Remove redundant cron check-in pairs

See appsignal/appsignal-ruby#1407 for context.
This commit reimplements that in Python.

Before submitting a batch of cron check-in events, check for any pairs
of events that are redundant with other pairs of cron check-in events.
This would happen, for example, when running a `cron` block for a given
identifier many times in quick succession.

A redundant pair of cron check-in events is a pair of start and
finish cron check-in events for a given identifier, given that another
pair of start and finish check-in events exists for that identifier.

Author: unflxw

.changesets/remove-redundant-cron-check-in-pairs.md

@@ -0,0 +1,6 @@
+---
+bump: patch
+type: change
+---
+
+Remove redundant cron check-in pairs. When more than one pair of start and finish cron check-in events is reported for the same identifier in the same period, only one of them will be reported to AppSignal.

src/appsignal/check_in/event.py

@@ -70,3 +70,49 @@ def describe(events: list[Event]) -> str:
 
     # This shouldn't happen.
     return "unknown check-in event"  # type: ignore[unreachable]
+
+
+def deduplicate_cron(events: list[Event]) -> None:
+    """Remove redundant pairs of cron events in-place, keeping only
+    the most recent complete pair."""
+    start_digests: dict[str, set[str | None]] = {}
+    finish_digests: dict[str, set[str | None]] = {}
+    complete_digests: dict[str, set[str | None]] = {}
+    keep_digest: dict[str, str | None] = {}
+
+    # Find complete pairs (start+finish) and track the latest one
+    for event in events:
+        if event["check_in_type"] != "cron" or event["kind"] not in ["start", "finish"]:
+            continue
+
+        identifier = event["identifier"]
+        digest = event.get("digest")
+        if identifier not in start_digests:
+            start_digests[identifier] = set()
+            finish_digests[identifier] = set()
+            complete_digests[identifier] = set()
+
+        if event["kind"] == "start":
+            start_digests[identifier].add(digest)
+            if digest in finish_digests[identifier]:
+                complete_digests[identifier].add(digest)
+                keep_digest[identifier] = digest
+        elif event["kind"] == "finish":
+            finish_digests[identifier].add(digest)
+            if digest in start_digests[identifier]:
+                complete_digests[identifier].add(digest)
+                keep_digest[identifier] = digest
+
+    # Iterate through the events and remove unwanted ones in place
+    i = 0
+    while i < len(events):
+        event = events[i]
+        if (
+            event["check_in_type"] == "cron"
+            and event.get("kind") in ("start", "finish")
+            and event.get("digest") in complete_digests.get(event["identifier"], set())
+            and event.get("digest") != keep_digest.get(event["identifier"])
+        ):
+            del events[i]
+        else:
+            i += 1

src/appsignal/check_in/scheduler.py

@@ -9,7 +9,7 @@
 from ..client import Client
 from ..config import Config
 from ..transmitter import transmit
-from .event import Event, describe, is_redundant
+from .event import Event, deduplicate_cron, describe, is_redundant
 
 
 class Scheduler:
@@ -133,7 +133,9 @@ def _stop_waker(self) -> None:
     def _push_events(self) -> None:
         if not self.events:
             return
-        self.queue.put(self.events.copy())
+        events_copy = self.events.copy()
+        deduplicate_cron(events_copy)
+        self.queue.put(events_copy)
         self.events.clear()
         self._start_waker(self.BETWEEN_TRANSMISSIONS_DEBOUNCE_SECONDS)
 

tests/check_in/test_event.py

@@ -1,7 +1,15 @@
 import json
+from itertools import permutations
 from typing import cast
 
-from appsignal.check_in.event import Event, cron, describe, heartbeat, is_redundant
+from appsignal.check_in.event import (
+    Event,
+    cron,
+    deduplicate_cron,
+    describe,
+    heartbeat,
+    is_redundant,
+)
 
 
 def test_describe_no_events():
@@ -93,3 +101,51 @@ def test_heartbeat_event_json_serialised():
     assert '"timestamp":' in serialised
     assert '"kind":' not in serialised
     assert '"digest":' not in serialised
+
+
+def test_deduplicate_cron_removes_redundant_pairs():
+    first_start = cron("checkin-name", "first", "start")
+    first_finish = cron("checkin-name", "first", "finish")
+    second_start = cron("checkin-name", "second", "start")
+    second_finish = cron("checkin-name", "second", "finish")
+    base_events = [first_start, first_finish, second_start, second_finish]
+
+    for events_tuple in permutations(base_events):
+        events = list(events_tuple)  # Convert tuple to list for in-place modification
+        deduplicate_cron(events)
+
+        assert len(events) == 2
+        assert events[0]["digest"] == events[1]["digest"]
+        assert {"start", "finish"} == {events[0]["kind"], events[1]["kind"]}
+
+
+def test_deduplicate_cron_keeps_unmatched_pairs():
+    first_start = cron("checkin-name", "first", "start")
+    second_start = cron("checkin-name", "second", "start")
+    second_finish = cron("checkin-name", "second", "finish")
+    third_finish = cron("checkin-name", "third", "finish")
+    base_events = [first_start, second_start, second_finish, third_finish]
+
+    for events_tuple in permutations(base_events):
+        events = list(events_tuple)  # Convert tuple to list for in-place modification
+        deduplicate_cron(events)
+
+        assert len(events) == 4
+        for event in [first_start, second_start, second_finish, third_finish]:
+            assert event in events
+
+
+def test_deduplicate_cron_keeps_different_identifiers():
+    first_start = cron("checkin-name", "first", "start")
+    first_finish = cron("checkin-name", "first", "finish")
+    second_start = cron("other-name", "second", "start")
+    second_finish = cron("other-name", "second", "finish")
+    base_events = [first_start, first_finish, second_start, second_finish]
+
+    for events_tuple in permutations(base_events):
+        events = list(events_tuple)  # Convert tuple to list for in-place modification
+        deduplicate_cron(events)
+
+        assert len(events) == 4
+        for event in [first_start, first_finish, second_start, second_finish]:
+            assert event in events