)
+ # if returned, it means the subscriber is already staged reload.
+ try:
+ record = command_manager._resolve(_cmd._hash)
+ except KeyError:
+ pass
+ else:
+ if id(_cmd) == id(record):
+ command_manager.delete(_cmd)
self.subscribers.pop(target.id, None)
for _key in keys:
self.trie[_key].remove(target.id) # type: ignore
diff --git a/arclet/entari/command/plugin.py b/arclet/entari/command/plugin.py
index 28fd654..155b75c 100644
--- a/arclet/entari/command/plugin.py
+++ b/arclet/entari/command/plugin.py
@@ -1,5 +1,6 @@
from __future__ import annotations
+import asyncio
from typing import Any
from typing_extensions import TypeVar, deprecated
@@ -28,7 +29,7 @@
async def _after_execute(ctx: Contexts, session: Session | None = None):
- result = ctx[RESULT]
+ result: str | MessageChain | _ExitException | None = ctx[RESULT]
event = ctx[EVENT]
if result is not None:
if isinstance(result, _ExitException):
@@ -55,7 +56,7 @@ def assign(self, path: str, value: Any = _seminal, or_not: bool = False, priorit
class AlconnaPluginDispatcher(PluginDispatcher[T]):
def __init__(self, plugin: Plugin, command: Alconna, need_reply_me: bool = False, need_notice_me: bool = False, use_config_prefix: bool = True, block: bool = True, skip_for_unmatch: bool = True): # noqa: E501
plugin._extra.setdefault("commands", []).append((command.prefixes, command.command))
- self.cache = LRU(10)
+ self.cache: "LRU[str, asyncio.Future]" = LRU(10) # noqa: UP037
self.supplier = AlconnaSuppiler(command, self.cache, block, skip_for_unmatch)
super().__init__(plugin, MessageCreatedEvent, command.path)
plugin.collect(
@@ -68,7 +69,15 @@ def __init__(self, plugin: Plugin, command: Alconna, need_reply_me: bool = False
@plugin.collect
def dispose():
- command_manager.delete(self.supplier.cmd)
+ _cmd = self.supplier.cmd
+ # if returned, it means the subscriber is already staged reload.
+ try:
+ record = command_manager._resolve(_cmd._hash)
+ except KeyError:
+ pass
+ else:
+ if id(_cmd) == id(record):
+ command_manager.delete(_cmd)
del self.supplier.cmd
del self.supplier
diff --git a/arclet/entari/core.py b/arclet/entari/core.py
index ce2cc35..1dd2c90 100644
--- a/arclet/entari/core.py
+++ b/arclet/entari/core.py
@@ -46,10 +46,10 @@
ITEM_SESSION,
ITEM_USER,
)
+from .event.api import SendResponse
from .event.base import MessageCreatedEvent, event_parse
from .event.config import ConfigReload
from .event.lifespan import AccountUpdate
-from .event.send import SendResponse
from .localdata import local_data
from .logger import apply_log_save, enable_rich_except, log
from .message import MessageChain
diff --git a/arclet/entari/event/api.py b/arclet/entari/event/api.py
new file mode 100644
index 0000000..a5c392c
--- /dev/null
+++ b/arclet/entari/event/api.py
@@ -0,0 +1,107 @@
+from dataclasses import dataclass
+from typing import TYPE_CHECKING, Any
+
+from arclet.letoderea import Contexts, Result, define, provide
+from satori import ChannelType
+from satori.client import Account
+from satori.exception import ActionFailed
+from satori.model import Channel, MessageObject
+
+from ..const import ITEM_ACCOUNT, ITEM_CHANNEL, ITEM_MESSAGE_CONTENT, ITEM_SESSION
+from ..message import MessageChain
+
+if TYPE_CHECKING:
+ from ..session import Session
+
+
+@dataclass
+class SendRequest:
+ account: Account
+ channel: str
+ message: MessageChain
+ session: "Session | None" = None
+
+ def check_result(self, value) -> Result[bool | MessageChain] | None:
+ if isinstance(value, bool | MessageChain):
+ return Result(value)
+
+
+before_send_pub = define(SendRequest, name="entari.event/before_send")
+
+
+@before_send_pub.gather
+async def send_req_gather(req: SendRequest, context: Contexts):
+ context[ITEM_ACCOUNT] = req.account
+ context[ITEM_MESSAGE_CONTENT] = req.message
+ if req.session:
+ context[ITEM_SESSION] = req.session
+ context[ITEM_CHANNEL] = req.session.channel
+ else:
+ try:
+ context[ITEM_CHANNEL] = await req.account.channel_get(req.channel)
+ except ActionFailed:
+ context[ITEM_CHANNEL] = Channel(
+ req.channel, ChannelType.DIRECT if req.channel.startswith("private:") else ChannelType.TEXT
+ )
+
+
+@dataclass
+class SendResponse:
+ account: Account
+ channel: str
+ message: MessageChain
+ result: list[MessageObject]
+ session: "Session | None" = None
+
+
+send_pub = define(SendResponse, name="entari.event/after_send")
+send_pub.providers.append(provide(list[MessageObject], call="$resp_result"))
+
+
+@send_pub.gather
+async def send_resp_gather(resp: SendResponse, context: Contexts):
+ context[ITEM_ACCOUNT] = resp.account
+ context[ITEM_MESSAGE_CONTENT] = resp.message
+ context["$resp_result"] = resp.result
+ if resp.session:
+ context[ITEM_SESSION] = resp.session
+ context[ITEM_CHANNEL] = resp.session.channel
+ else:
+ try:
+ context[ITEM_CHANNEL] = await resp.account.channel_get(resp.channel)
+ except ActionFailed:
+ context[ITEM_CHANNEL] = Channel(
+ resp.channel, ChannelType.DIRECT if resp.channel.startswith("private:") else ChannelType.TEXT
+ )
+
+
+@dataclass
+class APIRequest:
+ account: Account
+ name: str
+ params: dict[str, Any]
+
+
+before_api_pub = define(APIRequest, name="entari.event/before_api_call")
+
+
+@before_api_pub.gather
+async def call_req_gather(req: APIRequest, context: Contexts):
+ context[ITEM_ACCOUNT] = req.account
+
+
+@dataclass
+class APIResponse:
+ account: Account
+ name: str
+ params: dict[str, Any]
+ success: bool
+ result: Any
+
+
+after_api_pub = define(APIResponse, name="entari.event/after_api_call")
+
+
+@after_api_pub.gather
+async def call_resp_gather(resp: APIResponse, context: Contexts):
+ context[ITEM_ACCOUNT] = resp.account
diff --git a/arclet/entari/event/base.py b/arclet/entari/event/base.py
index 60261eb..f3234da 100644
--- a/arclet/entari/event/base.py
+++ b/arclet/entari/event/base.py
@@ -53,7 +53,7 @@ def _is_notice_me(message: MessageChain, account: Account):
def _remove_notice_me(message: MessageChain, account: Account):
- message = message.copy()
+ message = message.fork()
message.pop(0)
if _is_notice_me(message, account):
message.pop(0)
@@ -316,7 +316,7 @@ def __init__(self, account: Account, origin: OriginEvent):
super().__init__(account, origin)
self.content = MessageChain(self.message.message)
if self.content.has(Quote):
- self.quote = self.content.get(Quote, 1)[0]
+ self.quote = self.content.get_first(Quote)
self.content = self.content.exclude(Quote)
async def gather(self, context: Contexts):
diff --git a/arclet/entari/event/send.py b/arclet/entari/event/send.py
index de8991e..ebbb250 100644
--- a/arclet/entari/event/send.py
+++ b/arclet/entari/event/send.py
@@ -1,75 +1,10 @@
-from dataclasses import dataclass
-from typing import TYPE_CHECKING
+from warnings import warn
-from arclet.letoderea import Contexts, Result, define, provide
-from satori import ChannelType
-from satori.client import Account
-from satori.exception import ActionFailed
-from satori.model import Channel, MessageObject
+warn(
+ "arclet.entari.event.send is deprecated, please use arclet.entari.event.api instead",
+ DeprecationWarning,
+ stacklevel=2,
+)
-from ..const import ITEM_ACCOUNT, ITEM_CHANNEL, ITEM_MESSAGE_CONTENT, ITEM_SESSION
-from ..message import MessageChain
-
-if TYPE_CHECKING:
- from ..session import Session
-
-
-@dataclass
-class SendRequest:
- account: Account
- channel: str
- message: MessageChain
- session: "Session | None" = None
-
- def check_result(self, value) -> Result[bool | MessageChain] | None:
- if isinstance(value, bool | MessageChain):
- return Result(value)
-
-
-before_send_pub = define(SendRequest, name="entari.event/before_send")
-
-
-@before_send_pub.gather
-async def req_gather(req: SendRequest, context: Contexts):
- context[ITEM_ACCOUNT] = req.account
- context[ITEM_MESSAGE_CONTENT] = req.message
- if req.session:
- context[ITEM_SESSION] = req.session
- context[ITEM_CHANNEL] = req.session.channel
- else:
- try:
- context[ITEM_CHANNEL] = await req.account.channel_get(req.channel)
- except ActionFailed:
- context[ITEM_CHANNEL] = Channel(
- req.channel, ChannelType.DIRECT if req.channel.startswith("private:") else ChannelType.TEXT
- )
-
-
-@dataclass
-class SendResponse:
- account: Account
- channel: str
- message: MessageChain
- result: list[MessageObject]
- session: "Session | None" = None
-
-
-send_pub = define(SendResponse, name="entari.event/after_send")
-send_pub.providers.append(provide(list[MessageObject], call="$resp_result"))
-
-
-@send_pub.gather
-async def resp_gather(resp: SendResponse, context: Contexts):
- context[ITEM_ACCOUNT] = resp.account
- context[ITEM_MESSAGE_CONTENT] = resp.message
- context["$resp_result"] = resp.result
- if resp.session:
- context[ITEM_SESSION] = resp.session
- context[ITEM_CHANNEL] = resp.session.channel
- else:
- try:
- context[ITEM_CHANNEL] = await resp.account.channel_get(resp.channel)
- except ActionFailed:
- context[ITEM_CHANNEL] = Channel(
- resp.channel, ChannelType.DIRECT if resp.channel.startswith("private:") else ChannelType.TEXT
- )
+from .api import SendRequest as SendRequest # noqa: F401
+from .api import SendResponse as SendResponse # noqa: F401
diff --git a/arclet/entari/filter/__init__.py b/arclet/entari/filter/__init__.py
index dc683a7..039f493 100644
--- a/arclet/entari/filter/__init__.py
+++ b/arclet/entari/filter/__init__.py
@@ -1,18 +1,19 @@
-import asyncio
import inspect
from collections.abc import Awaitable, Callable
-from datetime import datetime
from typing import Final, TypeAlias
from typing_extensions import ParamSpec
-from arclet.letoderea import STOP, Propagator, enter_if, propagate
-from arclet.letoderea.utils import TCallable
+from arclet.letoderea import enter_if
from tarina import is_coroutinefunction
-from ..config import EntariConfig
-from ..message import MessageChain
from ..session import Session
from . import common
+from .limit import interval as interval
+from .limit import semaphore as semaphore
+from .message import endswith as endswith
+from .message import startswith as startswith
+from .permission import admins as admins
+from .permission import superusers as superusers
_SessionFilter: TypeAlias = Callable[[Session], bool] | Callable[[Session], Awaitable[bool]]
@@ -62,117 +63,3 @@ async def _(*args, _func=func, **kwargs):
filter_: Final[_Filter] = _Filter()
F = filter_
-
-
-class interval(Propagator):
- def __init__(self, value: float, limit_prompt: str | MessageChain | None = None, priority: int = 80):
- self.success = True
- self.value = value
- self.priority = priority
- self.limit_prompt = limit_prompt
- self.last_times: dict[str, datetime] = {}
-
- async def before(self, session: Session | None = None):
- session_id = (
- "$global" if not session else f"{session.account.platform}/{session.account.self_id}/{session.channel.id}"
- )
- last_time = self.last_times.get(session_id, None)
- if not last_time:
- return
- self.success = (datetime.now() - last_time).total_seconds() > self.value
- if not self.success:
- if session and self.limit_prompt:
- await session.send(self.limit_prompt)
- return STOP
-
- async def after(self, session: Session | None = None):
- session_id = (
- "$global" if not session else f"{session.account.platform}/{session.account.self_id}/{session.channel.id}"
- )
- self.last_times[session_id] = datetime.now()
-
- def compose(self):
- yield self.before, True, self.priority
- yield self.after, False, self.priority
-
- def __call__(self, func: TCallable) -> TCallable:
- return propagate(self)(func)
-
-
-class semaphore(Propagator):
- def __init__(self, count: int, limit_prompt: str | MessageChain | None = None, priority: int = 80):
- self.count = count
- self.limit_prompt = limit_prompt
- self.priority = priority
- self.semaphores: dict[str, asyncio.Semaphore] = {}
-
- async def before(self, session: Session | None = None):
- session_id = (
- "$global" if not session else f"{session.account.platform}/{session.account.self_id}/{session.channel.id}"
- )
- if session_id not in self.semaphores:
- self.semaphores[session_id] = asyncio.Semaphore(self.count)
- if not await self.semaphores[session_id].acquire():
- if session and self.limit_prompt:
- await session.send(self.limit_prompt)
- return STOP
-
- async def after(self, session: Session | None = None):
- session_id = (
- "$global" if not session else f"{session.account.platform}/{session.account.self_id}/{session.channel.id}"
- )
- if session_id not in self.semaphores:
- self.semaphores[session_id] = asyncio.Semaphore(self.count)
- self.semaphores[session_id].release()
-
- def compose(self):
- yield self.before, True, self.priority
- yield self.after, False, self.priority
-
- def __call__(self, func: TCallable) -> TCallable:
- return propagate(self)(func)
-
-
-class superusers(Propagator):
-
- async def check(self, session: Session | None = None):
- if not session:
- return STOP
- config = EntariConfig.instance.basic.superusers
- if session.account.platform not in config:
- return STOP
- if not session.event.user:
- return STOP
- if session.event.user.id not in config[session.account.platform]:
- return STOP
-
- def compose(self):
- yield self.check, True, 50
-
- def __call__(self, func: TCallable) -> TCallable:
- return propagate(self)(func)
-
-
-class admins(Propagator):
-
- async def check(self, session: Session | None = None):
- if not session:
- return STOP
- if session.event.member and session.event.member.roles:
- for role in session.event.member.roles:
- if any(keyword in role.id.lower() for keyword in ("admin", "administrator", "owner")):
- return
- config = EntariConfig.instance.basic.superusers
- if (
- session.account.platform in config
- and session.event.user
- and session.event.user.id in config[session.account.platform]
- ):
- return
- return STOP
-
- def compose(self):
- yield self.check, True, 50
-
- def __call__(self, func: TCallable) -> TCallable:
- return propagate(self)(func)
diff --git a/arclet/entari/filter/limit.py b/arclet/entari/filter/limit.py
new file mode 100644
index 0000000..5df1890
--- /dev/null
+++ b/arclet/entari/filter/limit.py
@@ -0,0 +1,77 @@
+import asyncio
+from datetime import datetime
+
+from arclet.letoderea import STOP, Propagator, propagate
+from arclet.letoderea.utils import TCallable
+
+from ..message import MessageChain
+from ..session import Session
+
+
+class interval(Propagator):
+ def __init__(self, value: float, limit_prompt: str | MessageChain | None = None, priority: int = 80):
+ self.success = True
+ self.value = value
+ self.priority = priority
+ self.limit_prompt = limit_prompt
+ self.last_times: dict[str, datetime] = {}
+
+ async def before(self, session: Session | None = None):
+ session_id = (
+ "$global" if not session else f"{session.account.platform}/{session.account.self_id}/{session.channel.id}"
+ )
+ last_time = self.last_times.get(session_id, None)
+ if not last_time:
+ return
+ self.success = (datetime.now() - last_time).total_seconds() > self.value
+ if not self.success:
+ if session and self.limit_prompt:
+ await session.send(self.limit_prompt)
+ return STOP
+
+ async def after(self, session: Session | None = None):
+ session_id = (
+ "$global" if not session else f"{session.account.platform}/{session.account.self_id}/{session.channel.id}"
+ )
+ self.last_times[session_id] = datetime.now()
+
+ def compose(self):
+ yield self.before, True, self.priority
+ yield self.after, False, self.priority
+
+ def __call__(self, func: TCallable) -> TCallable:
+ return propagate(self)(func)
+
+
+class semaphore(Propagator):
+ def __init__(self, count: int, limit_prompt: str | MessageChain | None = None, priority: int = 80):
+ self.count = count
+ self.limit_prompt = limit_prompt
+ self.priority = priority
+ self.semaphores: dict[str, asyncio.Semaphore] = {}
+
+ async def before(self, session: Session | None = None):
+ session_id = (
+ "$global" if not session else f"{session.account.platform}/{session.account.self_id}/{session.channel.id}"
+ )
+ if session_id not in self.semaphores:
+ self.semaphores[session_id] = asyncio.Semaphore(self.count)
+ if not await self.semaphores[session_id].acquire():
+ if session and self.limit_prompt:
+ await session.send(self.limit_prompt)
+ return STOP
+
+ async def after(self, session: Session | None = None):
+ session_id = (
+ "$global" if not session else f"{session.account.platform}/{session.account.self_id}/{session.channel.id}"
+ )
+ if session_id not in self.semaphores:
+ self.semaphores[session_id] = asyncio.Semaphore(self.count)
+ self.semaphores[session_id].release()
+
+ def compose(self):
+ yield self.before, True, self.priority
+ yield self.after, False, self.priority
+
+ def __call__(self, func: TCallable) -> TCallable:
+ return propagate(self)(func)
diff --git a/arclet/entari/filter/message.py b/arclet/entari/filter/message.py
new file mode 100644
index 0000000..c3f3cda
--- /dev/null
+++ b/arclet/entari/filter/message.py
@@ -0,0 +1,216 @@
+import re
+from typing import Any
+
+from arclet.letoderea import STOP, Contexts, Propagator, deref, propagate, provide
+from arclet.letoderea.utils import TCallable
+from nepattern import ANY, BasePattern, MatchMode, parser
+from satori import Text
+from tarina import Empty
+
+from ..const import ITEM_MESSAGE_CONTENT
+from ..message import MessageChain
+
+
+def _prefixed(pat: BasePattern):
+ if pat.mode not in (MatchMode.REGEX_MATCH, MatchMode.REGEX_CONVERT):
+ return pat
+ new_pat = pat.copy()
+ new_pat.regex_pattern = re.compile(f"^{new_pat.pattern}")
+ return new_pat
+
+
+def _suffixed(pat: BasePattern):
+ if pat.mode not in (MatchMode.REGEX_MATCH, MatchMode.REGEX_CONVERT):
+ return pat
+ new_pat = pat.copy()
+ new_pat.regex_pattern = re.compile(f"{new_pat.pattern}$")
+ return new_pat
+
+
+class startswith(Propagator):
+ def __init__(self, prefix: Any, include: bool = False, bind: str | None = None, priority: int = 80):
+ """
+ 前缀匹配
+
+ Args:
+ prefix: 需要匹配的前缀, 支持格式有 a|b , ['a', At(...)] 等
+ include: 指示消息链是否仅返回前缀被匹配的部分, 默认为 False
+ bind: 指定注入返回值的参数名称,未指定则注入到所有的 MessageChain 参数中
+ priority: 优先级
+ """
+ self.prefix = prefix
+ self.priority = priority
+ self.include = include
+ self.bind = bind
+
+ pattern = BasePattern(prefix, mode=MatchMode.REGEX_MATCH) if isinstance(prefix, str) else parser(prefix)
+ if pattern in (ANY, Empty):
+ raise ValueError(prefix)
+ self.pattern = _prefixed(pattern)
+
+ def providers(self):
+ if self.bind:
+ return [provide(MessageChain, self.bind, call=f"$startswith_{self.bind}", priority=4)]
+ return []
+
+ async def before(self, ctx: Contexts, message: MessageChain):
