feat(errors): Events API backend - Add error upsampling aggregation functions

[yuvmen]

Allow projects with error upsampling to use new sample_count(), sample_eps() and
sample_epm() function columns in Discover, returning the non extrapolated versions of these functions.

[github-actions]

🚨 Warning: This pull request contains Frontend and Backend changes!

It's discouraged to make changes to Sentry's Frontend and Backend in a single pull request. The Frontend and Backend are not atomically deployed. If the changes are interdependent of each other, they must be separated into two pull requests and be made forward or backwards compatible, such that the Backend or Frontend can be safely deployed independently.

Have questions? Please ask in the #discuss-dev-infra channel.

[roggenkemper]

can we separate the FE changes into a different PR from the BE ones?

[yuvmen]

yea Ill break it apart into FE and BE

[yuvmen]

@roggenkemper Separated it, the FE PR is now in: #95946

[roggenkemper]

could these loops be combined?

[codecov]

❌ 3 Tests Failed:

Tests completed	Failed	Passed	Skipped
26889	3	26886	204

View the top 3 failed test(s) by shortest run time

tests.snuba.api.endpoints.test_organization_events.OrganizationEventsErrorsDatasetEndpointTest::test_sample_eps_with_allowlisted_project

Stack Traces | 16.8s run time

#x1B[1m#x1B[.../api/endpoints/test_organization_events.py#x1B[0m:6913: in test_sample_eps_with_allowlisted_project
    assert response.status_code == 200, response.content
#x1B[1m#x1B[31mE   AssertionError: b'{"detail":"sample_eps is not a valid function"}'#x1B[0m
#x1B[1m#x1B[31mE   assert 400 == 200#x1B[0m
#x1B[1m#x1B[31mE    +  where 400 = <Response status_code=400, "application/json">.status_code#x1B[0m

tests.snuba.api.endpoints.test_organization_events.OrganizationEventsErrorsDatasetEndpointTest::test_sample_epm_with_allowlisted_project

Stack Traces | 17.1s run time

#x1B[1m#x1B[.../api/endpoints/test_organization_events.py#x1B[0m:6966: in test_sample_epm_with_allowlisted_project
    assert response.status_code == 200, response.content
#x1B[1m#x1B[31mE   AssertionError: b'{"detail":"sample_epm is not a valid function"}'#x1B[0m
#x1B[1m#x1B[31mE   assert 400 == 200#x1B[0m
#x1B[1m#x1B[31mE    +  where 400 = <Response status_code=400, "application/json">.status_code#x1B[0m

tests.snuba.api.endpoints.test_organization_events.OrganizationEventsErrorsDatasetEndpointTest::test_sample_count_with_allowlisted_project

Stack Traces | 17.6s run time

#x1B[1m#x1B[.../api/endpoints/test_organization_events.py#x1B[0m:6864: in test_sample_count_with_allowlisted_project
    assert response.status_code == 200, response.content
#x1B[1m#x1B[31mE   AssertionError: b'{"detail":"sample_count is not a valid function"}'#x1B[0m
#x1B[1m#x1B[31mE   assert 400 == 200#x1B[0m
#x1B[1m#x1B[31mE    +  where 400 = <Response status_code=400, "application/json">.status_code#x1B[0m

To view more test analytics, go to the Test Analytics Dashboard
_{📋 Got 3 mins? Take this short survey to help us improve Test Analytics.}

[cursor]

Bug: Upsampling Function Data Corruption

The convert_fields_for_upsampling function has two issues: it can lead to incorrect data in query results and fails to convert metadata correctly for empty results. Specifically, if both a standard function (e.g., count()) and its upsampled equivalent (e.g., upsampled_count()) are present, the standard function's key will incorrectly hold the upsampled value after conversion. Additionally, fields_meta keys are not converted when a query returns no data, as the conversion logic relies on keys being present in the data, resulting in incorrect field names in the metadata.

src/sentry/api/helpers/error_upsampling.py#L41-L62

sentry/src/sentry/api/helpers/error_upsampling.py

Lines 41 to 62 in 54fddec

	# Collect keys that need conversion and exist in data
	all_present_keys: set[str] = set()
	for result in data:
	all_present_keys.update(result.keys())
	# Filter the pre-ordered list to only include keys actually present
	keys_to_convert = [key for key in _ORDERED_CONVERSION_KEYS if key in all_present_keys]
	# Apply conversions to data
	for result in data:
	for original_key in keys_to_convert:
	if original_key in result:
	converted_key = _FUNCTION_KEY_CONVERSIONS[original_key]
	result[converted_key] = result[original_key]
	del result[original_key]
	# Apply conversions to fields_meta
	for original_key in keys_to_convert:
	if original_key in fields_meta:
	converted_key = _FUNCTION_KEY_CONVERSIONS[original_key]
	fields_meta[converted_key] = fields_meta[original_key]
	del fields_meta[original_key]

Fix in Cursor • Fix in Web

---

Was this report helpful? Give feedback by reacting with 👍 or 👎

[roggenkemper]

Should this use the flagpole definition now

[yuvmen]

hmm yea I can probably change the backend to use the flagpole instead of the existing option, but I am under some time pressure, that is a cleanup I can do and test afterwards, I don't need to do that yet really, I basically now have an option for the backend and a flappole for the frontend, and ill keep it like that for now and unite them later.