Inject result_type on publish to SNS, not on read

[dwhswenson]

Previously, notifiers injected result_type into the payloads for the template, so that it would always be available to templates.

However, I found an issue that, if two result types had identical payloads, the SNS FIFO deduplication would only publish one of the result types. Specifically, in eshgham-cron, if I was subscribed to ACTION_NEEDED (INACTIVATED + FAILED workflows), and I only had INACTIVATED workflows, I wasn't triggering my notification.

This fixes that by putting result_type in when we publish to SNS, instead of when we receive the message from SQS.

Yay for Codex catching the docs fixes. I would have forgotten to do that.

Assisted-by: Codex:gpt-5.4

[Copilot]

Pull request overview

This PR fixes SNS FIFO deduplication collisions by ensuring each published SNS message body includes a top-level result_type, rather than injecting result_type only at notification-read time.

Changes:

• Add build_result_message_payload() and use it in dispatch_sns_messages() so the SNS message body includes result_type (in addition to message attributes).
• Update notifier parsing to require result_type in the payload and validate it against the SQS messageAttributes.result_type.
• Update render CLI behavior, tests, and documentation to match the new data flow.

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
tests/test_render.py	Updates render CLI expectations for mismatched vs matching result_type in payloads.
tests/test_lambda_task.py	Adjusts SNS publish assertions and adds unit tests for distinct bodies and non-mutation behavior.
tests/notifications/test_print_handler.py	Updates fixture payloads to include result_type.
tests/notifications/test_email_handler.py	Updates email handler tests to include result_type in SQS event payloads.
tests/notifications/test_base.py	Reworks notifier parsing tests around required payload result_type and mismatch handling.
src/lambdacron/render.py	Uses publisher-style payload building to emulate SNS message bodies for template rendering.
src/lambdacron/notifications/print_handler.py	Removes include_result_type plumbing consistent with new parsing rules.
src/lambdacron/notifications/base.py	Enforces payload result_type presence and validates against message attributes.
src/lambdacron/lambda_task.py	Introduces build_result_message_payload() and injects result_type into published SNS message bodies.
docs/how-to/write-lambda-and-templates.md	Updates docs to reflect injection at publish-time instead of notifier-time.
docs/how-to/set-up-ses.md	Minor formatting adjustments.
docs/explanation/data-flow.md	Updates data flow examples and parsing description to match new behavior.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Requiring every incoming payload to include a non-empty string result_type is a behavior change that will cause any in-flight/older SQS messages (published before this change) to fail parsing and be retried indefinitely as batchItemFailures. Consider a backward-compatible transition: if payload.result_type is missing/empty but messageAttributes.result_type is present, inject it (and optionally log a warning / plan a deprecation), while still rejecting mismatches when both are present.

Suggested change

	if not isinstance(payload_result_type, str) or not payload_result_type:
	raise ValueError(
	"Result payload must include a non-empty string result_type"
	)
	attribute_result_type = self._extract_result_type(record)
	if attribute_result_type and attribute_result_type != payload_result_type:
	raise ValueError(
	"Result type mismatch between payload and message attributes "
	f"(payload {payload_result_type}, attribute {attribute_result_type})"
	)
	attribute_result_type = self._extract_result_type(record)
	if not isinstance(payload_result_type, str) or not payload_result_type:
	if attribute_result_type:
	logging.getLogger(__name__).warning(
	"Result payload missing result_type; using "
	"messageAttributes.result_type for backward compatibility"
	)
	payload["result_type"] = attribute_result_type
	payload_result_type = attribute_result_type
	else:
	raise ValueError(
	"Result payload must include a non-empty string result_type"
	)
	elif attribute_result_type and attribute_result_type != payload_result_type:
	raise ValueError(
	"Result type mismatch between payload and message attributes "
	f"(payload {payload_result_type}, attribute {attribute_result_type})"
	)

[dwhswenson]

We're not worried about backward compatibility here.

[Copilot]

selected = payload.get(result_type) means a missing result_type key produces the generic "must be a JSON object" error (because None fails the Mapping check). Consider explicitly detecting a missing key and raising a clearer error (e.g., that the task output does not contain the requested result type) to make the CLI failure mode easier to understand.

[dwhswenson]

Addressed in c699223

[ethanholz]

This makes sense to me, merging.