feat(issues): infer project platform if not set (#95402)

If a project ends up with no `platform` (e.g. by API creation), we
attempt to infer it from its next event by setting the project's
`platform` to the event's `platform`.


Related: 
-
https://linear.app/getsentry/issue/ID-442/investigate-and-fix-projects-with-no-platform
- #82176

---------

Co-authored-by: Katie Byers <lobsterkatie@gmail.com>

Author: cvxluo

src/sentry/event_manager.py

@@ -14,13 +14,14 @@
 from django.core.cache import cache
 from django.core.exceptions import ValidationError
 from django.db import IntegrityError, OperationalError, connection, router, transaction
-from django.db.models import Max
+from django.db.models import Max, Q
 from django.db.models.signals import post_save
 from django.utils.encoding import force_str
 from urllib3.exceptions import MaxRetryError, TimeoutError
 from usageaccountant import UsageUnit
 
 from sentry import (
+    audit_log,
     eventstore,
     eventstream,
     eventtypes,
@@ -37,6 +38,7 @@
     LOG_LEVELS_MAP,
     MAX_TAG_VALUE_LENGTH,
     PLACEHOLDER_EVENT_TITLES,
+    VALID_PLATFORMS,
     DataCategory,
     InsightModules,
 )
@@ -127,6 +129,7 @@
 from sentry.types.group import GroupSubStatus, PriorityLevel
 from sentry.usage_accountant import record
 from sentry.utils import metrics
+from sentry.utils.audit import create_system_audit_entry
 from sentry.utils.cache import cache_key_for_event
 from sentry.utils.circuit_breaker import (
     ERROR_COUNT_CACHE_KEY,
@@ -137,6 +140,7 @@
 from sentry.utils.event import has_event_minified_stack_trace, has_stacktrace, is_handled
 from sentry.utils.eventuser import EventUser
 from sentry.utils.metrics import MutableTags
+from sentry.utils.options import sample_modulo
 from sentry.utils.outcomes import Outcome, track_outcome
 from sentry.utils.projectflags import set_project_flag_and_signal
 from sentry.utils.safe import get_path, safe_execute, setdefault_path, trim
@@ -464,6 +468,11 @@ def save(
         # After calling _pull_out_data we get some keys in the job like the platform
         _pull_out_data([job], projects)
 
+        # Sometimes projects get created without a platform (e.g. through the API), in which case we
+        # attempt to set it based on the first event
+        if sample_modulo("sentry:infer_project_platform", project.id):
+            _set_project_platform_if_needed(project, job["event"])
+
         event_type = self._data.get("type")
         if event_type == "transaction":
             job["data"]["project"] = project.id
@@ -690,6 +699,30 @@ def _pull_out_data(jobs: Sequence[Job], projects: ProjectsMapping) -> None:
         job["groups"] = []
 
 
+def _set_project_platform_if_needed(project: Project, event: Event) -> None:
+    if project.platform:
+        return
+
+    if event.platform not in VALID_PLATFORMS or event.get_tag("sample_event") == "yes":
+        return
+
+    try:
+        updated = Project.objects.filter(
+            Q(id=project.id) & (Q(platform__isnull=True) | Q(platform=""))
+        ).update(platform=event.platform)
+
+        if updated:
+            create_system_audit_entry(
+                organization=project.organization,
+                target_object=project.id,
+                event=audit_log.get_event_id("PROJECT_EDIT"),
+                data={**project.get_audit_log_data(), "platform": event.platform},
+            )
+
+    except Exception:
+        logger.exception("Failed to infer and set project platform")
+
+
 @sentry_sdk.tracing.trace
 def _get_or_create_release_many(jobs: Sequence[Job], projects: ProjectsMapping) -> None:
     for job in jobs:

src/sentry/options/defaults.py

@@ -3480,3 +3480,11 @@
     default=False,
     flags=FLAG_AUTOMATOR_MODIFIABLE,
 )
+
+# Rollout for inferring project platform from events received
+register(
+    "sentry:infer_project_platform",
+    type=Float,
+    default=0.0,
+    flags=FLAG_AUTOMATOR_MODIFIABLE,
+)

tests/sentry/event_manager/test_project_platform_infer.py

@@ -0,0 +1,23 @@
+from sentry.testutils.cases import TestCase
+from sentry.testutils.helpers import override_options
+from sentry.testutils.helpers.eventprocessing import save_new_event
+
+
+class ProjectPlatformInferTest(TestCase):
+    @override_options({"sentry:infer_project_platform": 1.0})
+    def test_platform_inferred_on_event(self):
+        project = self.create_project()
+
+        save_new_event({"message": "test", "platform": "javascript"}, project)
+
+        project.refresh_from_db()
+        assert project.platform == "javascript"
+
+    @override_options({"sentry:infer_project_platform": 1.0})
+    def test_platform_does_not_override_existing_platform(self):
+        project = self.create_project(platform="python")
+
+        save_new_event({"message": "test", "platform": "javascript"}, project)
+
+        project.refresh_from_db()
+        assert project.platform == "python"