+ message = message.fork()
+ if message:
+ elem = message[0]
+ if isinstance(elem, Text) and (res := self.pattern.validate(elem.text)).success:
+ if self.include:
+ message = MessageChain(Text(str(res.value())))
+ else:
+ message[0] = Text(elem.text[len(str(res.value())) :].lstrip())
+ elif self.pattern.validate(elem).success:
+ if self.include:
+ message = MessageChain(elem)
+ else:
+ message.remove(elem)
+ else:
+ return STOP
+ if self.bind:
+ return {f"$startswith_{self.bind}": message}
+ if ITEM_MESSAGE_CONTENT in ctx:
+ return {ITEM_MESSAGE_CONTENT: message}
+ return {"$message": message}
+
+ def compose(self):
+ yield self.before, True, self.priority
+
+ def __call__(self, func: TCallable) -> TCallable:
+ return propagate(self)(func)
+
+
+class endswith(Propagator):
+ def __init__(self, suffix: Any, include: bool = False, bind: str | None = None, priority: int = 80):
+ """
+ 后缀匹配
+
+ Args:
+ suffix: 需要匹配的后缀, 支持格式有 a|b , ['a', At(...)] 等
+ include: 指示消息链是否仅返回后缀被匹配的部分, 默认为 False
+ bind: 指定注入返回值的参数名称,未指定则注入到所有的 MessageChain 参数中
+ priority: 优先级
+ """
+ self.suffix = suffix
+ self.priority = priority
+ self.include = include
+ self.bind = bind
+
+ pattern = BasePattern(suffix, mode=MatchMode.REGEX_MATCH) if isinstance(suffix, str) else parser(suffix)
+ if pattern in (ANY, Empty):
+ raise ValueError(suffix)
+ self.pattern = _suffixed(pattern)
+
+ def providers(self):
+ if self.bind:
+ return [provide(MessageChain, self.bind, call=f"$endswith_{self.bind}", priority=4)]
+ return []
+
+ async def before(self, ctx: Contexts, message: MessageChain):
+ message = message.fork()
+ if message:
+ elem = message[-1]
+ if isinstance(elem, Text) and (res := self.pattern.validate(elem.text)).success:
+ if self.include:
+ message = MessageChain(Text(str(res.value())))
+ else:
+ message[-1] = Text(elem.text[: elem.text.rfind(str(res.value()))].rstrip())
+ elif self.pattern.validate(elem).success:
+ if self.include:
+ message = MessageChain(elem)
+ else:
+ message.remove(elem)
+ else:
+ return STOP
+ if self.bind:
+ return {f"$endswith_{self.bind}": message}
+ if ITEM_MESSAGE_CONTENT in ctx:
+ return {ITEM_MESSAGE_CONTENT: message}
+ return {"$message": message}
+
+ def compose(self):
+ yield self.before, True, self.priority
+
+ def __call__(self, func: TCallable) -> TCallable:
+ return propagate(self)(func)
+
+
+class fullmatch(Propagator):
+ def __init__(
+ self, pattern: str | tuple[str, ...], ignorecase: bool = False, bind: str = "fullmatch", priority: int = 80
+ ):
+ """
+ 完全匹配
+
+ Args:
+ pattern: 指定消息全匹配字符串元组
+ ignorecase: 是否忽略大小写, 默认为 False
+ bind: 指定注入返回值的参数名称,默认为 "fullmatch"
+ priority: 优先级
+ """
+ if isinstance(pattern, str):
+ pattern = (pattern,)
+ self.pattern = tuple(map(str.casefold, pattern)) if ignorecase else pattern
+ self.ignorecase = ignorecase
+ self.priority = priority
+ self.bind = bind
+
+ def providers(self):
+ if self.bind:
+ return [provide(str, self.bind, call=f"$fullmatch_{self.bind}", priority=4)]
+ return []
+
+ async def before(self, ctx: Contexts, message: MessageChain):
+ text = message.extract_plain_text()
+ if not text:
+ return STOP
+ text = text.casefold() if self.ignorecase else text
+ if text in self.pattern:
+ return {f"$fullmatch_{self.bind}": text}
+ return STOP
+
+ def compose(self):
+ yield self.before, True, self.priority
+
+ def __call__(self, func: TCallable) -> TCallable:
+ return propagate(self)(func)
+
+
+class regexmatch(Propagator):
+ def __init__(self, pattern: str, flags: int | re.RegexFlag = 0, priority: int = 80):
+ """
+ 正则匹配,注意正则表达式匹配使用 search 而非 match,如需从头匹配请使用 `r"^xxx"` 来确保匹配开头
+
+ Args:
+ pattern: 需要匹配的正则表达式
+ flags: 正则匹配标志, 默认为 0
+ priority: 优先级
+ """
+ self.pattern = re.compile(pattern, flags)
+ self.priority = priority
+
+ def providers(self):
+ return [provide(re.Match, call="$regexmatch", priority=4)]
+
+ async def before(self, ctx: Contexts, message: MessageChain):
+ text = message.extract_plain_text()
+ if not text:
+ return STOP
+ if matched := self.pattern.search(text):
+ return {"$regexmatch": matched}
+ return STOP
+
+ def compose(self):
+ yield self.before, True, self.priority
+
+ def __call__(self, func: TCallable) -> TCallable:
+ return propagate(self)(func)
+
+
+def regex_origin():
+ return deref(re.Match)
+
+
+__all__ = ["startswith", "endswith", "fullmatch", "regexmatch", "regex_origin"]
diff --git a/arclet/entari/filter/permission.py b/arclet/entari/filter/permission.py
new file mode 100644
index 0000000..d337b74
--- /dev/null
+++ b/arclet/entari/filter/permission.py
@@ -0,0 +1,50 @@
+from arclet.letoderea import STOP, Propagator, propagate
+from arclet.letoderea.utils import TCallable
+
+from ..config import EntariConfig
+from ..session import Session
+
+
+class superusers(Propagator):
+
+ async def check(self, session: Session | None = None):
+ if not session:
+ return STOP
+ config = EntariConfig.instance.basic.superusers
+ if session.account.platform not in config:
+ return STOP
+ if not session.event.user:
+ return STOP
+ if session.event.user.id not in config[session.account.platform]:
+ return STOP
+
+ def compose(self):
+ yield self.check, True, 50
+
+ def __call__(self, func: TCallable) -> TCallable:
+ return propagate(self)(func)
+
+
+class admins(Propagator):
+
+ async def check(self, session: Session | None = None):
+ if not session:
+ return STOP
+ if session.event.member and session.event.member.roles:
+ for role in session.event.member.roles:
+ if any(keyword in role.id.lower() for keyword in ("admin", "administrator", "owner")):
+ return
+ config = EntariConfig.instance.basic.superusers
+ if (
+ session.account.platform in config
+ and session.event.user
+ and session.event.user.id in config[session.account.platform]
+ ):
+ return
+ return STOP
+
+ def compose(self):
+ yield self.check, True, 50
+
+ def __call__(self, func: TCallable) -> TCallable:
+ return propagate(self)(func)
diff --git a/arclet/entari/message.py b/arclet/entari/message.py
index 3b3c02e..0793e60 100644
--- a/arclet/entari/message.py
+++ b/arclet/entari/message.py
@@ -1,6 +1,6 @@
from __future__ import annotations
-from collections.abc import Awaitable, Callable, Iterable, Sequence
+from collections.abc import Awaitable, Callable, Iterable, Iterator, MutableSequence, Sequence
from copy import deepcopy
from dataclasses import dataclass
from typing import TYPE_CHECKING, Any, TypeAlias, TypeVar, Union, overload
@@ -27,33 +27,20 @@
MessageContainer = Union[str, Element, Sequence["MessageContainer"], "MessageChain[Element]"]
-class MessageChain(list[TE]):
- """消息序列
-
- Args:
- message: 消息内容
- """
+class MessageChain(MutableSequence[TE]):
+ """消息链, 被用于承载整个消息内容的数据结构, 包含有一有序列表, 包含有继承了 Element 的各式类实例."""
@overload
def __init__(self): ...
@overload
- def __init__(self: MessageChain[Text], message: str): ...
-
- @overload
- def __init__(self, message: TE): ...
-
- @overload
- def __init__(self: MessageChain[TE1], message: TE1): ...
+ def __init__(self, message: TE | Iterable[TE] | Sequence[TE]): ...
@overload
- def __init__(self, message: Iterable[TE]): ...
+ def __init__(self: MessageChain[TE1], message: TE1 | Iterable[TE1] | Sequence[TE1]): ...
@overload
- def __init__(self: MessageChain[TE1], message: Iterable[TE1]): ...
-
- @overload
- def __init__(self: MessageChain[Text], message: Iterable[str]): ...
+ def __init__(self: MessageChain[Text], message: str | Sequence[str]): ...
@overload
def __init__(self: MessageChain[Text | TE1], message: Iterable[str | TE1]): ...
@@ -62,7 +49,13 @@ def __init__(
self: MessageChain[Element],
message: Iterable[str | TE] | str | TE | None = None,
):
- super().__init__()
+ """从传入的序列(可以是元组 tuple, 也可以是列表 list) 创建消息链.
+ Args:
+ message (Iterable[str | TE] | str | TE): 包含且仅包含消息元素和字符串的序列
+ Returns:
+ MessageChain: 以传入的序列作为所承载消息的消息链
+ """
+ self.content: list[TE] = []
if message:
if isinstance(message, (str, Element)):
self.__iadd__(message)
@@ -71,32 +64,48 @@ def __init__(
self.__iadd__(i)
def __str__(self) -> str:
- return "".join(str(elem) for elem in self)
+ """获取以字符串形式表示的消息链, 且趋于通常你见到的样子.
+ Returns:
+ str: 以字符串形式表示的消息链
+ """
+ return "".join(str(elem) for elem in self.content)
def __repr__(self) -> str:
- return "[" + ", ".join(repr(elem) for elem in self) + "]"
+ """获取以字符串形式表示的消息链的详细信息.
+ Returns:
+ str: 以字符串形式表示的消息链的详细信息
+ """
+ return "[" + ", ".join(repr(elem) for elem in self.content) + "]"
@overload
def __add__(self, other: str) -> MessageChain[TE | Text]: ...
@overload
- def __add__(self, other: TE | Iterable[TE]) -> MessageChain[TE]: ...
+ def __add__(self, other: TE | Iterable[TE] | Sequence[TE]) -> MessageChain[TE]: ...
@overload
- def __add__(self, other: TE1 | Iterable[TE1]) -> MessageChain[TE | TE1]: ...
+ def __add__(self, other: TE1 | Iterable[TE1] | Sequence[TE1]) -> MessageChain[TE | TE1]: ...
def __add__(self, other: str | TE | TE1 | Iterable[TE | TE1]) -> MessageChain:
- result: MessageChain = self.fork()
+ """将另一个消息段或消息链添加到当前消息链.
+
+ Args:
+ other: 要添加的消息段或消息链
+
+ Returns:
+ 添加后的消息链
+ """
+ result: MessageChain[Element] = self.fork() # type: ignore
if isinstance(other, str):
- if result and isinstance(text := result[-1], Text):
- result[-1] = Text(text.text + other)
+ if result.content and isinstance(text := result[-1], Text):
+ result.content[-1] = Text(text.text + other)
else:
- result.append(Text(other))
+ result.content.append(Text(other))
elif isinstance(other, Element):
- if result and isinstance(result[-1], Text) and isinstance(other, Text):
- result[-1] = Text(result[-1].text + other.text)
+ if result.content and isinstance(text := result[-1], Text) and isinstance(other, Text):
+ result.content[-1] = Text(text.text + other.text)
else:
- result.append(other)
+ result.content.append(other)
elif isinstance(other, Iterable):
for elem in other:
result += elem
@@ -108,10 +117,10 @@ def __add__(self, other: str | TE | TE1 | Iterable[TE | TE1]) -> MessageChain:
def __radd__(self, other: str) -> MessageChain[Text | TE]: ...
@overload
- def __radd__(self, other: TE | Iterable[TE]) -> MessageChain[TE]: ...
+ def __radd__(self, other: TE | Iterable[TE] | Sequence[TE]) -> MessageChain[TE]: ...
@overload
- def __radd__(self, other: TE1 | Iterable[TE1]) -> MessageChain[TE1 | TE]: ...
+ def __radd__(self, other: TE1 | Iterable[TE1] | Sequence[TE1]) -> MessageChain[TE1 | TE]: ...
def __radd__(self, other: str | TE1 | Iterable[TE1]) -> MessageChain:
result = MessageChain(other)
@@ -119,15 +128,15 @@ def __radd__(self, other: str | TE1 | Iterable[TE1]) -> MessageChain:
def __iadd__(self, other: str | TE | Iterable[TE]) -> Self:
if isinstance(other, str):
- if self and isinstance(text := self[-1], Text):
- list.__setitem__(self, -1, Text(text.text + other))
+ if self.content and isinstance(text := self[-1], Text):
+ self.content[-1] = Text(text.text + other) # type: ignore
else:
- self.append(Text(other)) # type: ignore
+ self.content.append(Text(other)) # type: ignore
elif isinstance(other, Element):
- if self and (isinstance(text := self[-1], Text) and isinstance(other, Text)):
- list.__setitem__(self, -1, Text(text.text + other.text))
+ if self.content and (isinstance(text := self[-1], Text) and isinstance(other, Text)):
+ self.content[-1] = Text(text.text + other.text) # type: ignore
else:
- self.append(other)
+ self.content.append(other)
elif other:
for elem in other:
self.__iadd__(elem)
@@ -136,7 +145,7 @@ def __iadd__(self, other: str | TE | Iterable[TE]) -> Self:
return self
@overload
- def __getitem__(self, args: type[TE1]) -> MessageChain[TE1]:
+ def __getitem__(self, args: type[TE1], /) -> MessageChain[TE1]:
"""获取仅包含指定消息段类型的消息
Args:
@@ -147,7 +156,7 @@ def __getitem__(self, args: type[TE1]) -> MessageChain[TE1]:
"""
@overload
- def __getitem__(self, args: tuple[type[TE1], int]) -> TE1:
+ def __getitem__(self, args: tuple[type[TE1], int], /) -> TE1:
"""索引指定类型的消息段
Args:
@@ -158,7 +167,7 @@ def __getitem__(self, args: tuple[type[TE1], int]) -> TE1:
"""
@overload
- def __getitem__(self, args: tuple[type[TE1], slice]) -> MessageChain[TE1]:
+ def __getitem__(self, args: tuple[type[TE1], slice], /) -> MessageChain[TE1]:
"""切片指定类型的消息段
Args:
@@ -169,7 +178,7 @@ def __getitem__(self, args: tuple[type[TE1], slice]) -> MessageChain[TE1]:
"""
@overload
- def __getitem__(self, args: int) -> TE:
+ def __getitem__(self, args: int, /) -> TE:
"""索引消息段
Args:
@@ -180,7 +189,7 @@ def __getitem__(self, args: int) -> TE:
"""
@overload
- def __getitem__(self, args: slice) -> Self:
+ def __getitem__(self, args: slice, /) -> Self:
"""切片消息段
Args:
@@ -196,35 +205,72 @@ def __getitem__(
) -> TE | TE1 | MessageChain[TE1] | Self:
arg1, arg2 = args if isinstance(args, tuple) else (args, None)
if isinstance(arg1, int) and arg2 is None:
- return super().__getitem__(arg1)
+ return self.content[arg1]
if isinstance(arg1, slice) and arg2 is None:
- return MessageChain(super().__getitem__(arg1)) # type: ignore
+ return MessageChain(self.content[arg1]) # type: ignore
if TYPE_CHECKING:
assert not isinstance(arg1, slice | int)
if issubclass(arg1, Element) and arg2 is None:
- return MessageChain(elem for elem in self if isinstance(elem, arg1)) # type: ignore
+ return MessageChain(elem for elem in self.content if isinstance(elem, arg1)) # type: ignore
if issubclass(arg1, Element) and isinstance(arg2, int):
- return [elem for elem in self if isinstance(elem, arg1)][arg2]
+ return [elem for elem in self.content if isinstance(elem, arg1)][arg2]
if issubclass(arg1, Element) and isinstance(arg2, slice):
- return MessageChain([elem for elem in self if isinstance(elem, arg1)][arg2]) # type: ignore
+ return MessageChain([elem for elem in self.content if isinstance(elem, arg1)][arg2]) # type: ignore
raise ValueError("Incorrect arguments to slice") # pragma: no cover
- def __contains__(self, value: str | Element | type[Element]) -> bool:
- """检查消息段是否存在
+ def __setitem__(self, index: int, value: TE | str, /) -> None:
+ if isinstance(value, str):
+ value = Text(value) # type: ignore
+ self.content[index] = value # type: ignore
+
+ def __delitem__(self, index: int, /) -> None:
+ del self.content[index]
+
+ def __contains__(self, item: str | Element | type[Element] | Self | Sequence[str | Element]) -> bool:
+ """判断消息链中是否含有特定的内容.
Args:
- value: 消息段或消息段类型
+ item (str | Element | type[Element] | Self | Sequence[str | Element]): 需判断内容.
Returns:
消息内是否存在给定消息段或给定类型的消息段
"""
- if isinstance(value, type):
- return not not next((elem for elem in self if isinstance(elem, value)), None)
- if isinstance(value, str):
- value = Text(value)
- return super().__contains__(value)
+ if isinstance(item, type):
+ return not not next((elem for elem in self.content if isinstance(elem, item)), None)
+ if isinstance(item, Element):
+ return item in self.merge().content
+ if isinstance(item, (MessageChain, Sequence)):
+ return not not self.index_sub(item)
+
+ raise ValueError(f"{item} is not an acceptable argument!")
+
+ def merge(self, *, copy: bool = True) -> Self:
+ """合并相邻的 Text 项, 选择返回一个新的消息链实例
+
+ Returns:
+ MessageChain: 得到的新的消息链实例, 里面不应存在有任何的相邻的 Text 元素.
+ """
+
+ result = []
+
+ texts = []
+ for i in self.content:
+ if not isinstance(i, Text):
+ if texts:
+ result.append(Text("".join(texts)))
+ texts.clear() # 清空缓存
+ result.append(i)
+ else:
+ texts.append(i.text)
+ if texts:
+ result.append(Text("".join(texts)))
+ texts.clear() # 清空缓存
+ if copy:
+ return self.__class__(result)
+ self.content.clear()
+ self.content.extend(result)
+ return self
- def has(self, value: str | Element | type[Element]) -> bool:
- return value in self
+ has = __contains__
def index(self, value: str | Element | type[Element], *args: SupportsIndex) -> int:
"""索引消息段
@@ -243,114 +289,177 @@ def index(self, value: str | Element | type[Element], *args: SupportsIndex) -> i
first_elemment = next((elem for elem in self if isinstance(elem, value)), None)
if first_elemment is None:
raise ValueError(f"Element with type {value!r} is not in message")
- return super().index(first_elemment, *args)
+ return self.content.index(first_elemment, *args) # type: ignore
if isinstance(value, str):
value = Text(value)
- return super().index(value, *args) # type: ignore
+ return self.content.index(value, *args) # type: ignore
- def get(self, type_: type[TE], count: int | None = None) -> MessageChain[TE]:
- """获取指定类型的消息段
+ def index_sub(self, sub: MessageChain | Sequence[str | Element]) -> list[int]:
+ """判断消息链是否含有子链. 使用 KMP 算法.
Args:
- type_: 消息段类型
- count: 获取个数
+ sub (MessageChain | Sequence[str | Element]): 要判断的子链.
Returns:
- 构建的新消息
+ List[int]: 所有找到的下标.
+ """
+
+ def unzip(seq: Sequence[str | Element]) -> list[str | Element]:
+ res: list[str | Element] = []
+ for e in seq:
+ if isinstance(e, Text):
+ res.extend(e.text)
+ elif isinstance(e, str):
+ res.extend(e)
+ else:
+ res.append(e)
+ return res
+
+ pattern: list[str | Element] = unzip(sub.content) if isinstance(sub, MessageChain) else unzip(sub)
+
+ match_target: list[str | Element] = unzip(self.content)
+
+ if len(match_target) < len(pattern):
+ return []
+
+ fallback: list[int] = [0 for _ in pattern]
+ current_fb: int = 0 # current fallback index
+ for i in range(1, len(pattern)):
+ while current_fb and pattern[i] != pattern[current_fb]:
+ current_fb = fallback[current_fb - 1]
+ if pattern[i] == pattern[current_fb]:
+ current_fb += 1
+ fallback[i] = current_fb
+
+ match_index: list[int] = []
+ ptr = 0
+ for i, e in enumerate(match_target):
+ while ptr and e != pattern[ptr]:
+ ptr = fallback[ptr - 1]
+ if e == pattern[ptr]:
+ ptr += 1
+ if ptr == len(pattern):
+ match_index.append(i - ptr + 1)
+ ptr = fallback[ptr - 1]
+ return match_index
+
+ def get(self, element_class: type[TE1], count: int | None = None) -> MessageChain[TE1]:
+ """
+ 获取消息链中所有特定类型的消息元素
+
+ Args:
+ element_class (type[E]): 指定的消息元素的类型, 例如 "Text", "At", "Image" 等.
+ count (int, optional): 至多获取的元素个数
+
+ Returns:
+ MessageChain[E]: 获取到的符合要求的所有消息元素; 另: 可能是空列表([]).
"""
if count is None:
- return self[type_]
+ return self[element_class]
- iterator, filtered = (elem for elem in self if isinstance(elem, type_)), MessageChain()
- for _ in range(count):
- elem = next(iterator, None)
- if elem is None:
- break
- filtered.append(elem)
- return filtered # type: ignore
+ return MessageChain(elem for elem in self.content if isinstance(elem, element_class))[:count] # type: ignore
+
+ def get_one(self, element_class: type[TE1], index: int) -> TE1:
+ """获取消息链中第 index + 1 个特定类型的消息元素
+ Args:
+ element_class (type[Element]): 指定的消息元素的类型, 例如 "Text", "At", "Image" 等.
