[sentry-junior]

Creating a span metric monitor with an equation aggregate fails with a 400:

{"dataSources":{"aggregate":["Invalid Metric: equation|sum(a) / sum(b) is not a valid function"]}}

Spans Explore has supported equations since mid-2025, and Application Metrics monitors already accept them (#113105). For span monitors, _validate_aggregate sends the whole equation|… string to translate_aggregate_field, which can only handle a single function. So ratio monitors such as sum(gen_ai.usage.cache_read.input_tokens) / sum(gen_ai.usage.input_tokens) can't be created.

With the new organizations:spans-monitor-equations flag (FlagPole, api_expose=True), span equations get their own validation path:

• The equation is parsed against SPAN_DEFINITIONS. It must contain aggregate functions and no bare fields, because the subscription only sends aggregate expressions to Snuba.
• Every term has to be a function span monitors already accept (EAP_MONITOR_FUNCTIONS["spans"]). For example, count_sample() and count_if(...) are still rejected.
• The whole equation is resolved with the span SearchResolver, which catches bad argument types such as sum(span.op).
• The aggregate is stored unchanged in SnubaQuery.aggregate, the same equation|<expr> format Application Metrics uses.

Without the flag, span equations still fail, but with Span monitors do not support equations. instead of the "not a valid function" message.

Nothing downstream needed changing. PerformanceSpansEAPRpcEntitySubscription.build_rpc_request already turns a span equation into one formula expression, with an extrapolation mode on each aggregation. That is the shape Snuba's subscription validation requires. A new entity subscription test checks this, and the detector create test confirms the Snuba subscription gets created.

Out of scope:

• Monitor form editor for span equations. See feat(aci): Equation editor for span metric monitors behind a flag #130741
• Logs monitors, which still accept single functions only.
• Empty windows. get_aggregation_value_helper still turns an empty result into 0, so a "below X" ratio monitor fires on a window with no traffic. Ratio equations probably want to skip those updates; that's left as a follow-up.

via @mrduncan.

--

View Session [Sentry]

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 57ba0b5. Configure here.}

[saponifi3d]

lgtm! might be some cleanup for the tests if you wanted (paramterizing them might make the diff a little smaller).

[sentry-junior]

@saponifi3d Thanks for the suggestion. I folded the span equation validator cases into subTest tables in 4b4b5f9 (TestCase classes can't use pytest.mark.parametrize). The cases are the same, and the PR's test diff is 61 lines smaller (227 → 166).