Merge pull request #2527 from AppDaemon/start-immediate

Start-immediate

Author: acockburn

appdaemon/adapi.py

@@ -3225,7 +3225,6 @@ async def run_every(
         start: str | dt.time | dt.datetime | None = None,
         interval: TimeDeltaLike = 0,
         *args,
-        immediate: bool = False,
         random_start: TimeDeltaLike | None = None,
         random_end: TimeDeltaLike | None = None,
         pin: bool | None = None,
@@ -3242,7 +3241,9 @@ async def run_every(
                 intervals will be calculated forward from the start time, and the first trigger will be the first
                 interval in the future.
 
-                - If this is a ``str`` it will be parsed with :meth:`~appdaemon.adapi.ADAPI.parse_time()`.
+                - If this is ``now`` (default), then the first trigger will be now + interval
+                - If this is ``immediate``, then the first trigger will happen immediately
+                - Other ``str`` types will be parsed with :meth:`~appdaemon.adapi.ADAPI.parse_time()`.
                 - If this is a ``datetime.time`` object, the current date will be assumed.
                 - If this is a ``datetime.datetime`` object, it will be used as is.
 
@@ -3258,8 +3259,6 @@ async def run_every(
                 - If this is a ``timedelta`` object, the current date will be assumed.
 
             *args: Arbitrary positional arguments to be provided to the callback function when it is triggered.
-            immediate (bool, optional): Whether to immediately fire the callback or wait until the first interval if the
-                start time is now.
             random_start (int, optional): Start of range of the random time.
             random_end (int, optional): End of range of the random time.
             pin (bool, optional): Optional setting to override the default thread pinning behavior. By default, this is
@@ -3313,7 +3312,7 @@ def timed_callback(self, **kwargs): ...  # example callback
 
         """
         interval = utils.parse_timedelta(interval)
-        next_period = await self.AD.sched.get_next_period(interval, start, immediate=immediate)
+        next_period = await self.AD.sched.get_next_period(interval, start)
 
         self.logger.debug(
             "Registering %s for run_every in %s intervals, starting %s",

appdaemon/scheduler.py

@@ -483,22 +483,31 @@ async def get_next_period(
         self,
         interval: TimeDeltaLike,
         start: time | datetime | str | None = None,
-        *,
-        immediate: bool = False,
     ) -> datetime:
+        """Calculate the next execution time for a periodic interval.
+
+        If start is "immediate", returns the current time.
+        Otherwise, calculates a start time (defaulting to "now") and advances by the
+        interval until a future time is reached.
+        """
         interval = utils.parse_timedelta(interval)
         start = "now" if start is None else start
 
         # Get "now" once and use it consistently to avoid timing races
         now = await self.get_now()
-        aware_start = await self.parse_datetime(start, aware=True, now=now)
-        assert isinstance(aware_start, datetime) and aware_start.tzinfo is not None
-
-        # Skip forward to the next period if start is in the past
-        while aware_start < now or (immediate and aware_start <= now):
-            aware_start += interval
-
-        return aware_start
+        match start:
+            case "immediate":
+                return now
+            case "now", _:
+                aware_next = await self.parse_datetime(start, aware=True, now=now)
+                # Skip forward to the next period if start is in the past
+                # This makes the result in the first
+                while aware_next <= now:
+                    aware_next += interval
+
+        assert isinstance(aware_next, datetime) and aware_next.tzinfo is not None, \
+            "aware_start must be a timezone aware datetime"
+        return aware_next
 
     async def terminate_app(self, name: str):
         if app_sched := self.schedule.pop(name, False):

docs/HISTORY.md

@@ -9,7 +9,7 @@
 - Reload modified apps on SIGUSR2 - contributed by [chatziko](https://github.com/chatziko)
 - Using urlib to create endpoints from URLs - contributed by [cebtenzzre](https://github.com/cebtenzzre)
 - Added {py:meth}`~appdaemon.plugins.hass.hassapi.Hass.process_conversation` and {py:meth}`~appdaemon.plugins.hass.hassapi.Hass.reload_conversation` to the {ref}`Hass API <hass-api-usage>`.
-- Added `immediate` kwargs to `run_every` to control semantics around `start == "now"`
+- Added special value `immediate` to {py:meth}`~appdaemon.adapi.ADAPI.run_every` semantics for the `start` kwarg. See the method docs for more information.
 
 **Fixes**
 

tests/conf/apps/scheduler_test_app.py

@@ -7,9 +7,22 @@ class SchedulerTestAppMode(str, Enum):
     """Enum for different modes of the SchedulerTestApp."""
 
     RUN_EVERY = "run_every"
+    RUN_IN = "run_in"
 
 
 class SchedulerTestApp(ADAPI):
+    """
+    A test app to verify scheduler functionality.
+
+    Configuration Args:
+        mode (str, optional): The mode of operation. Defaults to 'run_every'.
+        register_delay (float, optional): Delay before setup in seconds. Defaults to 0.5.
+
+    RUN_EVERY:
+        interval (int): Interval in seconds for run_every. Required.
+        msg (str): Message to pass to callback. Required.
+        start (str, optional): Start time description. Defaults to "now".
+    """
     def initialize(self):
         self.set_log_level("DEBUG")
         self.log("SchedulerTestApp initialized")
@@ -20,10 +33,14 @@ def setup_callback(self, **kwargs) -> None:
         self.log(f"Running in {self.mode} mode")
         match self.mode:
             case SchedulerTestAppMode.RUN_EVERY:
-                start = self.args.get("start", "now")
-                interval = self.args["interval"]
-                msg = self.args["msg"]
-                self.run_every(self.run_every_callback, start=start, interval=interval, msg=msg)
+                match self.args:
+                    case {"interval": interval, "msg": str(msg)}:
+                        start = self.args.get("start", "now")
+                        self.run_every(self.run_every_callback, start=start, interval=interval, msg=msg)
+                        return
+            case SchedulerTestAppMode.RUN_IN:
+                pass
+        raise ValueError(f"Invalid arguments for {self.mode}")
 