+ index (int): 索引, 从 0 开始数
+ Returns:
+ T: 消息链第 index + 1 个特定类型的消息元素
+ """
+ return self.get(element_class)[index]
+
+ def get_first(self, element_class: type[TE1]) -> TE1:
+ """获取消息链中第 1 个特定类型的消息元素
+ Args:
+ element_class (type[Element]): 指定的消息元素的类型, 例如 "Text", "At", "Image" 等.
+ Returns:
+ T: 消息链第 1 个特定类型的消息元素
+ """
+ return self.get(element_class)[0]
+
+ def join(self, *chains: Self | Iterable[Self]) -> Self:
+ """将多个消息链连接起来, 并在其中插入自身.
+
+ Args:
+ *chains (Iterable[MessageChain]): 要连接的消息链.
+
+ Returns:
+ MessageChain: 连接后的消息链, 已对文本进行合并.
+ """
+ result: list[TE] = []
+ list_chains: list[MessageChain] = []
+ for chain in chains:
+ if isinstance(chain, MessageChain):
+ list_chains.append(chain)
+ else:
+ list_chains.extend(chain)
+
+ for chain in list_chains:
+ if chain is not list_chains[0]:
+ result.extend(deepcopy(self.content))
+ result.extend(deepcopy(chain.content))
+ return self.__class__(result).merge()
def count(self, value: type[Element] | str | Element) -> int:
- """计算指定消息段的个数
+ """计算指定消息元素的个数
Args:
- value: 消息段或消息段类型
+ value (str | Element | type[Element]): 消息元素或消息元素类型
Returns:
- 个数
+ int: 消息元素的个数
"""
if isinstance(value, str):
value = Text(value)
return (
len(self[value]) # type: ignore
if isinstance(value, type)
- else super().count(value) # type: ignore
+ else self.content.count(value) # type: ignore
)
def only(self, value: type[Element] | str | Element) -> bool:
- """检查消息中是否仅包含指定消息段
+ """检查消息中是否仅包含指定消息元素
Args:
- value: 指定消息段或消息段类型
+ value: 指定消息元素或消息元素类型
Returns:
- 是否仅包含指定消息段
+ bool: 是否仅包含指定消息元素
"""
if isinstance(value, type):
- return all(isinstance(elem, value) for elem in self)
+ return all(isinstance(elem, value) for elem in self.content)
if isinstance(value, str):
value = Text(value)
- return all(elem == value for elem in self)
-
- def join(self, iterable: Iterable[TE1 | MessageChain[TE1]]) -> MessageChain[TE | TE1]:
- """将多个消息连接并将自身作为分割
-
- Args:
- iterable: 要连接的消息
-
- Returns:
- 连接后的消息
- """
- ret = MessageChain()
- for index, msg in enumerate(iterable):
- if index != 0:
- ret.extend(self)
- if isinstance(msg, Element):
- ret.append(msg)
- else:
- ret.extend(msg.copy())
- return ret # type: ignore
+ return all(elem == value for elem in self.content)
- def copy(self) -> MessageChain[TE]:
+ def copy(self) -> Self:
"""深拷贝消息"""
return deepcopy(self)
- def fork(self) -> MessageChain[TE]:
+ def fork(self) -> Self:
"""浅拷贝消息"""
new = self.__class__()
- list.extend(new, self)
+ new.content = self.content[:]
return new
- def include(self, *types: type[Element]) -> MessageChain:
- """过滤消息
-
+ def exclude(self, *types: type[Element]) -> Self:
+ """将除了在给出的消息元素类型中符合的消息元素重新包装为一个新的消息链
Args:
- types: 包含的消息段类型
-
+ *types (type[Element]): 将排除在外的消息元素类型
Returns:
- 新构造的消息
+ MessageChain: 返回的消息链中不包含参数中给出的消息元素类型
"""
- return MessageChain(elem for elem in self if elem.__class__ in types)
-
- def exclude(self, *types: type[Element]) -> MessageChain:
- """过滤消息
+ return self.__class__([i for i in self.content if not isinstance(i, types)])
+ def include(self, *types: type[Element]) -> Self:
+ """将只在给出的消息元素类型中符合的消息元素重新包装为一个新的消息链
Args:
- types: 不包含的消息段类型
-
+ *types (type[Element]): 将只包含在内的消息元素类型
Returns:
- 新构造的消息
+ MessageChain: 返回的消息链中只包含参数中给出的消息元素类型
"""
- return MessageChain(elem for elem in self if elem.__class__ not in types)
+ return self.__class__([i for i in self.content if isinstance(i, types)])
def extract_plain_text(self) -> str:
"""提取消息内纯文本消息"""
@@ -363,7 +472,13 @@ def filter(self, predicate: Callable[[TE], bool]) -> MessageChain[TE]:
Args:
predicate: 过滤函数
"""
- return MessageChain(elem for elem in self if predicate(elem))
+ return MessageChain(elem for elem in self.content if predicate(elem))
+
+ def __iter__(self) -> Iterator[TE]:
+ yield from self.content
+
+ def __len__(self) -> int:
+ return len(self.content)
@overload
def map(self, func: Callable[[TE], TE1]) -> MessageChain[TE1]: ...
@@ -374,7 +489,7 @@ def map(self, func: Callable[[TE], T]) -> list[T]: ...
def map(self, func: Callable[[TE], TE1] | Callable[[TE], T]) -> MessageChain[TE1] | list[T]:
result1 = []
result2 = []
- for elem in self:
+ for elem in self.content:
result = func(elem)
if isinstance(result, Element):
result1.append(result)
@@ -418,7 +533,7 @@ def transform(self, rules: SyncVisitor[S], session: S = None) -> MessageChain:
转换后的消息
"""
output = MessageChain()
- for elem in self:
+ for elem in self.content:
result = self._visit_sync(elem, rules, session)
if result is True:
children = MessageChain(elem.children)
@@ -428,7 +543,7 @@ def transform(self, rules: SyncVisitor[S], session: S = None) -> MessageChain:
if isinstance(result, str | Element):
output += result
else:
- output.extend(result)
+ output.content.extend(result)
return output
async def transform_async(self, rules: AsyncVisitor[S], session: S = None) -> MessageChain:
@@ -442,7 +557,7 @@ async def transform_async(self, rules: AsyncVisitor[S], session: S = None) -> Me
转换后的消息
"""
output = MessageChain()
- for elem in self:
+ for elem in self.content:
result = await self._visit_async(elem, rules, session)
if result is True:
children = MessageChain(elem.children)
@@ -467,7 +582,7 @@ def split(self, pattern: str = " ") -> list[Self]:
result: list[Self] = []
tmp = []
- for seg in self:
+ for seg in self.content:
if isinstance(seg, Text):
split_result = seg.text.split(pattern)
for index, split_text in enumerate(split_result):
@@ -483,11 +598,7 @@ def split(self, pattern: str = " ") -> list[Self]:
tmp = []
return result
- def replace(
- self,
- old: str,
- new: str,
- ) -> Self:
+ def replace(self, old: str, new: str) -> Self:
"""替换消息中有关的文本
Args:
@@ -498,7 +609,7 @@ def replace(
UniMessage: 修改后的消息链, 若未替换则原样返回.
"""
result_list: list[TE] = []
- for seg in self:
+ for seg in self.content:
if isinstance(seg, Text):
result_list.append(seg.__class__(seg.text.replace(old, new)))
else:
@@ -515,9 +626,9 @@ def startswith(self, string: str) -> bool:
bool: 是否以给出的字符串开头
"""
- if not self or not isinstance(self[0], Text):
+ if not self.content or not isinstance(text := self.content[0], Text):
return False
- return list.__getitem__(self, 0).text.startswith(string)
+ return text.text.startswith(string)
def endswith(self, string: str) -> bool:
"""判断消息链是否以给出的字符串结尾
@@ -529,102 +640,236 @@ def endswith(self, string: str) -> bool:
bool: 是否以给出的字符串结尾
"""
- if not self or not isinstance(self[-1], Text):
+ if not self.content or not isinstance(text := self.content[-1], Text):
return False
- return list.__getitem__(self, -1).text.endswith(string)
+ return text.text.endswith(string)
+
+ def append(self, element: TE | str) -> None:
+ """
+ 向消息链最后追加单个元素
+
+ Args:
+ element (Element): 要添加的元素
+
+ Returns:
+ None
+ """
+ self.content.append(Text(element) if isinstance(element, str) else element) # type: ignore
+
+ def insert(self, index: int, value: Element | str, /) -> None:
+ if isinstance(value, str):
+ value = Text(value)
+ self.content.insert(index, value) # type: ignore
+
+ def extend(self, values: Iterable[Self | TE | Sequence[TE | str]]) -> None:
+ """
+ 向消息链最后添加元素/元素列表/消息链
+
+ Args:
+ *values (MessageChain | Element | list[Element | str]): 要添加的元素/元素容器.
+
+ Returns:
+ MessageChain: copy = True 时返回副本, 否则返回自己的引用.
+ """
+ result = []
+ for i in values:
+ if isinstance(i, Element):
+ result.append(i)
+ elif isinstance(i, str):
+ result.append(Text(i))
+ elif isinstance(i, MessageChain):
+ result.extend(i.content)
+ else:
+ for e in i:
+ if isinstance(e, str):
+ result.append(Text(e))
+ else:
+ result.append(e)
+ self.content.extend(result)
+
+ def empty(self) -> bool:
+ """
+ 判断消息链是否为空,包括判断是否仅包含空字符串。
- def removeprefix(self, prefix: str) -> Self:
+ Returns:
+ bool: 判断结果。
+ """
+
+ return not bool(self.content and str(self))
+
+ def pop(self, index: int = -1, /) -> TE:
+ """移除并返回指定位置的元素,默认移除最后一个元素。
+
+ Args:
+ index (int, optional): 要移除的元素的索引,默认为 -1(最后一个元素)。
+
+ Returns:
+ TE: 被移除的元素。
+ """
+ return self.content.pop(index) # type: ignore
+
+ def removeprefix(self, prefix: str, *, copy: bool = True) -> Self:
"""移除消息链前缀.
Args:
prefix (str): 要移除的前缀.
+ copy (bool, optional): 是否在副本上修改, 默认为 True.
Returns:
- UniMessage: 修改后的消息链.
+ MessageChain: 修改后的消息链, 若未移除则原样返回.
"""
- copy = list.copy(self)
- if not copy:
- return self.__class__(copy)
- seg = copy[0]
- if not isinstance(seg, Text):
- return self.__class__(copy)
- if seg.text.startswith(prefix):
- seg = seg.__class__(seg.text[len(prefix) :])
- if not seg.text:
- copy.pop(0)
- else:
- copy[0] = seg
- return self.__class__(copy)
+ elements = deepcopy(self.content) if copy else self.content
+ if not elements:
+ return self.copy() if copy else self
+ elem = elements[0]
+ if not isinstance(elem, Text):
+ return self.copy() if copy else self
+ if elem.text.startswith(prefix):
+ elem.text = elem.text[len(prefix) :]
+ if not elem.text:
+ elements.pop(0)
+ if copy:
+ return self.__class__(elements)
+ self.content.clear()
+ self.content.extend(elements)
+ return self
- def removesuffix(self, suffix: str) -> Self:
+ def removesuffix(self, suffix: str, *, copy: bool = True) -> Self:
"""移除消息链后缀.
Args:
suffix (str): 要移除的后缀.
+ copy (bool, optional): 是否在副本上修改, 默认为 True.
Returns:
- UniMessage: 修改后的消息链.
+ MessageChain: 修改后的消息链, 若未移除则原样返回.
"""
- copy = list.copy(self)
- if not copy:
- return self.__class__(copy)
- seg = copy[-1]
- if not isinstance(seg, Text):
- return self.__class__(copy)
- if seg.text.endswith(suffix):
- seg = seg.__class__(seg.text[: -len(suffix)])
- if not seg.text:
- copy.pop(-1)
- else:
- copy[-1] = seg
- return self.__class__(copy)
-
- def strip(self, *segments: str | Element | type[Element]) -> Self:
- return self.lstrip(*segments).rstrip(*segments)
-
- def lstrip(self, *segments: str | Element | type[Element]) -> Self:
- types = [i for i in segments if not isinstance(i, str)] or []
- chars = "".join([i for i in segments if isinstance(i, str)]) or None
- copy = list.copy(self)
- if not copy:
- return self.__class__(copy)
- while copy:
- seg = copy[0]
- if seg in types or seg.__class__ in types:
- copy.pop(0)
- elif isinstance(seg, Text):
- seg = seg.__class__(seg.text.lstrip(chars))
- if not seg.text:
- copy.pop(0)
+ elements = deepcopy(self.content) if copy else self.content
+ if not elements:
+ return self.copy() if copy else self
+ elem = elements[-1]
+ if not isinstance(elem, Text):
+ return self.copy() if copy else self
+ if elem.text.endswith(suffix):
+ elem.text = elem.text[: -len(suffix)]
+ if not elem.text:
+ elements.pop(-1)
+ if copy:
+ return self.__class__(elements)
+ self.content.clear()
+ self.content.extend(elements)
+ return self
+
+ def strip(self, *elements: str | type[Element] | Element, copy: bool = True) -> Self:
+ return self.lstrip(*elements, copy=copy).rstrip(*elements, copy=copy)
+
+ def lstrip(self, *elements: str | type[Element] | Element, copy: bool = True) -> Self:
+ types = [i for i in elements if not isinstance(i, str)] or []
+ chars = "".join([i for i in elements if isinstance(i, str)]) or None
+ content = deepcopy(self.content) if copy else self.content
+ if not content:
+ return self.copy() if copy else self
+ while content:
+ elem = content[0]
+ if elem in types or elem.__class__ in types:
+ content.pop(0)
+ elif isinstance(elem, Text):
+ text = elem.text.lstrip(chars)
+ if not text:
+ content.pop(0)
continue
- else:
- copy[0] = seg
+ elem.text = text
break
else:
break
- return self.__class__(copy)
-
- def rstrip(self, *segments: str | Element | type[Element]) -> Self:
- types = [i for i in segments if not isinstance(i, str)] or []
- chars = "".join([i for i in segments if isinstance(i, str)]) or None
- copy = list.copy(self)
- if not copy:
- return self.__class__(copy)
- while copy:
- seg = copy[-1]
- if seg in types or seg.__class__ in types:
- copy.pop(-1)
- elif isinstance(seg, Text):
- seg = seg.__class__(seg.text.rstrip(chars))
- if not seg.text:
- copy.pop(-1)
+ if copy:
+ return self.__class__(content)
+ self.content.clear()
+ self.content.extend(content)
+ return self
+
+ def rstrip(self, *elements: str | type[Element] | Element, copy: bool = True) -> Self:
+ types = [i for i in elements if not isinstance(i, str)] or []
+ chars = "".join([i for i in elements if isinstance(i, str)]) or None
+ content = deepcopy(self.content) if copy else self.content
+ if not content:
+ return self.copy() if copy else self
+ while content:
+ elem = content[-1]
+ if elem in types or elem.__class__ in types:
+ content.pop(-1)
+ elif isinstance(elem, Text):
+ text = elem.text.rstrip(chars)
+ if not text:
+ content.pop(-1)
continue
- else:
- copy[-1] = seg
+ elem.text = text
break
else:
break
- return self.__class__(copy)
+ if copy:
+ return self.__class__(content)
+ self.content.clear()
+ self.content.extend(content)
+ return self
+
+ def replace_chain(self, old: Sequence[Element], new: Sequence[Element]) -> Self:
+ """替换消息链中的一部分. (在副本上操作)
+
+ Args:
+ old (MessageChain): 要替换的消息链.
+ new (MessageChain): 替换后的消息链.
+
+ Returns:
+ MessageChain: 修改后的消息链, 若未替换则原样返回.
+ """
+ if not isinstance(old, MessageChain):
+ old = MessageChain(old)
+ if not isinstance(new, MessageChain):
+ new = MessageChain(new)
+ index_list: list[int] = self.index_sub(old)
+
+ def unzip(chain: MessageChain) -> list[str | Element]:
+ unzipped: list[str | Element] = []
+ for e in chain.content:
+ if isinstance(e, Text):
+ unzipped.extend(e.text)
+ else:
+ unzipped.append(e)
+ return unzipped
+
+ unzipped_new: list[str | Element] = unzip(new)
+ unzipped_old: list[str | Element] = unzip(old)
+ unzipped_self: list[str | Element] = unzip(self)
+ unzipped_result: list[str | Element] = []
+ last_end: int = 0
+ for start in index_list:
+ unzipped_result.extend(unzipped_self[last_end:start])
+ last_end = start + len(unzipped_old)
+ unzipped_result.extend(unzipped_new)
+ unzipped_result.extend(unzipped_self[last_end:])
+
+ # Merge result
+ result_list: list[TE] = []
+ char_stk: list[str] = []
+ for v in unzipped_result:
+ if isinstance(v, str):
+ char_stk.append(v)
+ else:
+ result_list.append(Text("".join(char_stk))) # type: ignore
+ char_stk = []
+ result_list.append(v) # type: ignore
+ if char_stk:
+ result_list.append(Text("".join(char_stk))) # type: ignore
+ return self.__class__(result_list)
+
+ def __bool__(self):
+ return bool(self.content and str(self))
+
+ def __eq__(self, value: object, /):
+ if not isinstance(value, MessageChain):
+ return False
+ return value.content == self.content
def display(self):
texts = []
diff --git a/arclet/entari/plugin/__init__.py b/arclet/entari/plugin/__init__.py
index d0cb445..db6c173 100644
--- a/arclet/entari/plugin/__init__.py
+++ b/arclet/entari/plugin/__init__.py
@@ -7,7 +7,7 @@
from pathlib import Path
from typing import TYPE_CHECKING, Any, Literal, TypeVar, overload
-from arclet.letoderea import Subscriber, on, publish
+from arclet.letoderea import Subscriber, on
from arclet.letoderea.effect import AsyncDisposable, Disposable
from arclet.letoderea.utils import Resultable
from tarina import init_spec
@@ -15,19 +15,19 @@
from ..config import EntariConfig, config_model_keys, config_model_validate
from ..event.config import ConfigReload
-from ..event.lifespan import Ready
-from ..event.plugin import PluginLoadedFailed
-from ..exceptions import RegisterNotInPluginError, ReusablePluginError, StaticPluginDispatchError
-from ..logger import log
+from ..exceptions import StaticPluginDispatchError
from ..message import Fragment, MessageChain, Render
from ..session import COMPONENTS, Session, component_transform
-from ..utils import escape_tag
+from .loader import find_plugin as find_plugin
+from .loader import load_plugin as load_plugin
+from .loader import reload_plugin as reload_plugin
+from .loader import reload_subplugin as reload_subplugin
+from .loader import unload_plugin as unload_plugin
from .model import TS, Plugin, PluginDispatcher, current_plugin
from .model import PluginMetadata as PluginMetadata
from .model import PluginRole as PluginRole
from .model import RootlessPlugin as RootlessPlugin
from .model import keeping as keeping
-from .module import import_plugin
from .module import package as package
from .module import requires as requires
from .service import plugin_service
@@ -200,108 +200,6 @@ def dispatch(event: type, name: str | None = None) -> PluginDispatcher:
return get_plugin(1).dispatch(event, name=name)
-def load_plugin(
- path: str, config: dict | None = None, recursive_guard: set[str] | None = None, prelude: bool = False
-) -> Plugin | None:
- """
- 以导入路径方式加载模块
-
- Args:
- path (str): 模块路径
- config (dict): 模块配置
- recursive_guard (set[str]): 递归保护
- prelude (bool): 是否为前置插件
- """
- if config is not None:
- config["$path"] = path
- else:
- for k, names in EntariConfig.instance._plugin_names.items():
- if path in names:
- config = EntariConfig.instance.plugin.get(k, {})
- config["$path"] = k
- break
- else:
- config = {"$path": path}
- if prelude:
- config["$static"] = True
- if recursive_guard is None:
- recursive_guard = set()
- path = path.replace("::", "arclet.entari.builtins.")
- while path in plugin_service._subplugined:
- path = plugin_service._subplugined[path]
- if path in plugin_service._apply:
- if path in plugin_service.plugins:
- return plugin_service.plugins[path]
- log.plugin.trace(f"loaded rootless plugin {path!r}")
- return plugin_service._apply[path][0](config)
- if plug := find_plugin(path):
- plugin_service._direct_plugins.add(plug.path)
- return plug
- try:
- mod = import_plugin(path, config=config)
- if not mod:
- mod = next(
- (import_plugin(_path, config=config) for _path in EntariConfig.instance._plugin_names.get(path, [])),
- None,
- )
- if not mod:
- log.plugin.error(f"cannot found plugin {path!r}")
- publish(PluginLoadedFailed(path))
- return
- plugin_service._direct_plugins.add(mod.__name__)
- if mod.__name__ in plugin_service.referents and plugin_service.referents[mod.__name__]:
- referents = plugin_service.referents[mod.__name__].copy()
- # plugin_service.referents[mod.__name__].clear()
- for referent in referents:
- if referent in recursive_guard:
- continue
- if referent.startswith(mod.__name__):
- continue
- if referent in plugin_service._subplugined and mod.__name__.startswith(
- plugin_service._subplugined[referent]
- ):
- continue
- if referent in plugin_service.plugins:
- plugin_service.referents[mod.__name__].discard(referent)
- log.plugin.debug(f"reloading {escape_tag(mod.__name__)}'s referent {referent!r}")
- unload_plugin(referent)
- if not (plug := load_plugin(referent)):
- plugin_service.referents[mod.__name__].add(referent)
- else:
- publish(Ready(), plug._scope)
- recursive_guard.add(referent)
-
- return mod.__plugin__
- except (ImportError, RegisterNotInPluginError, ReusablePluginError, StaticPluginDispatchError):
- return
- except Exception as e:
- log.plugin.exception(f"failed to load plugin {path!r}: {e}", exc_info=e)
- publish(PluginLoadedFailed(path))
- return
-
-
-def load_plugins(dir_: str | os.PathLike | Path):
- """加载指定目录下的所有插件"""
- path = dir_ if isinstance(dir_, Path) else Path(dir_)
- if not path.is_dir():
- raise NotADirectoryError(f"{path} is not a directory")
- path: Path = path.resolve() # .relative_to(Path.cwd())
- syspaths = [Path(p).resolve() for p in sys.path if p]
- prefixes = [p for p in syspaths if path.is_relative_to(p)]
- if prefixes:
- prefix = max(prefixes, key=lambda p: len(p.parts))
- else:
- prefix = Path.cwd()
- for p in path.iterdir():
- if p.suffix in (".py", "") and p.stem not in {"__init__", "__pycache__"}:
- p = p.resolve().relative_to(prefix)
- if len(p.parts) > 1:
- plg = ".".join(p.parts[:-1:1]) + "." + p.stem
- else:
- plg = p.stem
- load_plugin(plg)
-
-
if TYPE_CHECKING:
@init_spec(PluginMetadata)
@@ -370,6 +268,43 @@ def _reload(event: ConfigReload):
get_config = plugin_config
+def load_plugins(dir_: str | os.PathLike | Path):
+ """加载指定目录下的所有插件"""
+ path = dir_ if isinstance(dir_, Path) else Path(dir_)
+ if not path.is_dir():
+ raise NotADirectoryError(f"{path} is not a directory")
+ path: Path = path.resolve() # .relative_to(Path.cwd())
+ syspaths = [Path(p).resolve() for p in sys.path if p]
+ prefixes = [p for p in syspaths if path.is_relative_to(p)]
+ if prefixes:
+ prefix = max(prefixes, key=lambda p: len(p.parts))
+ else:
+ prefix = Path.cwd()
+ for p in path.iterdir():
+ if p.suffix in (".py", "") and p.stem not in {"__init__", "__pycache__"}:
+ p = p.resolve().relative_to(prefix)
+ if len(p.parts) > 1:
+ plg = ".".join(p.parts[:-1:1]) + "." + p.stem
+ else:
+ plg = p.stem
+ load_plugin(plg)
+
+
+def find_plugin_by_file(file: str) -> Plugin | None:
+ path = Path(file).resolve()
+ for plugin in plugin_service.plugins.values():
+ if plugin.module.__file__ == str(path):
+ return plugin
+ if plugin.module.__file__ and Path(plugin.module.__file__).parent == path:
+ return plugin
+ path1 = Path(path)
+ while path1.parent != path1:
+ if str(path1) == plugin.module.__file__:
+ return plugin
+ path1 = path1.parent
+ return None
+
+
def declare_static():
"""声明当前插件为静态插件"""
_plugin = get_plugin(1)
@@ -400,43 +335,6 @@ def restore():
return get_plugin(1).restore()
-def find_plugin(name: str) -> Plugin | None:
- """根据插件名称查找插件"""
- if name in plugin_service.plugins:
- return plugin_service.plugins[name]
- if name in EntariConfig.instance.plugin_prefixes:
- for prefix in EntariConfig.instance.plugin_prefixes[name]:
- if f"{prefix}{name}" in plugin_service.plugins:
- return plugin_service.plugins[f"{prefix}{name}"]
- if not name.count(".") and f"entari_plugin_{name}" in plugin_service.plugins:
- return plugin_service.plugins[f"entari_plugin_{name}"]
-
-
-def find_plugin_by_file(file: str) -> Plugin | None:
- path = Path(file).resolve()
- for plugin in plugin_service.plugins.values():
- if plugin.module.__file__ == str(path):
- return plugin
- if plugin.module.__file__ and Path(plugin.module.__file__).parent == path:
- return plugin
- path1 = Path(path)
- while path1.parent != path1:
- if str(path1) == plugin.module.__file__:
- return plugin
- path1 = path1.parent
- return None
-
-
-def unload_plugin(plugin: str):
- plugin = plugin.replace("::", "arclet.entari.builtins.")
- while plugin in plugin_service._subplugined:
- plugin = plugin_service._subplugined[plugin]
- if not (_plugin := find_plugin(plugin)):
- return False
- _plugin.dispose()
- return True
-
-
async def unload_plugin_async(plugin: str):
plugin = plugin.replace("::", "arclet.entari.builtins.")
while plugin in plugin_service._subplugined:
diff --git a/arclet/entari/plugin/loader.py b/arclet/entari/plugin/loader.py
new file mode 100644
index 0000000..28736c8
--- /dev/null
+++ b/arclet/entari/plugin/loader.py
@@ -0,0 +1,425 @@
+import ast
+import asyncio
+import sys
+from typing import Any
+
+from arclet.letoderea import publish
+
+from ..config import EntariConfig
+from ..event.lifespan import Ready
+from ..event.plugin import PluginLoadedFailed
+from ..exceptions import RegisterNotInPluginError, ReusablePluginError, StaticPluginDispatchError
+from ..logger import log
+from .model import Plugin, current_plugin
+from .module import import_plugin
+from .service import plugin_service
+
+
+def collect_module_level_names(nodes: ast.Module) -> set[str]:
+ """收集模块层表达式引用的名字(函数体除外;含装饰器/默认值/注解/基类/类级语句)"""
+ used: set[str] = set()
+
+ def visit_expr(expr: ast.expr):
+ if isinstance(expr, ast.Name) and isinstance(expr.ctx, ast.Load):
+ used.add(expr.id)
+ return
+ for child in ast.iter_child_nodes(expr):
+ if isinstance(child, ast.expr):
+ visit_expr(child)
+ elif isinstance(child, ast.keyword) and child.value is not None:
+ visit_expr(child.value)
+
+ def visit_stmt(stmt: ast.stmt):
+ if isinstance(stmt, (ast.FunctionDef, ast.AsyncFunctionDef)):
+ for dec in stmt.decorator_list:
+ visit_expr(dec)
+ for default in [*stmt.args.defaults, *[d for d in stmt.args.kw_defaults if d]]:
+ visit_expr(default)
+ for arg in stmt.args.posonlyargs + stmt.args.args + stmt.args.kwonlyargs:
+ if arg.annotation:
+ visit_expr(arg.annotation)
+ if stmt.args.vararg and stmt.args.vararg.annotation:
+ visit_expr(stmt.args.vararg.annotation)
+ if stmt.args.kwarg and stmt.args.kwarg.annotation:
+ visit_expr(stmt.args.kwarg.annotation)
+ return
+ if isinstance(stmt, ast.ClassDef):
+ for dec in stmt.decorator_list:
+ visit_expr(dec)
+ for base in stmt.bases:
+ visit_expr(base)
+ for kw in stmt.keywords:
+ visit_expr(kw.value)
+ for s in stmt.body:
+ visit_stmt(s)
+ return
+ for child in ast.iter_child_nodes(stmt):
+ if isinstance(child, ast.expr):
+ visit_expr(child)
+ elif isinstance(child, ast.stmt):
+ visit_stmt(child)
+ elif isinstance(child, ast.keyword) and child.value is not None:
+ visit_expr(child.value)
+
+ for stmt in nodes.body:
+ visit_stmt(stmt)
+ return used
+
+
+def _uses_module_level(plugin: Plugin, path: str) -> bool:
+ """检测插件是否在模块层使用 path 的绑定名(基类/模块级装饰器/顶层实例化等)"""
+ if not plugin._inspect:
+ return True
+ bound = {name for name, (target, _) in plugin.bindings.items() if target == path or target.startswith(path + ".")}
+ if not bound:
+ return False
+ return bool(collect_module_level_names(plugin._inspect.nodes) & bound)
+
+
+_MISSING = object()
+
+
+def _rebind_imports(plugin: Plugin, path: str) -> bool:
+ """将插件对 path(或其子树)的 import 绑定改写为新对象;False 表示有绑定无法满足(升级全量)
+
+ 模块绑定(attr 为 None)仅在名字等同于目标模块或其末组件时写回:`import a.b` 风格的名字是顶层包,
+ 其子模块链由 promote 的父属性写回维护,不在此处理。
+ 绑定目标为 path 的父包、且父包绑定记录将该名字解析到 path 时,
+ 经父模块 getattr 取新值(父包在 sorted 依赖序中先于其子模块处理,故已重绑完成)。
+ """
+ module = plugin.module
+ parent_pkg = path.rpartition(".")[0]
+ for name, (target, attr) in plugin.bindings.items():
+ if target == path or target.startswith(path + "."):
+ if target not in plugin_service.plugins:
+ continue
+ new_module = plugin_service.plugins[target].module
+ if attr is None:
+ if name != target and name != target.rpartition(".")[-1]:
+ continue
+ value: Any = new_module
+ else:
+ value = getattr(new_module, attr, _MISSING)
+ if value is _MISSING:
+ return False
+ module.__dict__[name] = value
+ elif (
+ parent_pkg
+ and attr is not None
+ and target == parent_pkg
+ and parent_pkg in plugin_service.bindings
+ and parent_pkg in plugin_service.plugins
+ ):
+ parent_chain = plugin_service.bindings[parent_pkg]
+ if parent_chain.get(attr or name, (None, None))[0] == path:
+ value = getattr(plugin_service.plugins[parent_pkg].module, attr or name, _MISSING)
+ if value is _MISSING:
+ return False
+ module.__dict__[name] = value
+ return True
+
+
+def public_fingerprint(plugin: Plugin) -> str | None:
+ """插件公开面指纹:排序后的导出名 + 各名字的定义节点的dump(导入名仅记录类型)
+
+ 无 `__all__` 时导出名 = 非下划线开头的模块 dict 名字;定义节点 dump 覆盖函数体与类体,任何行为差异都反映为指纹差异。
+ """
+ if not plugin._inspect:
+ return None
+ module = plugin.module
+ names = getattr(module, "__all__", None)
+ if names is None:
+ names = sorted(n for n in module.__dict__ if not n.startswith("_"))
+ else:
+ names = sorted(names)
+ node_map: dict[str, ast.stmt] = {}
+ for stmt in plugin._inspect.nodes.body:
+ if isinstance(stmt, (ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)):
+ node_map[stmt.name] = stmt
+ elif isinstance(stmt, ast.Assign):
+ for target in stmt.targets:
+ if isinstance(target, ast.Name):
+ node_map[target.id] = stmt
+ elif isinstance(stmt, ast.AnnAssign) and isinstance(stmt.target, ast.Name):
+ node_map[stmt.target.id] = stmt
+ parts: list[str] = []
+ for name in names:
+ if node := node_map.get(name):
+ parts.append(f"{name}:{ast.dump(node, include_attributes=False)}")
+ else:
+ value = module.__dict__.get(name, None)
+ parts.append(f"{name}:imported:{type(value).__name__}")
+ return "\n".join(parts)
+
+
+def _fingerprint_changed(plugin: Plugin) -> bool:
+ """比较插件公开面指纹(存储的旧指纹 vs 现算新指纹),并更新存储
+
+ 旧指纹缺失(未存储过)时视为已变化:无法证明公开面未变 → 保守级联。
+ """
+ path = plugin.path
+ old = plugin_service.fingerprints.get(path)
+ new = public_fingerprint(plugin)
+ plugin_service.fingerprints[path] = new or ""
+ return old is None or old != new
+
+
+def _rebind_dep(dep: Plugin, path: str):
+ """重绑依赖方对 path 的 import 绑定并恢复其可用状态"""
+ if _rebind_imports(dep, path):
+ log.plugin.debug(f"rebound {dep.id!r}'s imports on {path}")
+ dep.enable()
+ else:
+ log.plugin.warning(f"cannot satisfy {dep.id!r}'s imports on {path}, falling back to full reload")
+ _cascade_dep(dep.id, plugin_service.referents.get(path, set()), set())
+
+
+def _cascade_dep(dep_id: str, referent_set: set[str], recursive_guard: set[str]):
+ """对依赖方执行全量级联(unload + load + Ready),失败时恢复 referent 边"""
+ referent_set.discard(dep_id)
+ log.plugin.debug(f"reloading {dep_id!r}")
+ unload_plugin(dep_id)
+ if not (plug := load_plugin(dep_id)):
+ referent_set.add(dep_id)
+ else:
+ publish(Ready(), plug._scope)
+ recursive_guard.add(dep_id)
+
+
+def _handle_dependents(plugin: Plugin, recursive_guard: set[str] | None = None):
+ """插件完成(重)加载后处理依赖方:公开面未变 → 全部重绑;否则按模块层使用细粒度重绑或级联
+
+ 依赖方 = referents 图与绑定索引的并集;公开面指纹未变时新对象与旧对象行为等价,模块层绑定无需重执行;
+ 指纹变化时仅模块层使用方升级全量,惰性依赖方重绑即可。
+ """
+ if recursive_guard is None:
+ recursive_guard = set()
+ path = plugin.path
+ dependents = set(plugin_service.referents.get(path, ())) | set(plugin_service.dependents_of(path, ensure=False))
+ if not dependents:
+ return
+ surface_changed = _fingerprint_changed(plugin)
+ referent_set = plugin_service.referents.setdefault(path, set())
+ current = current_plugin.get(None)
+ for dep_id in plugin_service.topo_dependents(dependents):
+ if dep_id in recursive_guard:
+ continue
+ if current is not None and (dep_id == current.id or dep_id.startswith(current.id + ".")):
+ # 如果依赖方是当前正在加载的插件或其子插件,说明它们在同一 load_plugins 调用链中被导入,
+ # 且已在当前上下文中处理过依赖关系。
+ continue
+ if dep_id == path or dep_id.startswith(path + "."):
+ continue
+ dep = plugin_service.plugins.get(dep_id)
+ if dep is None:
+ if not (plug := load_plugin(dep_id)):
+ referent_set.add(dep_id)
+ continue
+ publish(Ready(), plug._scope)
+ recursive_guard.add(dep_id)
+ continue
+ if not surface_changed:
+ _rebind_dep(dep, path)
+ continue
+ if _uses_module_level(dep, path):
+ _cascade_dep(dep_id, referent_set, recursive_guard)
+ else:
+ _rebind_dep(dep, path)
+
+
+def load_plugin(
+ path: str,
+ config: dict | None = None,
+ recursive_guard: set[str] | None = None,
+ prelude: bool = False,
+ staged: bool = False,
+) -> Plugin | None:
+ """
+ 以导入路径方式加载模块
+
+ Args:
+ path (str): 模块路径
+ config (dict): 模块配置
+ recursive_guard (set[str]): 递归保护
+ prelude (bool): 是否为前置插件
+ staged (bool): 是否为暂存加载
+ """
+ if config is not None:
+ config["$path"] = path
+ else:
+ for k, names in EntariConfig.instance._plugin_names.items():
+ if path in names:
+ config = EntariConfig.instance.plugin.get(k, {})
+ config["$path"] = k
+ break
+ else:
+ config = {"$path": path}
+ if prelude:
+ config["$static"] = True
+ if recursive_guard is None:
+ recursive_guard = set()
+ path = path.replace("::", "arclet.entari.builtins.")
+ while path in plugin_service._subplugined:
+ path = plugin_service._subplugined[path]
+ if path in plugin_service._apply:
+ if path in plugin_service.plugins:
+ return plugin_service.plugins[path]
+ log.plugin.trace(f"loaded rootless plugin {path!r}")
+ return plugin_service._apply[path][0](config)
+ if not staged and (plug := find_plugin(path)):
+ plugin_service._direct_plugins.add(plug.path)
+ return plug
+ try:
+ mod = import_plugin(path, config=config, staged=staged)
+ if not mod:
+ mod = next(
+ (
+ import_plugin(_path, config=config, staged=staged)
+ for _path in EntariConfig.instance._plugin_names.get(path, [])
+ ),
+ None,
+ )
+ if not mod:
+ log.plugin.error(f"cannot found plugin {path!r}")
+ publish(PluginLoadedFailed(path))
+ return
+ plugin_service._direct_plugins.add(mod.__name__)
+ if not staged:
+ _handle_dependents(mod.__plugin__, recursive_guard)
+ return mod.__plugin__
+ except (ImportError, RegisterNotInPluginError, ReusablePluginError, StaticPluginDispatchError):
+ return
+ except Exception as e:
+ log.plugin.exception(f"failed to load plugin {path!r}: {e}", exc_info=e)
+ publish(PluginLoadedFailed(path))
+ return
+
+
+def find_plugin(name: str) -> Plugin | None:
+ """根据插件名称查找插件"""
+ if name in plugin_service.plugins:
+ return plugin_service.plugins[name]
+ if name in EntariConfig.instance.plugin_prefixes:
+ for prefix in EntariConfig.instance.plugin_prefixes[name]:
+ if f"{prefix}{name}" in plugin_service.plugins:
+ return plugin_service.plugins[f"{prefix}{name}"]
+ if not name.count(".") and f"entari_plugin_{name}" in plugin_service.plugins:
+ return plugin_service.plugins[f"entari_plugin_{name}"]
+
+
+def unload_plugin(plugin: str):
+ """卸载插件及其子插件"""
+ plugin = plugin.replace("::", "arclet.entari.builtins.")
+ while plugin in plugin_service._subplugined:
+ plugin = plugin_service._subplugined[plugin]
+ if not (_plugin := find_plugin(plugin)):
+ return False
+ _plugin.dispose()
+ return True
+
+
+def promote_staged(plugin: Plugin):
+ """将暂存插件及其子插件正式注册进 plugin_service,并恢复启用状态与父绑定
+
+ 子插件按加载顺序已记入父插件 subplugins 列表;
+ 父属性写回(插件导入链)使模块中的属性指向新模块对象。
+ """
+ for sid in [plugin.id, *plugin.subplugins]:
+ if staged := plugin_service._staged.pop(sid, None):
+ plugin_service.plugins[sid] = staged
+ if sid != plugin.id:
+ plugin_service._subplugined[sid] = plugin.id
+ staged.check_disable()
+ if plugin_service.status.blocking:
+ publish(Ready(), staged._scope)
+ plugin_service._unloaded.discard(plugin.id)
+ for sid in plugin.subplugins:
+ if sid not in plugin_service.plugins or sid not in plugin_service._subplugined:
+ continue
+ parent = plugin_service.plugins.get(plugin_service._subplugined[sid])
+ if parent is not None:
+ parent.module.__dict__[sid.rpartition(".")[-1]] = plugin_service.plugins[sid].module
+
+
+def _collect_subtree(plugin: Plugin) -> list[str]:
+ """DFS 收集插件树的全部子插件 id(父在前,去重)"""
+ ids: list[str] = []
+ seen: set[str] = set()
+
+ def walk(plug: Plugin):
+ for sid in plug.subplugins:
+ if sid in seen:
+ continue
+ seen.add(sid)
+ ids.append(sid)
+ if sub := plugin_service.plugins.get(sid):
+ walk(sub)
+
+ walk(plugin)
+ return ids
+
+
+async def reload_plugin(path: str, conf: dict | None = None) -> bool:
+ """原子重载:导入失败时旧插件继续运行;成功后处理依赖方"""
+ path = path.replace("::", "arclet.entari.builtins.")
+ while path in plugin_service._subplugined:
+ path = plugin_service._subplugined[path]
+ if not (plugin := find_plugin(path)):
+ return False
+ if plugin.is_static:
+ return False
+ _conf = conf if conf is not None else plugin.config.copy()
+ old_subplugins = _collect_subtree(plugin)
+ for name in old_subplugins:
+ sys.modules.pop(name, None)
+ log.plugin.debug(f"staged loading {path!r}, old plugin keeps running until swap")
+ if not (new_plugin := load_plugin(path, _conf, staged=True)):
+ log.plugin.error(f"failed to load staged plugin {path!r}, old plugin keeps running")
+ return False
+ if tasks := plugin.dispose(replacing=True):
+ await asyncio.wait(tasks)
+ promote_staged(new_plugin)
+ # 恢复未随 staged exec 重新导入的子插件(load_plugins/config 方式加载的模块在
+ # 暂存加载期间 load_plugin 上溯命中旧插件早退,且不记入新插件 subplugins)
+ for sub_id in old_subplugins:
+ if sub_id in new_plugin.subplugins or sub_id in plugin_service.plugins:
+ continue
+ try:
+ mod = import_plugin(sub_id)
+ except Exception as e:
+ log.plugin.error(f"failed to restore sub-plugin {sub_id!r}: {e!r}")
+ continue
+ if mod is None:
+ log.plugin.error(f"cannot restore sub-plugin {sub_id!r}: module not found")
+ continue
+ log.plugin.debug(f"restored sub-plugin {sub_id!r} not re-imported by staged reload")
+ _handle_dependents(new_plugin)
+ return True
+
+
+async def reload_subplugin(path: str, conf: dict | None = None) -> bool:
+ """子插件粒度重载:仅替换子插件自身,父插件与兄弟经绑定索引重绑
+
+ 父插件在模块层使用该子插件名字(基类/模块级装饰器/顶层实例化)→ 回退整树重载;
+ 子插件自身经暂存机制原子替换(失败时旧子插件继续运行)。
+ """
+ path = path.replace("::", "arclet.entari.builtins.")