     @property
     def mode(self) -> SchedulerTestAppMode:

tests/conftest.py

@@ -28,7 +28,7 @@ async def ad_obj(running_loop: asyncio.BaseEventLoop, ad_cfg: AppDaemonConfig, l
     for cfg in ad.logging.config.values():
         logger_ = logging.getLogger(cfg["name"])
         logger_.propagate = True
-        logger_.setLevel("DEBUG")
+    #     logger_.setLevel("DEBUG")
 
     await ad.app_management._process_import_paths()
     ad.app_management.dependency_manager = DependencyManager(python_files=list(), config_files=list())

tests/functional/test_run_every.py

@@ -1,11 +1,13 @@
 import logging
 import re
 import uuid
-from datetime import timedelta
+from datetime import datetime, timedelta
 from functools import partial
 from itertools import product
+from typing import cast
 
 import pytest
+import pytz
 from appdaemon.types import TimeDeltaLike
 from appdaemon.utils import parse_timedelta
 
@@ -29,9 +31,12 @@ async def test_run_every(
     n: int = 2,
 ) -> None:
     interval = parse_timedelta(interval)
-    run_time = (interval * n) + timedelta(seconds=0.01)
-    register_delay = 0.1
-    run_time += timedelta(seconds=register_delay)  # Accounts for the delay in registering the callback
+
+    # Calculate base runtime for 'n' occurrences plus a small buffer to account for the delay in registering the callback
+    register_delay = timedelta(seconds=0.2)
+    run_time = (interval * (n + 1)) + register_delay
+
+    # If start time is future "now + offset", add offset to ensure coverage
     if (parts := re.split(r"\s+[\+]\s+", start)) and len(parts) == 2:
         _, offset = parts
         run_time += parse_timedelta(offset)
@@ -41,11 +46,62 @@ async def test_run_every(
     app_args = dict(start=start, interval=interval, msg=test_id, register_delay=register_delay)
     async with run_app_for_time(app_name, run_time=run_time.total_seconds(), **app_args) as (ad, caplog):
         check_interval_partial = partial(check_interval, caplog, f"kwargs: {{'msg': '{test_id}',")
+        check_interval_partial(n, interval)
 
-        if start.startswith("now -"):
-            check_interval_partial(n, interval)
-        else:
-            check_interval_partial(n + 1, interval)
+        cb_count = await ad.state.get_state('test', 'admin', f'app.{app_name}', 'instancecallbacks')
+        assert cast(int, cb_count) >= n, "Callback didn't get called enough times."
 
         # diffs = utils.time_diffs(utils.filter_caplog(caplog, test_id))
         # logger.debug(diffs)
+
+
+@pytest.mark.asyncio(loop_scope="session")
+@pytest.mark.parametrize("start", ["now", "immediate"])
+async def test_run_every_start_time(
+    run_app_for_time: AsyncTempTest,
+    start: str,
+) -> None:
+    interval = timedelta(seconds=0.5)
+    run_time = timedelta(seconds=1)
+    register_delay = timedelta(seconds=0.1)
+
+    match start:
+        case "now":
+            n = 1
+        case "immediate":
+            n = 2
+
+    app_name = "scheduler_test_app"
+    test_id = str(uuid.uuid4())
+    app_args = dict(start=start, interval=interval, msg=test_id, register_delay=register_delay)
+    async with run_app_for_time(app_name, run_time=run_time.total_seconds(), **app_args) as (ad, caplog):
+        check_interval(
+            caplog,
+            f"kwargs: {{'msg': '{test_id}',",
+            n=n,
+            interval=interval
+        )
+
+        cb_count = await ad.state.get_state('test', 'admin', f'app.{app_name}', 'instancecallbacks')
+        assert cast(int, cb_count) >= (n + 1), "Callback didn't get called enough times."
+
+now = datetime.now(pytz.utc)
+START_TIMES = ["now", now, now.time(), now.isoformat()]
+
+@pytest.mark.asyncio(loop_scope="session")
+@pytest.mark.parametrize("start", START_TIMES)
+async def test_run_every_start_time_types(
+    run_app_for_time: AsyncTempTest,
+    start: str,
+) -> None:
+    interval = timedelta(seconds=0.25)
+    run_time = timedelta(seconds=1)
+    register_delay = timedelta(seconds=0.1)
+    n = 3
+
+    app_name = "scheduler_test_app"
+    test_id = str(uuid.uuid4())
+    app_args = dict(start=start, interval=interval, msg=test_id, register_delay=register_delay)
+    async with run_app_for_time(app_name, run_time=run_time.total_seconds(), **app_args) as (ad, caplog):
+        cb_count = await ad.state.get_state('test', 'admin', f'app.{app_name}', 'instancecallbacks')
+        assert cast(int, cb_count) >= (n + 1), "Callback didn't get called enough times."