+ if path not in plugin_service.plugins or path not in plugin_service._subplugined:
+ return False
+ plugin = plugin_service.plugins[path]
+ parent_id = plugin_service._subplugined[path]
+ parent = plugin_service.plugins.get(parent_id)
+ if parent and _uses_module_level(parent, path):
+ log.plugin.debug(f"parent {parent_id!r} uses {path!r} at module level, full tree reload")
+ return await reload_plugin(parent_id, parent.config.copy())
+ _conf = conf if conf is not None else plugin.config.copy()
+ log.plugin.debug(f"staged loading sub-plugin {path!r}, old sub-plugin keeps running until swap")
+ if not (mod := import_plugin(path, config=_conf, staged=True)):
+ log.plugin.error(f"failed to load staged sub-plugin {path!r}, old sub-plugin keeps running")
+ return False
+ if tasks := plugin.dispose(replacing=True):
+ await asyncio.wait(tasks)
+ new_plugin = mod.__plugin__ # type: ignore
+ promote_staged(new_plugin)
+ _handle_dependents(new_plugin)
+ return True
diff --git a/arclet/entari/plugin/model.py b/arclet/entari/plugin/model.py
index c66e495..4f0888b 100644
--- a/arclet/entari/plugin/model.py
+++ b/arclet/entari/plugin/model.py
@@ -1,5 +1,6 @@
from __future__ import annotations
+import ast
import asyncio
import inspect
import re
@@ -216,6 +217,12 @@ def wrapper(func: TCallable) -> TCallable:
return wrapper
+@dataclass(slots=True)
+class PluginInspect:
+ nodes: ast.Module
+ dump: str
+
+
@dataclass
class Plugin:
id: str
@@ -226,13 +233,18 @@ class Plugin:
is_static: bool = False
path: str = field(init=False)
uid: str | None = None
+ _inspect: PluginInspect | None = field(default=None, repr=False)
_metadata: PluginMetadata | None = None
_is_disposed: bool = False
_services: dict[str, Service] = field(init=False, default_factory=dict)
_config_key: str = field(init=False)
- # _scope: Scope = field(init=False)
_extra: dict[str, Any] = field(default_factory=dict, init=False) # extra metadata for inspection
- _apply: Callable[[Plugin], Any] | None = field(default=None, init=False)
+ _apply: Callable[[Plugin], Any] | None = field(default=None, init=False, repr=False)
+
+ @property
+ def bindings(self) -> dict[str, tuple[str, str | None]]:
+ """本插件模块导入层绑定的 {名字: (目标模块, 属性)};attr 为 None 表示模块绑定"""
+ return plugin_service.bindings.get(self.id, {})
@property
def reusable(self) -> bool:
@@ -271,6 +283,46 @@ def metadata(self, value: PluginMetadata):
):
value.config.__doc__ = value.description or value.name
+ def __post_init__(self):
+ uid_index = self.id.rfind("@")
+ self.path = self.id[:uid_index] if uid_index != -1 else self.id
+ self.uid = self.id[uid_index + 1 :] if uid_index != -1 else None
+ if self.id in plugin_service.plugins and not self.id.startswith("."):
+ # 原子重载暂存:id 冲突时注册进 _staged,scope 用唯一 id 并置 disabled
+ self._scope = _make_scope(self).of(f"{self.id}@staging")
+ self._scope.disable()
+ plugin_service._staged[self.id] = self # type: ignore
+ else:
+ self._scope = _make_scope(self).of(self.id)
+ plugin_service.plugins[self.id] = self # type: ignore
+ self.effect = self._scope.effect
+ self._config_key = self.config.pop("$path", self.id)
+ if filter_expr := self.config.get("$filter", ""):
+ self._scope.propagators.append(FilterPropagator(filter_expr))
+ # if self._metadata and self._metadata.depend_services:
+ # self._scope.propagators.append(inject(*self._metadata.depend_services, _is_global=True)) # type: ignore
+ # self._extra["injected_services"] = [
+ # s.id if isinstance(s, type) else s for s in self._metadata.depend_services
+ # ]
+ if "$disable" in self.config and isinstance(self.config["$disable"], str):
+
+ async def _check_reload(event: ConfigReload):
+ if event.scope == "basic":
+ self.check_disable()
+
+ sub = on(ConfigReload, _check_reload)
+ self.collect(sub.dispose)
+
+ self.is_static = self.config.pop("$static", False)
+ if self.id not in plugin_service._keep_values:
+ plugin_service._keep_values[self.id] = {}
+ if self.path not in plugin_service.referents:
+ plugin_service.referents[self.path] = set()
+ if self.path not in plugin_service.references:
+ plugin_service.references[self.path] = set()
+ plugin_service._unloaded.discard(self.id)
+ finalize(self, self.dispose, is_cleanup=True)
+
def exec_apply(self):
if not self._apply:
return
@@ -345,9 +397,9 @@ def disable(self):
continue
plugin_service.plugins[ret].disable()
tasks = set()
- t = self._clean_service()
- t.add_done_callback(tasks.discard)
- tasks.add(t)
+ if (t := self._clean_service()) is not None:
+ t.add_done_callback(tasks.discard)
+ tasks.add(t)
self._scope.disable()
if "$disable" not in self.config or isinstance(self.config["$disable"], bool):
self.config["$disable"] = True
@@ -388,40 +440,6 @@ def restore(self):
"""回收所有副作用"""
return self._scope._effect_manager.dispose()
- def __post_init__(self):
- uid_index = self.id.rfind("@")
- self.path = self.id[:uid_index] if uid_index != -1 else self.id
- self.uid = self.id[uid_index + 1 :] if uid_index != -1 else None
- self._scope = _make_scope(self).of(self.id)
- self.effect = self._scope.effect
- plugin_service.plugins[self.id] = self # type: ignore
- self._config_key = self.config.pop("$path", self.id)
- if filter_expr := self.config.get("$filter", ""):
- self._scope.propagators.append(FilterPropagator(filter_expr))
- # if self._metadata and self._metadata.depend_services:
- # self._scope.propagators.append(inject(*self._metadata.depend_services, _is_global=True)) # type: ignore
- # self._extra["injected_services"] = [
- # s.id if isinstance(s, type) else s for s in self._metadata.depend_services
- # ]
- if "$disable" in self.config and isinstance(self.config["$disable"], str):
-
- async def _check_reload(event: ConfigReload):
- if event.scope == "basic":
- self.check_disable()
-
- sub = on(ConfigReload, _check_reload)
- self.collect(sub.dispose)
-
- self.is_static = self.config.pop("$static", False)
- if self.id not in plugin_service._keep_values:
- plugin_service._keep_values[self.id] = {}
- if self.path not in plugin_service.referents:
- plugin_service.referents[self.path] = set()
- if self.path not in plugin_service.references:
- plugin_service.references[self.path] = set()
- plugin_service._unloaded.discard(self.id)
- finalize(self, self.dispose, is_cleanup=True)
-
def _clean_service(self):
manager = it(Launart)
@@ -432,24 +450,34 @@ def _gen(service: Service):
yield service
_services = [s for serv in self._services.values() for s in _gen(serv)]
+ if not _services:
+ return
+
+ async def _clean_one(service: Service):
+ if not manager.task_group:
+ return
+ plugin_service.service_waiter.clear(service.id)
+ if service.id not in manager.task_group.sideload_trackers:
+ return
+ try:
+ tracker = manager.task_group.sideload_trackers[service.id]
+ manager.remove_component(service)
+ await asyncio.wait([tracker, add_task(service.status.wait_for("finished"))])
+ except (ValueError, KeyError):
+ pass
async def _clean(services: list[Service]):
if not manager.task_group:
return
- for serv in services:
- plugin_service.service_waiter.clear(serv.id)
- if serv.id not in manager.task_group.sideload_trackers:
- continue
- try:
- tracker = manager.task_group.sideload_trackers[serv.id]
- manager.remove_component(serv)
- await asyncio.wait([tracker, add_task(serv.status.wait_for("finished"))])
- except (ValueError, KeyError):
- pass
+ await asyncio.gather(*(_clean_one(serv) for serv in services))
return add_task(_clean(_services))
- def dispose(self, *, is_cleanup: bool = False):
+ def dispose(self, *, is_cleanup: bool = False, replacing: bool = False):
+ """拆卸插件
+
+ replacing=True 表示是重载插件下的卸载(reload_plugin/reload_subplugin)
+ """
if not is_cleanup and self.is_static:
return # static plugin can only be disposed in cleanup phase
plugin_service._unloaded.add(self.id)
@@ -457,16 +485,14 @@ def dispose(self, *, is_cleanup: bool = False):
return
if not self.id.startswith(".") and self.id not in plugin_service._subplugined:
log.plugin.debug(f"disposing plugin {self.id}")
+ _was_staged = self.id in plugin_service._staged
self._is_disposed = True
tasks = set()
- t = self._clean_service()
- t.add_done_callback(tasks.discard)
- tasks.add(t)
+ if (t := self._clean_service()) is not None:
+ t.add_done_callback(tasks.discard)
+ tasks.add(t)
self._services.clear()
- if self.module.__spec__ and self.module.__spec__.cached:
- Path(self.module.__spec__.cached).unlink(missing_ok=True)
sys.modules.pop(self.module.__name__, None)
- tasks.update(self.restore())
delattr(self.module, "__plugin__")
if self.subplugins:
subplugs = [i.removeprefix(self.id)[1:] for i in self.subplugins]
@@ -474,17 +500,23 @@ def dispose(self, *, is_cleanup: bool = False):
log.plugin.trace(f"disposing sub-plugin {', '.join(subplugs)} of {self.id}")
for subplug in self.subplugins:
if subplug not in plugin_service.plugins:
- plugin_service._subplugined.pop(subplug, None)
+ if subplug in plugin_service._staged:
+ tasks.update(
+ plugin_service._staged[subplug].dispose(is_cleanup=is_cleanup, replacing=replacing)
+ )
+ else:
+ plugin_service._subplugined.pop(subplug, None)
continue
try:
- tasks.update(plugin_service.plugins[subplug].dispose(is_cleanup=is_cleanup))
+ tasks.update(plugin_service.plugins[subplug].dispose(is_cleanup=is_cleanup, replacing=replacing))
plugin_service._subplugined.pop(subplug, None)
except Exception as e:
log.plugin.error(f"failed to dispose sub-plugin {subplug} caused by {e!r}")
plugin_service.plugins.pop(subplug, None)
self.subplugins.clear()
- if not is_cleanup:
+ if not is_cleanup and not _was_staged:
publish(PluginUnloaded(self.id))
+ if not is_cleanup and not _was_staged and not replacing:
for ref in plugin_service.references.pop(self.path):
if ref not in plugin_service.plugins:
continue
@@ -505,7 +537,10 @@ def dispose(self, *, is_cleanup: bool = False):
except Exception as e:
log.plugin.error(f"failed to dispose referent plugin {ref} caused by {e!r}")
plugin_service.plugins.pop(ref, None)
- for ret in plugin_service.referents[self.path].copy():
+ # bindings-only 依赖方在卸载时同样需要停用,否则 A 永久卸载后其持有僵尸绑定继续运行
+ _dependents = set(plugin_service.referents[self.path])
+ _dependents.update(plugin_service.dependents_of(self.path, ensure=True))
+ for ret in _dependents:
if ret not in plugin_service.plugins:
continue
if (
@@ -515,10 +550,14 @@ def dispose(self, *, is_cleanup: bool = False):
):
continue
tasks.update(plugin_service.plugins[ret].disable())
- self._scope.dispose()
+ tasks.update(self._scope.dispose())
self._scope.propagators.clear()
- del plugin_service.plugins[self.id]
+ if self.id in plugin_service.plugins:
+ del plugin_service.plugins[self.id]
+ else:
+ plugin_service._staged.pop(self.id, None)
del self.module
+ del self._inspect
return tasks
def dispatch(self, event, name: str | None = None):
@@ -575,6 +614,12 @@ def service(self, serv: TS | type[TS]) -> TS:
plugin_service.service_waiter.assign(serv.id)
return serv
+ def restore_kept_state(self):
+ """重载后将保持的模块级可变对象重新绑定到模块 dict"""
+ for kept in plugin_service._keep_values.get(self.id, {}).values():
+ if kept.module_attr:
+ self.module.__dict__[kept.module_attr] = kept.obj
+
class RootlessPlugin(Plugin):
# fmt: off
@@ -615,9 +660,10 @@ def validate(self, func):
class KeepingVariable(Generic[T]):
- def __init__(self, obj: T, dispose: Callable[[T], None] | Callable[[T], Awaitable[None]] | None = None):
+ def __init__(self, obj: T, dispose=None, module_attr=None):
self.obj = obj
self._dispose = None
+ self.module_attr = module_attr
if hasattr(self.obj, "dispose"):
_dispose = self.obj.dispose.__func__ # type: ignore
if _is_awaitable(_dispose):
@@ -641,7 +687,7 @@ async def dispose(self):
# fmt: off
-def keeping(id_: str, obj: T | None = None, obj_factory: Callable[[], T] | None = None, dispose: Callable[[T], None] | Callable[[T], Awaitable[None]] | None = None) -> T: # noqa: E501
+def keeping(id_: str, obj: T | None = None, obj_factory: Callable[[], T] | None = None, dispose: Callable[[T], None] | Callable[[T], Awaitable[None]] | None = None, module_attr: str | None = None) -> T: # noqa: E501
# fmt: on
if not (plug := current_plugin.get(None)):
raise LookupError("no plugin context found")
@@ -650,5 +696,5 @@ def keeping(id_: str, obj: T | None = None, obj_factory: Callable[[], T] | None
raise ValueError("Either `obj` or `obj_factory` must be provided")
_obj = obj_factory() if obj_factory else obj
plug._extra.setdefault("kept_variables", []).append(id_)
- plugin_service._keep_values[plug.id][id_] = KeepingVariable(cast(T, _obj), dispose) # type: ignore
+ plugin_service._keep_values[plug.id][id_] = KeepingVariable(cast(T, _obj), dispose, module_attr) # type: ignore
return plugin_service._keep_values[plug.id][id_].obj # type: ignore
diff --git a/arclet/entari/plugin/model.pyi b/arclet/entari/plugin/model.pyi
index 8935786..536feb9 100644
--- a/arclet/entari/plugin/model.pyi
+++ b/arclet/entari/plugin/model.pyi
@@ -1,3 +1,4 @@
+import ast
import asyncio
from enum import Enum
from collections.abc import Callable
@@ -196,6 +197,11 @@ class PluginMetadata:
def inject(*services: type[Service] | str | DependService) -> Callable[[TCallable], TCallable]: ...
@overload
def inject(*services: type[Service] | str | DependService, _is_global: Literal[True]) -> Check: ...
+@dataclass(slots=True)
+class PluginInspect:
+ nodes: ast.Module
+ dump: str
+
@dataclass
class Plugin:
id: str
@@ -206,6 +212,7 @@ class Plugin:
is_static: bool = ...
path: str = ...
uid: str | None = ...
+ _inspect: PluginInspect | None = ...
_metadata: PluginMetadata | None = ...
_is_disposed: bool = ...
_services: dict[str, Service] = field(init=False, default_factory=dict)
@@ -214,6 +221,8 @@ class Plugin:
_extra: dict[str, Any] = field(default_factory=dict, init=False) # extra metadata for inspection
_apply: Callable[[Plugin], Any] | None = field(default=None, init=False)
+ @property
+ def bindings(self) -> dict[str, tuple[str, str | None]]: ...
@property
def reusable(self) -> bool: ...
@property
@@ -240,8 +249,8 @@ class Plugin:
def effect(self, execute: Callable[[], AsyncEffect], label: str = "") -> AsyncDisposable[Awaitable[None]]: ...
def collect(self, *disposes: Disposable | AsyncDisposable) -> Self: ...
def restore(self) -> set[asyncio.Task]: ...
- def _clean_service(self) -> asyncio.Task: ...
- def dispose(self, *, is_cleanup: bool = False) -> set[asyncio.Task]: ...
+ def _clean_service(self) -> asyncio.Task | None: ...
+ def dispose(self, *, is_cleanup: bool = False, replacing: bool = False) -> set[asyncio.Task]: ...
@overload
def dispatch(self, event: type[Resultable[T]], name: str | None = None) -> PluginDispatcher[T]: ...
@overload
@@ -346,6 +355,7 @@ class Plugin:
def proxy(self) -> ModuleType: ...
def subproxy(self, sub_id: str) -> ModuleType: ...
def service(self, serv: TS | type[TS]) -> TS: ...
+ def restore_kept_state(self) -> None: ...
class RootlessPlugin(Plugin):
@classmethod
@@ -361,15 +371,28 @@ class RootlessPlugin(Plugin):
class KeepingVariable(Generic[T]):
obj: T
_dispose: Callable[[T], Awaitable[None]] | None
- def __init__(self, obj: T, dispose: Callable[[T], None] | Callable[[T], Awaitable[None]] | None = None): ...
+ module_attr: str | None
+ def __init__(
+ self,
+ obj: T,
+ dispose: Callable[[T], None] | Callable[[T], Awaitable[None]] | None = None,
+ module_attr: str | None = None,
+ ): ...
async def dispose(self): ...
@overload
-def keeping(id_: str, obj: T, *, dispose: Callable[[T], None] | Callable[[T], Awaitable[None]] | None = None) -> T: ...
+def keeping(
+ id_: str,
+ obj: T,
+ *,
+ dispose: Callable[[T], None] | Callable[[T], Awaitable[None]] | None = None,
+ module_attr: str | None = None,
+) -> T: ...
@overload
def keeping(
id_: str,
*,
obj_factory: Callable[[], T],
dispose: Callable[[T], None] | Callable[[T], Awaitable[None]] | None = None,
+ module_attr: str | None = None,
) -> T: ...
diff --git a/arclet/entari/plugin/module.py b/arclet/entari/plugin/module.py
index 2e5b8eb..031aed9 100644
--- a/arclet/entari/plugin/module.py
+++ b/arclet/entari/plugin/module.py
@@ -5,11 +5,11 @@
import tokenize
from collections.abc import Sequence
from importlib import _bootstrap, _bootstrap_external # type: ignore
-from importlib.abc import MetaPathFinder
from importlib.machinery import ExtensionFileLoader, ModuleSpec, PathFinder, SourceFileLoader
from importlib.metadata import Distribution, PackageNotFoundError, distribution, distributions
from importlib.util import module_from_spec, resolve_name
from io import BytesIO
+from os import PathLike
from pathlib import Path
from types import ModuleType
from typing import Any
@@ -22,7 +22,7 @@
from ..event.plugin import PluginLoadedFailed, PluginLoadedSuccess
from ..exceptions import RegisterNotInPluginError, ReusablePluginError, StaticPluginDispatchError
from ..logger import log
-from .model import Plugin, PluginMetadata, current_plugin
+from .model import Plugin, PluginInspect, PluginMetadata, current_plugin
from .service import plugin_service
_SUBMODULE_WAITLIST: dict[str, set[str]] = {}
@@ -75,12 +75,33 @@ def _ensure_plugin(names: list[str], sub: bool, pid: str, pname: str, prefix="")
_IMPORTING.add(f"{prefix}{name}")
+def _resolve_from_target(node: ast.ImportFrom, pname: str, is_init: bool) -> str | None:
+ """from-import 的目标模块全限定名(相对导入按当前模块解析)"""
+ if node.level == 0:
+ return node.module
+ parts = pname.split(".")
+ pkg = parts if is_init else parts[:-1]
+ if node.level > len(pkg) + 1:
+ return None
+ base = pkg if node.level == 1 else pkg[: 1 - node.level]
+ if node.module:
+ return ".".join([*base, node.module])
+ return ".".join(base) if base else None
+
+
+def _record_binding(pid: str, name: str, target: str, attr: str | None):
+ """记录名字级 import 绑定(name → (target, attr)),供重载侧查询与改写"""
+ if not target:
+ return
+ plugin_service.bindings.setdefault(pid, {})[name] = (target, attr)
+
+
# fmt: off
class _Visitor(ast.NodeVisitor):
- def __init__(self, pid: str, pname: str, path: str, plg_lineno: list[int], sub_lineno: list[int], ns_lineno: list[int]): # noqa: E501
+ def __init__(self, pid: str, pname: str, path: bytes | str | PathLike[str], plg_lineno: list[int], sub_lineno: list[int], ns_lineno: list[int]): # noqa: E501
self.pid = pid
self.pname = pname
- self.path = path
+ self.path = path.decode() if isinstance(path, bytes) else f"{Path(path)}"
self.signed_plugin_lineno = plg_lineno
self.signed_subplugin_lineno = sub_lineno
self.signed_namespace_lineno = ns_lineno
@@ -96,6 +117,8 @@ def visit_Import(self, node: ast.Import):
if self._in_type_checking():
return
+ for alias in node.names:
+ _record_binding(self.pid, alias.asname or alias.name.split(".")[0], alias.name, None)
if node.lineno in self.signed_plugin_lineno or all(x.name in _ENSURE_IS_PLUGIN for x in node.names):
_ensure_plugin([alias.name for alias in node.names], False, self.pid, self.pname)
elif node.lineno in self.signed_subplugin_lineno or all(x.name in _SUBMODULE_WAITLIST.get(self.pname, ()) for x in node.names): # noqa: E501
@@ -105,6 +128,15 @@ def visit_ImportFrom(self, node: ast.ImportFrom):
name = self.pname
if self._in_type_checking():
return
+ target = _resolve_from_target(node, name, self.path.endswith("__init__.py"))
+ if target:
+ for alias in node.names:
+ if alias.name == "*":
+ continue
+ if node.level == 1 and node.module is None:
+ _record_binding(self.pid, alias.asname or alias.name, f"{target}.{alias.name}", None)
+ else:
+ _record_binding(self.pid, alias.asname or alias.name, target, alias.name)
if node.module is None: # from . import xxx
_ensure_plugin([alias.name for alias in node.names], node.lineno not in self.signed_plugin_lineno, self.pid, name, f"{name}.") # noqa: E501
elif node.level == 0: # from xxx import xxx
@@ -194,6 +226,8 @@ def __init__(self, fullname: str, path: str, plugin_id: str, parent_plugin_id: s
self.loaded = False
self.plugin_id = plugin_id
self.parent_plugin_id = parent_plugin_id
+ self._inspect: PluginInspect = None # type: ignore
+ self.staged = False
super().__init__(fullname, path)
def get_code(self, fullname):
@@ -204,13 +238,12 @@ def get_code(self, fullname):
"""
source_path = self.get_filename(fullname)
- source_bytes = None
- if source_bytes is None:
- source_bytes = self.get_data(source_path)
+ # --- SourceFileLoader's cache handler removed ---
+ source_bytes = self.get_data(source_path)
code_object = self.source_to_code(source_bytes, source_path)
return code_object
- def source_to_code(self, data, path=""):
+ def source_to_code(self, data, path="", *, _optimize: int = -1):
"""Return the code object compiled from source.
The 'data' argument can be any object type that compile() supports.
@@ -236,33 +269,39 @@ def source_to_code(self, data, path=""):
nodes = ast.parse(data, type_comments=True)
except SyntaxError:
return _bootstrap._call_with_frames_removed( # type: ignore
- compile, data, path, "exec", dont_inherit=True, optimize=-1
+ compile, data, path, "exec", dont_inherit=True, optimize=_optimize
)
visitor = _Visitor(self.plugin_id, name, path, plg_lineno, sub_lineno, ns_lineno)
visitor.visit(nodes)
-
+ self._inspect = PluginInspect(nodes, ast.dump(nodes, include_attributes=False))
return _bootstrap._call_with_frames_removed( # type: ignore
- compile, nodes, path, "exec", dont_inherit=True, optimize=-1
+ compile, nodes, path, "exec", dont_inherit=True, optimize=_optimize
)
def create_module(self, spec) -> ModuleType | None:
+ if self.staged:
+ return super().create_module(spec)
if self.name in plugin_service.plugins:
self.loaded = True
return plugin_service.plugins[self.name].proxy()
if self.name in plugin_service._subplugined:
self.loaded = True
return plugin_service.plugins[plugin_service._subplugined[self.name]].subproxy(self.name)
- if (
- any((k.startswith(self.name) and k.rfind("@") != -1) for k in plugin_service.plugins)
- and self.plugin_id.rfind("@") == -1
- ):
- raise ReusablePluginError(f"reusable plugin {self.name!r} cannot be imported directly")
+ _check_reusable(self.name, self.plugin_id)
return super().create_module(spec)
def exec_module(self, module: ModuleType, config: dict[str, Any] | None = None) -> None:
is_sub = False
- if plugin := plugin_service.plugins.get(self.parent_plugin_id) if self.parent_plugin_id else None:
- plugin.subplugins.append(self.plugin_id)
+ plugin = (
+ # 暂存期间 plugins 中仍是旧顶插件,须优先取 _staged 中的新插件,
+ # 否则新子插件的父链接指向旧插件
+ (plugin_service._staged.get(self.parent_plugin_id) or plugin_service.plugins.get(self.parent_plugin_id))
+ if self.parent_plugin_id
+ else None
+ )
+ if plugin:
+ if self.plugin_id not in plugin.subplugins:
+ plugin.subplugins.append(self.plugin_id)
plugin_service._subplugined[self.plugin_id] = plugin.id
is_sub = True
if config is None or not {k: v for k, v in config.items() if k not in ("$path", "$static")}:
@@ -305,7 +344,10 @@ def exec_module(self, module: ModuleType, config: dict[str, Any] | None = None)
if not plugin.is_static:
token1 = scope_ctx.set(plugin._scope)
try:
- super().exec_module(module)
+ code = self.get_code(module.__name__)
+ if code is None:
+ raise ImportError(f"cannot load module {module.__name__r} when get_code() returns None")
+ _bootstrap._call_with_frames_removed(exec, code, module.__dict__) # type: ignore
except RegisterNotInPluginError as e:
deleted = []
for frame in reversed(inspect.trace()):
@@ -322,7 +364,10 @@ def exec_module(self, module: ModuleType, config: dict[str, Any] | None = None)
_ensure_plugin(deleted[-1:], False, self.plugin_id, self.name)
_ENSURE_IS_PLUGIN.update(deleted[:-1])
try:
- super().exec_module(module)
+ code = self.get_code(module.__name__)
+ if code is None:
+ raise ImportError(f"cannot load module {module.__name__r} when get_code() returns None")
+ _bootstrap._call_with_frames_removed(exec, code, module.__dict__) # type: ignore
except Exception as e1:
if isinstance(e1, RegisterNotInPluginError):
log.plugin.error(f"failed to load plugin {self.plugin_id!r}:\n{e1.msg}")
@@ -335,7 +380,7 @@ def exec_module(self, module: ModuleType, config: dict[str, Any] | None = None)
if isinstance(e, (ImportError, StaticPluginDispatchError, ReusablePluginError)):
raise e1 from None
else:
- raise ImportError(f"{e1!r} in {self.name!r}", name=self.name, path=self.path) from None
+ raise ImportError(f"{e1!r} in {self.name!r}", name=self.name, path=self.path)
except Exception as e:
log.plugin.exception(f"failed to load plugin {self.plugin_id!r} caused by {e!r}", exc_info=e)
plugin.dispose()
@@ -343,7 +388,7 @@ def exec_module(self, module: ModuleType, config: dict[str, Any] | None = None)
if isinstance(e, (ImportError, StaticPluginDispatchError, ReusablePluginError)):
raise
else:
- raise ImportError(f"{e!r} in {self.name!r}", name=self.name, path=self.path) from None
+ raise ImportError(f"{e!r} in {self.name!r}", name=self.name, path=self.path)
finally:
# leave plugin context
delattr(module, "__cached__")
@@ -356,16 +401,20 @@ def exec_module(self, module: ModuleType, config: dict[str, Any] | None = None)
if metadata and not plugin.metadata:
plugin.metadata = metadata
plugin._apply = getattr(module, "__plugin_apply__", None)
+ plugin._inspect = self._inspect
+ plugin.restore_kept_state()
+ del self._inspect
+ staged = "staged " if self.staged else ""
if not is_sub:
if plugin._apply:
- log.plugin.success(f"loaded plugin {self.plugin_id!r} partially applied")
+ log.plugin.success(f"{staged}loaded plugin {self.plugin_id!r} partially applied")
else:
- log.plugin.success(f"loaded plugin {self.plugin_id!r}")
+ log.plugin.success(f"{staged}loaded plugin {self.plugin_id!r}")
else:
- log.plugin.trace(f"loaded sub-plugin {plugin.id!r} of {self.parent_plugin_id!r}")
+ log.plugin.trace(f"{staged}loaded sub-plugin {plugin.id!r} of {self.parent_plugin_id!r}")
if not plugin._apply:
publish(PluginLoadedSuccess(self.plugin_id))
- if plugin_service.status.blocking:
+ if plugin_service.status.blocking and not self.staged:
if plugin._apply:
plugin.exec_apply()
plugin.check_disable()
@@ -405,20 +454,46 @@ def _path_find_spec(fullname, path=None, target=None) -> ModuleSpec | None:
return spec
-class _PluginFinder(MetaPathFinder):
+def _as_plugin(
+ module_spec: ModuleSpec, fullname: str, module_origin: str, plugin_id: str, staged: bool = False
+) -> ModuleSpec:
+ loader = PluginLoader(fullname, module_origin, plugin_id)
+ loader.staged = staged
+ module_spec.loader = loader
+ return module_spec
+
+
+def _as_submodule(
+ module_spec: ModuleSpec, fullname: str, module_origin: str, plugin_id: str, parent: str, staged: bool = False
+) -> ModuleSpec:
+ loader = PluginLoader(fullname, module_origin, plugin_id, parent)
+ loader.staged = staged
+ module_spec.loader = loader
+ return module_spec
+
+
+def _check_reusable(name: str, plugin_id: str) -> None:
+ if any(k.startswith(name) and k.rfind("@") != -1 for k in plugin_service.plugins) and plugin_id.rfind("@") == -1:
+ raise ReusablePluginError(f"reusable plugin {name!r} cannot be imported directly")
+
+
+class _PluginFinder(PathFinder):
@classmethod
def find_spec(
cls,
fullname: str,
- path: Sequence[str] | None,
+ path: Sequence[str] | None = None,
target: ModuleType | None = None,
origin_id_: str | None = None,
- ):
+ force: bool = False,
+ staged: bool = False,
+ ) -> ModuleSpec | None:
# get the module spec using the default path-finder
module_spec = _path_find_spec(fullname, path, target)
if not module_spec:
return
module_origin = module_spec.origin
+ plugin_id = origin_id_ or fullname
# if the module has no origin, it might be a namespace package or a built-in module.
# We only care about namespace packages here, as built-in modules should not be treated as plugins.
# For namespace packages, we can still return the spec without modification,
@@ -433,30 +508,46 @@ def find_spec(
return
# current import statement is within a plugin.
if plug := current_plugin.get(None):
+ # only record outside import, as inside import are already recorded by the plugin's loader.
+ if module_spec.name != plug.module.__name__ and not module_spec.name.startswith(plug.module.__name__ + "."):
+ _record_binding(plug.id, module_spec.name.split(".")[0], module_spec.name, None)
# if the module being imported is the same as the plugin's module,
# return the plugin's module spec directly to avoid infinite recursion.
- if plug.module.__spec__ and plug.module.__spec__.origin == module_spec.origin:
+ if plug.module.__spec__ and plug.module.__spec__.origin == module_origin:
return plug.module.__spec__
# get the top-level plugin id (the parent) of the current plugin
- plugin_id = plug.id
- while plugin_id in plugin_service._subplugined:
- plugin_id = plugin_service._subplugined[plugin_id]
+ parent_id = plug.id
+ while parent_id in plugin_service._subplugined:
+ parent_id = plugin_service._subplugined[parent_id]
# if the module being imported is a submodule of the top-level plugin,
- if module_spec.name.startswith(plugin_service.plugins[plugin_id].module.__name__ + "."):
- module_spec.loader = PluginLoader(fullname, module_origin, origin_id_ or fullname, plugin_id)
- return module_spec
- # if the module being imported is in the waitlist of the top-level plugin,
+ # or if the module being imported is in the waitlist of the top-level plugin,
# it means it is marked as a submodule by the plugin author.
- if module_spec.name in _SUBMODULE_WAITLIST.get(plugin_id, ()):
- module_spec.loader = PluginLoader(fullname, module_origin, origin_id_ or fullname, plugin_id)
- # plugin_service.referents.setdefault(module_spec.name, set()).add(plug.id)
- # _SUBMODULE_WAITLIST[plug.module.__name__].remove(module_spec.name)
- return module_spec
+ if module_spec.name.startswith(
+ plugin_service.plugins[parent_id].module.__name__ + "."
+ ) or module_spec.name in _SUBMODULE_WAITLIST.get( # noqa: E501
+ parent_id, ()
+ ):
+ return _as_submodule(
+ module_spec,
+ fullname,
+ module_origin,
+ plugin_id,
+ parent_id,
+ staged=staged or plug.id in plugin_service._staged,
+ ) # noqa: E501
# in the following cases, the module is imported directly (probably from Entari App)
# 1. the module is already a plugin.
if module_spec.name in plugin_service.plugins:
- module_spec.loader = PluginLoader(fullname, module_origin, origin_id_ or fullname)
- return module_spec
+ if module_spec.name in plugin_service._subplugined:
+ return _as_submodule(
+ module_spec,
+ fullname,
+ module_origin,
+ plugin_id,
+ plugin_service._subplugined[module_spec.name],
+ staged=staged,
+ ) # noqa: E501
+ return _as_plugin(module_spec, fullname, module_origin, plugin_id, staged=staged)
# 2. the module is marked as a plugin by the plugin author, or followed the naming convention for plugins.
marked = (
module_spec.name in _ENSURE_IS_PLUGIN
@@ -487,7 +578,7 @@ def find_spec(
except (KeyError, ValueError):
pass
if marked:
- module_spec.loader = PluginLoader(fullname, module_origin, origin_id_ or fullname)
+ _as_plugin(module_spec, fullname, module_origin, plugin_id, staged=staged)
# if there already exists a plugin that is importing this module,
# we should add the plugin as a referent of this module
if plug:
@@ -495,33 +586,32 @@ def find_spec(
return module_spec
# 3. the module is marked as a submodule by other plugin, or it is a submodule of a plugin.
if module_spec.name in plugin_service._subplugined:
- module_spec.loader = PluginLoader(
- fullname, module_origin, origin_id_ or fullname, plugin_service._subplugined[module_spec.name]
- )
- return module_spec
+ return _as_submodule(
+ module_spec,
+ fullname,
+ module_origin,
+ plugin_id,
+ plugin_service._subplugined[module_spec.name],
+ staged=staged,
+ ) # noqa: E501
# 4. if the module is already a plugin, but it is assigned an unique id (usage of reusable plugin),
# it cannot be imported directly, otherwise it will break the uniqueness of the plugin instance.
- if (
- any(k.startswith(module_spec.name) and k.rfind("@") != -1 for k in plugin_service.plugins)
- and (origin_id_ or fullname).rfind("@") == -1
- ):
- raise ReusablePluginError(f"reusable plugin {module_spec.name!r} cannot be imported directly")
+ _check_reusable(module_spec.name, plugin_id)
# 5. the module is a submodule of a plugin, but it is not marked as a submodule by the plugin author,
- # we should still treat it as a submodule of the plugin to avoid breaking existing plugins
+ # we should still treat it as a submodule of the plugin to avoid breaking existing plugins.
+ # notice: cannot merge two conditions below, because some spec with submodule_search_locations (Namespace),
+ # their parent is the spec itself, not the parent module name.
if module_spec.parent and module_spec.parent in plugin_service.plugins:
- module_spec.loader = PluginLoader(fullname, module_origin, origin_id_ or fullname, module_spec.parent)
- return module_spec
- # 6. the module is a submodule of a plugin, but it is not marked as a submodule by the plugin author,
- # we should still treat it as a submodule of the plugin to avoid breaking existing plugins
- if module_spec.name.rpartition(".")[0] in plugin_service.plugins:
- module_spec.loader = PluginLoader(
- fullname, module_origin, origin_id_ or fullname, module_spec.name.rpartition(".")[0]
- )
- return module_spec
+ return _as_submodule(module_spec, fullname, module_origin, plugin_id, module_spec.parent, staged=staged)
+ if (parent_name := module_spec.name.rpartition(".")[0]) and parent_name in plugin_service.plugins:
+ return _as_submodule(module_spec, fullname, module_origin, plugin_id, parent_name, staged=staged)
+ # 6. force-wrap as a plugin when explicitly requested by import_plugin.
+ if force:
+ return _as_plugin(module_spec, fullname, module_origin, plugin_id, staged=staged)
return
-def find_spec(id_, package=None) -> ModuleSpec | None:
+def import_plugin(id_, package=None, config: dict | None = None, staged: bool = False) -> ModuleType | None:
uid_index = id_.rfind("@")
name = id_ if uid_index == -1 else id_[:uid_index]
fullname = resolve_name(name, package) if name.startswith(".") else name
@@ -542,20 +632,13 @@ def find_spec(id_, package=None) -> ModuleSpec | None:
if _current in plugin_service.plugins:
parent = plugin_service.plugins[_current].module
enter_plugin = True
- _current += "."
- continue
- if _current in _ENSURE_IS_PLUGIN:
- parent = import_plugin(_current)
- if parent:
+ elif _current in _ENSURE_IS_PLUGIN or enter_plugin:
+ if parent := import_plugin(_current):
enter_plugin = True
else:
parent = __import__(_current, fromlist=["__path__"])
- _current += "."
- continue
- if enter_plugin and (parent := import_plugin(_current)):
- pass
+ enter_plugin = False
else:
- enter_plugin = False
parent = __import__(_current, fromlist=["__path__"])
_current += "."
if parent is None:
@@ -566,27 +649,17 @@ def find_spec(id_, package=None) -> ModuleSpec | None:
parent_path = parent.__path__
else:
parent_path = None
- if isinstance(parent_path, _bootstrap_external._NamespacePath): # type: ignore
- parent_path = _NamespacePath(parent_path._name, parent_path._path, PathFinder._get_spec) # type: ignore
- if spec := _PluginFinder.find_spec(fullname, parent_path, origin_id_=id_):
- return spec
- module_spec = _path_find_spec(fullname, parent_path, None)
- if not module_spec:
- return
- module_origin = module_spec.origin
- if not module_origin:
- return
- if isinstance(module_spec.loader, ExtensionFileLoader):
+ spec = _PluginFinder.find_spec(fullname, parent_path, origin_id_=id_, force=True, staged=staged)
+ if not spec:
return
- module_spec.loader = PluginLoader(fullname, module_origin, id_)
- return module_spec
-
-
-def import_plugin(id_, package=None, config: dict | None = None):
- spec = find_spec(id_, package)
- if spec:
- mod = module_from_spec(spec)
- if spec.loader:
+ mod = module_from_spec(spec)
+ spec._initializing = True # type: ignore
+ try:
+ if spec.loader is None:
+ if spec.submodule_search_locations is None:
+ raise ImportError("missing loader", name=spec.name)
+ # A namespace package so do nothing.
+ else:
if isinstance(spec.loader, PluginLoader):
spec.loader.exec_module(mod, config=config)
protected_modules = set()
@@ -604,8 +677,10 @@ def import_plugin(id_, package=None, config: dict | None = None):
_IMPORTING.clear()
else:
spec.loader.exec_module(mod)
- return mod
- return
+ sys.modules[mod.__name__] = mod
+ finally:
+ spec._initializing = False # type: ignore
+ return mod
sys.meta_path.insert(0, _PluginFinder())
diff --git a/arclet/entari/plugin/service.py b/arclet/entari/plugin/service.py
index 680f8ab..02bc593 100644
--- a/arclet/entari/plugin/service.py
+++ b/arclet/entari/plugin/service.py
@@ -42,13 +42,26 @@ class PluginManagerService(Service):
id = "entari.plugin.manager"
plugins: dict[str, Plugin]
- _keep_values: dict[str, dict[str, KeepingVariable]]
+ """插件字典,键为插件ID,值为插件对象"""
referents: dict[str, set[str]]
+ """插件引用字典,键为插件ID,值为引用该插件的其他插件ID集合"""
references: dict[str, set[str]]
+ """插件被引用字典,键为插件ID,值为该插件引用的其他插件ID集合"""
+ bindings: dict[str, dict[str, tuple[str, str | None]]]
+ """插件导入绑定字典,键为插件ID,值为该插件中导入的其他插件的绑定信息 {名字: (目标模块, 属性)}"""
+ fingerprints: dict[str, str]
+ """插件指纹字典,键为插件ID,值为该插件的指纹字符串"""
+ service_waiter: ServiceWaiters
+ _keep_values: dict[str, dict[str, KeepingVariable]]
_direct_plugins: set[str]
+ """直接插件集合,存储所有直接加载(反过来即只由插件导入的插件)的插件ID"""
_unloaded: set[str]
+ """卸载插件集合,存储所有已卸载的插件ID"""
_subplugined: dict[str, str]
+ """子插件字典,键为子插件ID,值为父插件ID"""
_apply: dict[str, tuple[Callable[[dict[str, Any]], RootlessPlugin], bool]]
+ _staged: dict[str, Plugin]
+ """插件暂存"""
def __init__(self):
super().__init__()
@@ -60,6 +73,9 @@ def __init__(self):
self._unloaded = set()
self._subplugined = {}
self._apply = {}
+ self.bindings = {}
+ self.fingerprints = {}
+ self._staged = {}
self.service_waiter = ServiceWaiters()
@property
@@ -70,6 +86,59 @@ def required(self) -> set[str]:
def stages(self) -> set[Phase]:
return {"preparing", "cleanup", "blocking"}
+ def dependents_of(self, path: str, ensure: bool = True) -> list[str]:
+ """path(及其子树)的依赖方插件:直接导入者 + 经父包再导出链一层
+
+ 子树匹配使整树重载时,依赖子插件的下游插件同样被处理(子插件随树重建,绑定需重绑/级联)。
+ 插件(或其子模块)对自身子树的绑定属内部边,不构成依赖方,直接跳过。
+
+ Args:
+ path (str): 插件ID或其子模块路径
+ ensure (bool, optional): 是否确保返回的插件ID存在于已加载插件中. Defaults to True.
+
+ Returns:
+ list[str]: 依赖方插件ID列表
+ """
+ parent_pkg = path.rpartition(".")[0]
+ result: list[str] = []
+ for plug_id, bindings in self.bindings.items():
+ if plug_id == path or plug_id.startswith(path + "."):
+ continue
+ for name, (target, attr) in bindings.items():
+ if target == path or target.startswith(path + "."):
+ result.append(plug_id)
+ break
+ if parent_pkg and attr is not None and target == parent_pkg and parent_pkg in self.bindings:
+ parent_chain = self.bindings[parent_pkg]
+ if parent_chain.get(attr or name, (None, None))[0] == path:
+ result.append(plug_id)
+ break
+ if ensure:
+ result = [r for r in result if r in self.plugins]
+ return result
+
+ def topo_dependents(self, dependents: set[str]) -> list[str]:
+ """依赖方按 references 图拓扑排序:上游(被依赖者)优先于下游(依赖者)
+
+ 依赖者 C 若在 B 之前级联,会绑定旧 B,随后 B 重载时, C 已在 recursive_guard 中被跳过 → 静默。
+ 拓扑序保证 B 先重载。
+ """
+ ordered: list[str] = []
+ visited: set[str] = set()
+
+ def visit(dep_id: str):
+ if dep_id in visited:
+ return
+ visited.add(dep_id)
+ for ref in self.references.get(dep_id, ()):
+ if ref in dependents:
+ visit(ref)
+ ordered.append(dep_id)
+
+ for dep_id in sorted(dependents):
+ visit(dep_id)
+ return ordered
+
async def launch(self, manager: Launart):
servs = []
diff --git a/arclet/entari/plugin/swap.py b/arclet/entari/plugin/swap.py
new file mode 100644
index 0000000..4b22d99
--- /dev/null
+++ b/arclet/entari/plugin/swap.py
@@ -0,0 +1,596 @@
+import ast
+import builtins
+import copy
+import inspect
+import sys
+import types
+from collections.abc import Iterator, Sequence
+from dataclasses import dataclass
+from types import ModuleType
+from typing import Any
+
+from arclet.letoderea.scope import scope_ctx
+
+from ..logger import log
+from .model import Plugin, PluginInspect, current_plugin
+
+
+@dataclass(slots=True)
+class FunctionChange:
+ qualname: str
+ ordinal: int
+ count: int
+ node: ast.FunctionDef | ast.AsyncFunctionDef
+ signature_changed: bool
+ append: bool = False
+
+
+def list_dump(nodes: Sequence[ast.AST]) -> str:
+ """Copy from ast.py line 167-170"""
+ return f"[{', '.join(ast.dump(x, include_attributes=False) for x in nodes)}]"
+
+
+def iter_functions(body: list[ast.stmt]) -> Iterator[tuple[str, int, ast.FunctionDef | ast.AsyncFunctionDef]]:
+ """按 qualname 收集模块顶层函数与类方法(含嵌套类)"""
+
+ def _walk(body: list[ast.stmt], prefix: str) -> Iterator[tuple[str, int, ast.FunctionDef | ast.AsyncFunctionDef]]:
+ seen: dict[str, int] = {}
+ for stmt in body:
+ if isinstance(stmt, (ast.FunctionDef, ast.AsyncFunctionDef)):
+ qualname = f"{prefix}{stmt.name}"
+ ordinal = seen.get(stmt.name, 0)
+ seen[stmt.name] = ordinal + 1
+ yield qualname, ordinal, stmt
+ elif isinstance(stmt, ast.ClassDef):
+ yield from _walk(stmt.body, f"{prefix}{stmt.name}.")
+
+ yield from _walk(body, "")
+
+
+def _signature_names(node: ast.FunctionDef | ast.AsyncFunctionDef | ast.Lambda) -> set[str]:
+ names = {arg.arg for arg in node.args.posonlyargs + node.args.args + node.args.kwonlyargs}
+ if node.args.vararg:
+ names.add(node.args.vararg.arg)
+ if node.args.kwarg:
+ names.add(node.args.kwarg.arg)
+ return names
+
+
+def _comp_targets(gen: ast.comprehension) -> set[str]:
+ targets: set[str] = set()
+ stack: list[ast.AST] = [gen.target]
+ while stack:
+ node = stack.pop()
+ if isinstance(node, ast.Name):
+ targets.add(node.id)
+ elif isinstance(node, (ast.Tuple, ast.List)):
+ stack.extend(node.elts)
+ elif isinstance(node, ast.Starred):
+ stack.append(node.value)
+ return targets
+
+
+def _structurally_same(o: ast.stmt, n: ast.stmt) -> bool:
+ """结构比较:函数/方法体跳过(由 qualname 配对分析),其余须 dump 相等"""
+ if isinstance(o, (ast.FunctionDef, ast.AsyncFunctionDef)) and isinstance(
+ n, (ast.FunctionDef, ast.AsyncFunctionDef)
+ ):
+ return True
+
+ if isinstance(o, ast.ClassDef) and isinstance(n, ast.ClassDef):
+ if list_dump(o.decorator_list) != list_dump(n.decorator_list):
+ return False
+ if list_dump(o.bases) != list_dump(n.bases):
+ return False
+ if list_dump(o.keywords) != list_dump(n.keywords):
+ return False
+ if len(o.body) != len(n.body):
+ return False
+ return all(_structurally_same(x, y) for x, y in zip(o.body, n.body))
+ return ast.dump(o, include_attributes=False) == ast.dump(n, include_attributes=False)
+
+
+def _import_names(stmt: ast.stmt) -> dict[str, str]:
+ """import 语句绑定的 {名字: 目标模块}"""
+ if isinstance(stmt, ast.Import):
+ return {alias.asname or alias.name.split(".")[0]: alias.name for alias in stmt.names}
+ if isinstance(stmt, ast.ImportFrom):
+ module = "." * (stmt.level or 0) + (stmt.module or "")
+ return {alias.asname or alias.name: f"{module}.{alias.name}" for alias in stmt.names if alias.name != "*"}
+ return {}
+
+
+def _align_prefix(
+ old_body: list[ast.stmt], new_body: list[ast.stmt]
+) -> tuple[dict[str, str], dict[str, str], list[tuple[ast.stmt, ast.stmt]], list[ast.stmt]] | None:
+ """对齐新旧模块前缀:允许 import 语句就地变化/增删,其余语句须结构相同
+
+ 返回 (旧 import 绑定, 新 import 绑定, 对齐的非 import 语句对, 新尾部)。
+ None 表示结构无法匹配(需全量重载)。
+ """
+ old_imports: dict[str, str] = {}
+ new_imports: dict[str, str] = {}
+ pairs: list[tuple[ast.stmt, ast.stmt]] = []
+ i = j = 0
+ while i < len(old_body) and j < len(new_body):
+ o, n = old_body[i], new_body[j]
+ if isinstance(o, (ast.Import, ast.ImportFrom)) and isinstance(n, (ast.Import, ast.ImportFrom)):
+ old_imports.update(_import_names(o))
+ new_imports.update(_import_names(n))
+ i += 1
+ j += 1
+ elif isinstance(o, (ast.Import, ast.ImportFrom)):
+ old_imports.update(_import_names(o))
+ i += 1
+ elif isinstance(n, (ast.Import, ast.ImportFrom)):
+ new_imports.update(_import_names(n))
+ j += 1
+ elif _structurally_same(o, n):
+ pairs.append((o, n))
+ i += 1
+ j += 1
+ else:
+ return None
+ if i != len(old_body):
+ return None
+ return old_imports, new_imports, pairs, new_body[j:]
+
+
+def _changed_import_names(old_imports: dict[str, str], new_imports: dict[str, str]) -> set[str]:
+ """新旧 import 绑定的变更名集合(新增/移除/改绑)"""
+ common = set(old_imports) & set(new_imports)
+ return (set(old_imports) ^ set(new_imports)) | {n for n in common if old_imports[n] != new_imports[n]}
+
+
+class _GlobalNameCollector:
+ """作用域感知的名字收集器:收集会在模块全局或 builtins 解析的名字
+
+ 规则:Name(Load) 在自身与所有外层函数作用域均未绑定 → 全局名。
+ """
+
+ __slots__ = ("used", "stack")
+
+ def __init__(self):
+ self.used: set[str] = set()
+ self.stack: list[set[str]] = [set()]
+
+ def _bound(self, name: str) -> bool:
+ return any(name in scope for scope in self.stack)
+
+ def _bind(self, name: str):
+ self.stack[-1].add(name)
+
+ def collect(self, stmts: list[ast.stmt], initial: set[str] | None = None) -> set[str]:
+ self.used = set()
+ self.stack = [initial or set()]
+ for stmt in stmts:
+ self.visit_stmt(stmt)
+ return self.used
+
+ def visit_expr(self, expr: ast.expr):
+ if isinstance(expr, ast.Name):
+ if isinstance(expr.ctx, ast.Load) and not self._bound(expr.id):
+ self.used.add(expr.id)
+ elif isinstance(expr.ctx, ast.Store):
+ self._bind(expr.id)
+ elif isinstance(expr, ast.Lambda):
+ self.stack.append(_signature_names(expr))
+ self.visit_expr(expr.body)
+ self.stack.pop()
+ elif isinstance(expr, (ast.ListComp, ast.SetComp, ast.DictComp, ast.GeneratorExp)):
+ comps = expr.generators
+ if comps:
+ self.visit_expr(comps[0].iter)
+ self.stack.append({name for gen in comps for name in _comp_targets(gen)})
+ for gen in comps:
+ if gen is not comps[0]:
+ self.visit_expr(gen.iter)
+ for cond in gen.ifs:
+ self.visit_expr(cond)
+ if isinstance(expr, ast.DictComp):
+ self.visit_expr(expr.key)
+ self.visit_expr(expr.value)
+ else:
+ self.visit_expr(expr.elt)
+ self.stack.pop()
+ else:
+ for child in ast.iter_child_nodes(expr):
+ if isinstance(child, ast.expr):
+ self.visit_expr(child)
+ elif isinstance(child, ast.keyword) and child.value is not None:
+ self.visit_expr(child.value)
+
+ def visit_stmt(self, stmt: ast.stmt):
+ if isinstance(stmt, (ast.FunctionDef, ast.AsyncFunctionDef)):
+ self._bind(stmt.name)
+ for dec in stmt.decorator_list:
+ self.visit_expr(dec)
+ for default in [*stmt.args.defaults, *[d for d in stmt.args.kw_defaults if d]]:
+ self.visit_expr(default)
+ for arg in stmt.args.posonlyargs + stmt.args.args + stmt.args.kwonlyargs:
+ if arg.annotation:
+ self.visit_expr(arg.annotation)
+ if stmt.args.vararg and stmt.args.vararg.annotation:
+ self.visit_expr(stmt.args.vararg.annotation)
+ if stmt.args.kwarg and stmt.args.kwarg.annotation:
+ self.visit_expr(stmt.args.kwarg.annotation)
+ self.stack.append(_signature_names(stmt))
+ for body_stmt in stmt.body:
+ self.visit_stmt(body_stmt)
+ self.stack.pop()
+ elif isinstance(stmt, ast.ClassDef):
+ self._bind(stmt.name)
+ for dec in stmt.decorator_list:
+ self.visit_expr(dec)
+ for base in stmt.bases:
+ self.visit_expr(base)
+ for kw in stmt.keywords:
+ self.visit_expr(kw.value)
+ self.stack.append(set())
+ for body_stmt in stmt.body:
+ self.visit_stmt(body_stmt)
+ self.stack.pop()
+ elif isinstance(stmt, ast.Assign):
+ for target in stmt.targets:
+ self.visit_expr(target)
+ self.visit_expr(stmt.value)
+ elif isinstance(stmt, ast.AnnAssign):
+ self.visit_expr(stmt.target)
+ self.visit_expr(stmt.annotation)
+ if stmt.value:
+ self.visit_expr(stmt.value)
+ elif isinstance(stmt, ast.AugAssign):
+ self.visit_expr(stmt.target)
+ self.visit_expr(stmt.value)
+ elif isinstance(stmt, (ast.For, ast.AsyncFor)):
+ self.visit_expr(stmt.target)
+ self.visit_expr(stmt.iter)
+ for body_stmt in [*stmt.body, *stmt.orelse]:
+ self.visit_stmt(body_stmt)
+ elif isinstance(stmt, (ast.With, ast.AsyncWith)):
+ for item in stmt.items:
+ self.visit_expr(item.context_expr)
+ if item.optional_vars:
+ self.visit_expr(item.optional_vars)
+ for body_stmt in stmt.body:
+ self.visit_stmt(body_stmt)
+ elif isinstance(stmt, ast.Try) or (sys.version_info >= (3, 11) and isinstance(stmt, ast.TryStar)):
+ for body_stmt in [*stmt.body, *stmt.orelse, *stmt.finalbody]:
+ self.visit_stmt(body_stmt)
+ for handler in stmt.handlers:
+ if handler.type:
+ self.visit_expr(handler.type)
+ if handler.name:
+ self._bind(handler.name)
+ for body_stmt in handler.body:
+ self.visit_stmt(body_stmt)
+ elif isinstance(stmt, ast.Import):
+ for alias in stmt.names:
+ self._bind(alias.asname or alias.name.split(".")[0])
+ elif isinstance(stmt, ast.ImportFrom):
+ for alias in stmt.names:
+ self._bind(alias.asname or alias.name)
+ elif isinstance(stmt, ast.If):
+ self.visit_expr(stmt.test)
+ for body_stmt in [*stmt.body, *stmt.orelse]:
+ self.visit_stmt(body_stmt)
+ elif isinstance(stmt, ast.While):
+ self.visit_expr(stmt.test)
+ for body_stmt in [*stmt.body, *stmt.orelse]:
+ self.visit_stmt(body_stmt)
+ else:
+ for child in ast.iter_child_nodes(stmt):
+ if isinstance(child, ast.expr):
+ self.visit_expr(child)
+ elif isinstance(child, ast.stmt):
+ self.visit_stmt(child)
+
+
+def _collect_unchanged_references(
+ pairs: list[tuple[ast.stmt, ast.stmt]], changes: list[FunctionChange], new_nodes: ast.Module
+) -> set[str]:
+ """收集未变更代码(顶层语句、类级语句、未变更函数/方法体)引用的全局名"""
+ collector = _GlobalNameCollector()
+ used: set[str] = set()
+ for _, new_stmt in pairs:
+ if isinstance(new_stmt, (ast.Import, ast.ImportFrom, ast.FunctionDef, ast.AsyncFunctionDef)):
+ continue
+ if isinstance(new_stmt, ast.ClassDef):
+ class_stmts = [s for s in new_stmt.body if not isinstance(s, (ast.FunctionDef, ast.AsyncFunctionDef))]
+ if class_stmts:
+ used |= collector.collect(class_stmts)
+ continue
+ used |= collector.collect([new_stmt])
+ changed_nodes = {c.node for c in changes}
+ for _, _, fn in iter_functions(new_nodes.body):
+ if fn in changed_nodes:
+ continue
+ used |= collector.collect(fn.body, _signature_names(fn))
+ return used
+
+
+def classify(old_nodes: ast.Module, new_nodes: ast.Module) -> list[FunctionChange] | None:
+ """比较新旧模块的 AST,返回可以热替换的函数列表
+
+ 函数配对:先按(名称组, 完整节点 dump 相同)锚定未变化函数(容忍位置交换/重排),
+ 剩余函数按组内顺序配对比较(容忍原地编辑);重命名(组键变化)与增删(组内数量不匹配)走全量重载。
+ 文件底部追加的新函数 走 append 路径(整句执行,含装饰器注册,要求旧 body 是新的结构前缀且尾部新增全为函数定义)。
+ 前缀中的 import 语句允许就地变化/增删,条件是变更的绑定名不被任何未变更代码引用(否则走全量重载)。
+
+ Returns:
+ 可就地替换的函数列表,None 表示模块整体需要全量重载
+ """
+ old_body = old_nodes.body
+ new_body = new_nodes.body
+ aligned = _align_prefix(old_body, new_body)
+ if aligned is None:
+ return None
+ old_imports, new_imports, pairs, tail = aligned
+ if tail and not all(
+ isinstance(s, (ast.FunctionDef, ast.AsyncFunctionDef, ast.Import, ast.ImportFrom)) for s in tail
+ ):
+ return None
+ old_groups: dict[str, list[tuple[int, ast.FunctionDef | ast.AsyncFunctionDef]]] = {}
+ new_groups: dict[str, list[tuple[int, ast.FunctionDef | ast.AsyncFunctionDef]]] = {}
+ for qualname, ordinal, fn in iter_functions([o for o, _ in pairs]):
+ old_groups.setdefault(qualname, []).append((ordinal, fn))
+ for qualname, ordinal, fn in iter_functions([n for _, n in pairs]):
+ new_groups.setdefault(qualname, []).append((ordinal, fn))
+ if set(old_groups.keys()) != set(new_groups.keys()):
+ return None
+ changes: list[FunctionChange] = []
+ for qualname, old_list in old_groups.items():
+ new_list = new_groups[qualname]
+ if len(old_list) != len(new_list):
+ return None
+ used_old: set[int] = set()
+ used_new: set[int] = set()
+ for i, (_, o_node) in enumerate(old_list):
+ for j, (_, n_node) in enumerate(new_list):
+ if (
+ i not in used_old
+ and j not in used_new
+ and ast.dump(o_node, include_attributes=False) == ast.dump(n_node, include_attributes=False)
+ ):
+ used_old.add(i)
+ used_new.add(j)
+ break
+ rest_old = [old_list[i] for i in range(len(old_list)) if i not in used_old]
+ rest_new = [new_list[j] for j in range(len(new_list)) if j not in used_new]
+ for (o_ord, o_node), (n_ord, n_node) in zip(rest_old, rest_new):
+ if list_dump(n_node.decorator_list) != list_dump(o_node.decorator_list):
+ return None
+ signature_changed = n_node.__class__ is not o_node.__class__ or ast.dump(
+ n_node.args, include_attributes=False
+ ) != ast.dump(o_node.args, include_attributes=False)
+ # if signature_changed or list_dump(n_node.body) != list_dump(o_node.body):
+ changes.append(FunctionChange(qualname, o_ord, len(old_list), n_node, signature_changed))
+ for stmt in tail:
+ if isinstance(stmt, (ast.FunctionDef, ast.AsyncFunctionDef)):
+ changes.append(FunctionChange(stmt.name, 0, 1, stmt, signature_changed=False, append=True))
+ if changed_import_names := _changed_import_names(old_imports, new_imports):
+ used = _collect_unchanged_references(pairs, changes, new_nodes)
+ if changed_import_names & used:
+ log.plugin.debug(
+ f"changed import names {sorted(changed_import_names & used)!r} are referenced "
+ f"by unchanged code, fallback to full reload"
+ )
+ return None
+ return changes
+
+
+def collect_global_names(fn: ast.FunctionDef | ast.AsyncFunctionDef) -> set[str]:
+ """收集新函数体中所有会在模块全局或 builtins 解析的名字(含嵌套函数)"""
+ collector = _GlobalNameCollector()
+ return collector.collect(fn.body, initial=_signature_names(fn))
+
+
+def _unwrap_fn(target: Any) -> types.FunctionType | None:
+ """从模块/类绑定或 Subscriber 中取出裸函数对象"""
+ if isinstance(target, types.FunctionType):
+ return target
+ if isinstance(target, (classmethod, staticmethod)):
+ inner = target.__func__
+ return inner if isinstance(inner, types.FunctionType) else None
+ if hasattr(target, "callable_target"):
+ inner = getattr(target, "callable_target")
+ return inner if isinstance(inner, types.FunctionType) else None
+ return None
+
+
+def _resolve_old_fn(plugin: Plugin, qualname: str, ordinal: int, count: int) -> types.FunctionType | None:
+ """按 (qualname, 序号) 解析活模块中的旧函数对象;返回 None 表示不可就地替换
+
+ 最后一个同名定义经模块/类绑定解析(模块执行时同名后者胜出);
+ 非最后一个同名定义只能经插件 scope 的 Subscriber 按名字+注册序解析
+ (即 `_` 惯用法:多个 `@plugin.listen` 装饰的同名函数)。
+ """
+ module = plugin.module
+ parts = qualname.split(".")
+ if len(parts) > 1:
+ if ordinal != count - 1:
+ return None
+ parent = module.__dict__.get(parts[0])
+ if parent is None:
+ return None
+ for part in parts[1:-1]:
+ parent = inspect.getattr_static(parent, part, None)
+ if parent is None:
+ return None
+ return _unwrap_fn(inspect.getattr_static(parent, parts[-1], None))
+ if ordinal == count - 1:
+ return _unwrap_fn(module.__dict__.get(parts[0]))
+ candidates = [
+ slot.subscriber.callable_target
+ for slot in plugin._scope.subscribers
+ if isinstance(slot.subscriber.callable_target, types.FunctionType)
+ and slot.subscriber.callable_target.__name__ == parts[0]
+ ]
+ if len(candidates) != count:
+ return None
+ return candidates[ordinal]
+
+
+def _build_new_fn(new_node: ast.FunctionDef | ast.AsyncFunctionDef, module: ModuleType) -> types.FunctionType | None:
+ """剥掉装饰器后在临时命名空间执行新函数定义,返回新函数对象"""
+ node = copy.deepcopy(new_node)
+ node.decorator_list = []
+ ast.fix_missing_locations(node)
+ try:
+ code = compile(
+ ast.Module(body=[node], type_ignores=[]),
+ module.__name__,
+ "exec",
+ dont_inherit=True,
+ optimize=-1,
+ )
+ ns = dict(module.__dict__)
+ exec(code, ns)
+ except Exception as e:
+ log.plugin.error(f"failed to build new function {node.name!r}: {e!r}")
+ return None
+ fn = ns.get(node.name)
+ return fn if isinstance(fn, types.FunctionType) else None
+
+
+def _decorator_global_names(node: ast.FunctionDef | ast.AsyncFunctionDef) -> set[str]:
+ """装饰器表达式中的全局引用(append 路径整句执行时需解析)"""
+ used: set[str] = set()
+ for dec in node.decorator_list:
+ for name in ast.walk(dec):
+ if isinstance(name, ast.Name) and isinstance(name.ctx, ast.Load):
+ used.add(name.id)
+ return used
+
+
+def _exec_imports(plugin: Plugin, nodes: ast.Module) -> bool:
+ """将新源码中的 import 语句按源码顺序执行并写入模块中"""
+ module = plugin.module
+ for stmt in nodes.body:
+ if not isinstance(stmt, (ast.Import, ast.ImportFrom)):
+ continue
+ stmt_cp = copy.deepcopy(stmt)
+ ast.fix_missing_locations(stmt_cp)
+ code = compile(
+ ast.Module(body=[stmt_cp], type_ignores=[]),
+ module.__name__,
+ "exec",
+ dont_inherit=True,
+ optimize=-1,
+ )
+ try:
+ exec(code, module.__dict__)
+ except Exception as e:
+ log.plugin.error(f"failed to exec import statement at line {stmt.lineno}: {e!r}")
+ return False
+ return True
+
+
+def _exec_append(plugin: Plugin, node: ast.FunctionDef | ast.AsyncFunctionDef) -> bool:
+ """在插件上下文(current_plugin + scope)中整句执行新增语句,完成装饰器注册"""
+ module = plugin.module
+ stmt = copy.deepcopy(node)
+ ast.fix_missing_locations(stmt)
+ code = compile(
+ ast.Module(body=[stmt], type_ignores=[]),
+ module.__name__,
+ "exec",
+ dont_inherit=True,
+ optimize=-1,
+ )
+ token = current_plugin.set(plugin)
+ try:
+ if not plugin.is_static:
+ token1 = scope_ctx.set(plugin._scope)
+ try:
+ exec(code, module.__dict__)
+ finally:
+ scope_ctx.reset(token1)
+ else:
+ exec(code, module.__dict__)
+ finally:
+ current_plugin.reset(token)
+ return True
+
+
+def swap_functions(plugin: Plugin, new_nodes: ast.Module, changes: list[FunctionChange]) -> bool:
+ """就地替换函数实现,如果会影响插件自身或下游依赖方则返回 False,调用方走全量重载"""
+ module = plugin.module
+ import_names = {
+ name
+ for stmt in new_nodes.body
+ if isinstance(stmt, (ast.Import, ast.ImportFrom))
+ for name in _import_names(stmt)
+ }
+ resolved: list[tuple[types.FunctionType, types.FunctionType, FunctionChange]] = []
+ for change in changes:
+ new_fn = change.node
+ free_names = collect_global_names(new_fn)
+ missing = {
+ name
+ for name in free_names
+ if name not in module.__dict__ and name not in vars(builtins) and name not in import_names
+ }
+ if missing:
+ log.plugin.warning(
+ f"cannot hot-swap {change.qualname!r}: free names {missing!r} missing, "
+ "fallback to full reload"
+ )
+ return False
+ if change.append:
+ free_names |= _decorator_global_names(new_fn)
+ missing = {
+ name
+ for name in free_names
+ if name not in module.__dict__ and name not in vars(builtins) and name not in import_names
+ }
+ if missing:
+ log.plugin.warning(
+ f"cannot append {change.qualname!r}: free names {missing!r} missing, "
+ "fallback to full reload"
+ )
+ return False
+ continue
+ old_fn = _resolve_old_fn(plugin, change.qualname, change.ordinal, change.count)
+ if old_fn is None:
+ log.plugin.warning(f"cannot resolve {change.qualname!r} in module, fallback to full reload")
+ return False
+ if change.signature_changed:
+ from ..command import _commands
+
+ if any(sub.callable_target is old_fn for sub in _commands.subscribers.values()):
+ log.plugin.warning(
+ f"signature of command target {change.qualname!r} changed, " "fallback to full reload"
+ )
+ return False
+ new_function = _build_new_fn(new_fn, module)
+ if new_function is None:
+ return False
+ resolved.append((old_fn, new_function, change))
+ if not _exec_imports(plugin, new_nodes):
+ return False
+ for old_fn, new_function, change in resolved:
+ old_fn.__code__ = new_function.__code__
+ old_fn.__defaults__ = new_function.__defaults__
+ old_fn.__kwdefaults__ = new_function.__kwdefaults__
+ old_fn.__annotations__ = new_function.__annotations__
+ for slot in plugin._scope.subscribers:
+ if slot.subscriber.callable_target is not old_fn:
+ continue
+ sub = slot.subscriber
+ sub.cancel_running()
+ if change.signature_changed:
+ sub.callable_target = new_function
+ try:
+ sub._recompile()
+ except Exception as e:
+ log.plugin.error(f"failed to recompile subscriber of {change.qualname!r}: {e!r}")
+ for change in changes:
+ if change.append:
+ try:
+ _exec_append(plugin, change.node)
+ except Exception as e:
+ log.plugin.error(f"failed to append {change.qualname!r}: {e!r}")
+ return False
+ plugin._inspect = PluginInspect(new_nodes, ast.dump(new_nodes, include_attributes=False))
+ return True
diff --git a/arclet/entari/session.py b/arclet/entari/session.py
index 854c2b7..45d52a0 100644
--- a/arclet/entari/session.py
+++ b/arclet/entari/session.py
@@ -1,6 +1,9 @@
import asyncio
+import inspect
import secrets
from collections.abc import Awaitable, Callable, Iterable
+from functools import wraps
+from types import MethodType
from typing import Any, Generic, NoReturn, cast, overload
from typing_extensions import TypeVar
@@ -26,6 +29,7 @@
from . import command
from .config import EntariConfig
+from .event.api import APIRequest, APIResponse, SendRequest, SendResponse
from .event.base import (
FriendRequestEvent,
GuildMemberRequestEvent,
@@ -35,7 +39,6 @@
Reply,
SatoriEvent,
)
-from .event.send import SendRequest, SendResponse
from .message import MessageChain, Render
TEvent = TypeVar("TEvent", bound=SatoriEvent, default=SatoriEvent)
@@ -55,14 +58,51 @@ async def rule(elem: Element, sess: "Session"):
return await content.transform_async(rule, session)
+STATIC_METHODS = frozenset(
+ {
+ "__init__",
+ "call_api",
+ "request_internal",
+ "send",
+ "send_message",
+ "send_private_message",
+ "update_message",
+ "message_create",
+ }
+)
+
+
class EntariProtocol(ApiProtocol):
# fmt: off
+ def __init__(self, account: Account["EntariProtocol"]):
+ super().__init__(account)
+ funcs = inspect.getmembers(self, predicate=lambda x: isinstance(x, MethodType))
+ for name, func in funcs:
+ if name in STATIC_METHODS:
+ continue
+
+ @wraps(func)
+ async def wrapper(*args, _func=func, _sig=inspect.signature(func), **kwargs):
+ bounds = _sig.bind(*args, **kwargs)
+ bounds.apply_defaults()
+ try:
+ if result := await es.post(APIRequest(self.account, _func.__name__, bounds.arguments)):
+ ans = result.value
+ else:
+ ans = await _func(**bounds.arguments)
+ success = True
+ except Exception as e:
+ ans = e
+ success = False
+ await es.publish(APIResponse(self.account, _func.__name__, bounds.arguments, success, ans))
+ return ans
+ setattr(self, name, wrapper)
+
async def send_message(self, channel: str | Channel, message: str | Iterable[str | Element], at_sender: At | None = None, reply_to: Quote | None = None, referrer: dict[str, Any] | None = None) -> list[MessageObject]: # noqa: E501
"""发送消息。返回一个 `MessageReceipt` 对象构成的数组。
- Args:
- channel (str | Channel): 要发送的频道 ID
+ Args: channel (str | Channel): 要发送的频道 ID
message (str | Iterable[str | Element]): 要发送的消息
at_sender (At | None): 是否 @ 发送者,默认为 None
reply_to (Quote | None): 是否作为回复发送,默认为 None
@@ -123,7 +163,7 @@ async def message_create(self, channel_id: str, content: str | Iterable[str | El
msg = await component_transform(sess, msg)
referrer = {k: v for k, v in referrer.items() if k != "source"}
sess.elements = msg
- btns = select(msg, Button)
+ btns = select(msg.content, Button)
for btn in btns:
if btn.type != "link" and not btn.id:
btn.id = secrets.token_urlsafe(16)
diff --git a/entari.schema.json b/entari.schema.json
index 46c7d76..8872b08 100644
--- a/entari.schema.json
+++ b/entari.schema.json
@@ -77,6 +77,23 @@
"default": false,
"description": "是否利用元数据进行插件导入检测(可能会增加启动时间)",
"title": "Check Metadata"
+ },
+ "str_as_message": {
+ "type": "boolean",
+ "default": true,
+ "description": "发送字符串时是否自动转换为消息链",
+ "title": "Str As Message"
+ },
+ "superusers": {
+ "type": "object",
+ "additionalProperties": {
+ "type": "array",
+ "items": {
+ "type": "string"
+ }
+ },
+ "description": "超级用户配置,键为平台名称,值为该平台的超级用户 ID 列表",
+ "title": "Superusers"
}
},
"additionalProperties": false,
@@ -102,7 +119,14 @@
"title": "Host"
},
"port": {
- "type": "integer",
+ "oneOf": [
+ {
+ "type": "string"
+ },
+ {
+ "type": "integer"
+ }
+ ],
"default": 5140,
"description": "服务器端口",
"title": "Port"
@@ -156,7 +180,14 @@
"title": "Host"
},
"port": {
- "type": "integer",
+ "oneOf": [
+ {
+ "type": "string"
+ },
+ {
+ "type": "integer"
+ }
+ ],
"default": 8080,
"description": "本机 Webhook 服务器端口",
"title": "Port"
@@ -185,9 +216,16 @@
"title": "Server Host"
},
"server_port": {
- "type": "integer",
+ "oneOf": [
+ {
+ "type": "string"
+ },
+ {
+ "type": "integer"
+ }
+ ],
"default": 5140,
- "description": "发送请求的目标服务器端口",
+ "description": "发送请求的目标服务器端port",
"title": "Server Port"
},
"server_path": {
@@ -306,14 +344,21 @@
"type": "string"
},
"plugins": {
- "description": "List of plugins under the prefix",
- "items": {
- "type": "string",
- "description": "Plugin name"
- },
- "title": "Plugins",
- "type": "array",
- "uniqueItems": true
+ "anyOf": [
+ {
+ "type": "string"
+ },
+ {
+ "items": {
+ "type": "string",
+ "description": "Plugin name"
+ },
+ "type": "array",
+ "uniqueItems": true
+ }
+ ],
+ "description": "List of plugins under the prefix, or select an item of $files to apply plugins",
+ "title": "Plugins"
}
},
"required": [
@@ -425,6 +470,12 @@
"description": "是否记录发送的消息",
"title": "Record Send"
},
+ "short_message": {
+ "type": "boolean",
+ "default": false,
+ "description": "是否在日志中使用简短的消息内容",
+ "title": "Short Message"
+ },
"$disable": {
"type": "string",
"description": "Expression for whether disable this plugin"
@@ -443,9 +494,13 @@
},
".scheduler": {
"type": "object",
- "description": "Simple Scheduler with interval / crontab task; no configuration required",
- "additionalProperties": true,
+ "title": "_SchedulerConf",
"properties": {
+ "debug": {
+ "type": "boolean",
+ "default": true,
+ "title": "Debug"
+ },
"$disable": {
"type": "string",
"description": "Expression for whether disable this plugin"
@@ -458,7 +513,9 @@
"type": "string",
"description": "Plugin filter expression, which will be evaluated in the context of the plugin"
}
- }
+ },
+ "additionalProperties": false,
+ "description": "Simple Scheduler with interval / crontab task"
},
"reusable@1": {
"type": "object",
diff --git a/entari.yml b/entari.yml
index 5c21ebd..3836cc7 100644
--- a/entari.yml
+++ b/entari.yml
@@ -42,9 +42,7 @@ plugins:
output: bar!
example_plugin2:
$filter: >
- (platform in ("onebot", "milky") and member.roles[0].id ne 'member')
- or
- platform eq "console"
+ (platform in ("onebot", "milky") and member.roles[0].id ne 'member') or platform eq "console"
example_plugin3:
$filter: message == "test_plugin3"
?example_plugin4: {}
diff --git a/example_plugin.py b/example_plugin.py
index f51ddf8..7232de0 100644
--- a/example_plugin.py
+++ b/example_plugin.py
@@ -32,7 +32,7 @@ async def cleanup():
@plug.dispatch(MessageCreatedEvent)
@filter_.public
-async def _(session: Session):
+async def invoke_test(session: Session):
if session.content == "test":
resp = await session.send("This message will recall in 5s...", at_sender=True)
@@ -49,13 +49,13 @@ async def filter_content(session: Session):
@disp_message.on().if_(filter_content)
-async def _(session: Session):
+async def filter_content1(session: Session):
return await session.send("Filter: public message, to me, and content is 'aaa'")
@disp_message
@filter_.public & filter_.to_me & filter_(lambda sess: str(sess.content) != "aaa")
-async def _(session: Session):
+async def filter_content2(session: Session):
return await session.send("Filter: public message, to me, but content is not 'aaa'")
diff --git a/example_plugins/example_plugin5/foo/bar.py b/example_plugins/example_plugin5/foo/bar.py
index 23f5ad6..58e104f 100644
--- a/example_plugins/example_plugin5/foo/bar.py
+++ b/example_plugins/example_plugin5/foo/bar.py
@@ -6,6 +6,6 @@
metadata(__file__, description="A test plugin 5")
-@command.on("exam5 ")
-def exam5(x: int, y: int):
+@command.on("exam5 {x} {y}")
+async def exam5(x: int, y: int):
return f"example_plugin5: {x} * {y} = {calc(x, y)}"
diff --git a/example_plugins/example_plugin8.py b/example_plugins/example_plugin8.py
new file mode 100644
index 0000000..7ae902c
--- /dev/null
+++ b/example_plugins/example_plugin8.py
@@ -0,0 +1,35 @@
+from arclet.entari.filter.message import startswith, regexmatch, regex_origin
+from arclet.entari import MessageCreatedEvent, MessageChain, Session, listen, Image, Text, User
+
+
+@listen(MessageCreatedEvent).if_(startswith("!hello"))
+async def hello_listener1(sess: Session, message: MessageChain, user: User):
+ await sess.send(f"Hello! This is a response from the hello_listener. {user}")
+ await sess.send(message)
+
+
+@listen(MessageCreatedEvent).if_(startswith(Image, include=True))
+async def image_listener(sess: Session, message: MessageChain):
+ await sess.send("Hello! This is a response from the image_listener.")
+ await sess.send(message)
+
+
+@listen(MessageCreatedEvent).if_(startswith("!world", bind="world"))
+async def hello_listener2(sess: Session, message: MessageChain, world: MessageChain):
+ await sess.send("Hello! This is a response from the hello_listener2.")
+ await sess.send(message)
+ await sess.send(world)
+
+
+@listen(MessageCreatedEvent).if_(regexmatch(r"test (\d+)", flags=2))
+async def regex_listener(
+ sess: Session,
+ message: MessageChain,
+ match = regex_origin(),
+ group1: str = regex_origin().group(1),
+ dicts: dict = regex_origin().groupdict(),
+):
+ await sess.send(f"Hello! This is a response from the regex_listener. You said: {message}")
+ await sess.send(f"Matched: {Text(str(match))}")
+ await sess.send(f"Matched group 1: {group1}")
+ await sess.send(f"Matched dicts: {dicts}")
diff --git a/pdm.lock b/pdm.lock
index 57651c3..eaae681 100644
--- a/pdm.lock
+++ b/pdm.lock
@@ -5,7 +5,7 @@
groups = ["default", "cli", "cron", "dev", "dotenv", "full", "msgspec", "pydantic", "reload", "rich", "toml", "yaml"]
strategy = ["inherit_metadata"]
lock_version = "4.5.0"
-content_hash = "sha256:ff5971fff00cca2f21031effa0c543be4be9b16d453f356335f0dbdc7d66a559"
+content_hash = "sha256:f182f588fcce2a6a56161db2f42695f7ec5769d67554dee2ebc73a2390b5bce5"
[[metadata.targets]]
requires_python = ">=3.10"
@@ -221,7 +221,7 @@ files = [
[[package]]
name = "arclet-letoderea"
-version = "0.21.3"
+version = "0.22.0"
requires_python = "<4.0,>=3.10"
summary = "A high-performance, simple-structured event system, relies on asyncio"
groups = ["default"]
@@ -230,8 +230,8 @@ dependencies = [
"typing-extensions>=4.12.0",
]
files = [
- {file = "arclet_letoderea-0.21.3-py3-none-any.whl", hash = "sha256:ea1e235a037a193c734fbbe754969db9242d551290b11a0c0babae4e17d56d9d"},
- {file = "arclet_letoderea-0.21.3.tar.gz", hash = "sha256:4a2588d521afcd18d356804dc4289209be07b885e1d7851c8e37c8c03b954af4"},
+ {file = "arclet_letoderea-0.22.0-py3-none-any.whl", hash = "sha256:7b0abe3e5d532cc4db8527f97bafeefb698fcb83d5094a41959eb3d0f5d7c847"},
+ {file = "arclet_letoderea-0.22.0.tar.gz", hash = "sha256:eab31c29a42aa61265944ec6aab31f94fb7465df77b53408875869cd70c4009c"},
]
[[package]]
diff --git a/pyproject.toml b/pyproject.toml
index 16fe674..b7e053b 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -6,7 +6,7 @@ authors = [
{name = "RF-Tar-Railt",email = "rf_tar_railt@qq.com"},
]
dependencies = [
- "arclet-letoderea<0.22.0,>=0.21.3",
+ "arclet-letoderea<0.23.0,>=0.22.0",
"arclet-alconna<2.0.0,>=1.8.44",
"satori-python-core<1.4.0,>=1.3.5",
"satori-python-client<1.4.0,>=1.3.2",
@@ -32,6 +32,25 @@ classifiers = [
"Framework :: AsyncIO",
"Operating System :: OS Independent",
]
+keywords = [
+ "asyncio",
+ "framework",
+ "instant-messaging",
+ "satori",
+ "entari",
+ "plugin-system",
+ "hmr",
+ "hot",
+ "module",
+ "replacement",
+ "reload",
+ "watch",
+ "development",
+ "hot-reload",
+ "hot-module-replacement",
+ "high-performance",
+ "event-driven",
+]
[project.urls]
homepage = "https://arclet.top"