an assumption holds or refuses the data, and Calliope's `errors: warn` checks have no place in a spec

[FBumann]

Prompt: "Let’s add one parent issue for caliope and file focused issues for individual topics or link existing ones."

Note

The following content was generated by AI.

Calliope has three checks at errors: warn. A mathspec assumption holds, or the data is refused, so the port in #770 leaves these checks out. The recommendation: keep warnings out of the language, and record that in limits.md.

Case, why refuse, what to change

Case. Calliope v0.7.0 base.yaml has the checks flow_in_for_supply_tech, flow_out_for_demand_tech and no_storage_discharge_depth_when_clustering, each at errors: warn. The model still builds when one of them fails.

Why refuse. limits.md makes a check language because "two consumers must not answer differently". A warning does not change which data a consumer accepts, so two consumers that treat it differently build the same model. A warning is lint for data prep or for the consumer. #11 (checks as assumptions:, landed in #589) closed without a discussion of severity, so nothing records this decision.

The alternative, severity: warn on an assumption, is a small field. But each consumer must then honour it, and the typesetter must decide whether to print it. No second caller has asked for it.

What to change.

• Add a row to limits.md → Deliberate non-primitives: "An assumption that only warns. Why refused: it does not change which data is accepted. Instead: a check in data preparation or in the consumer."
• In docs(calliope): all of calliope's math is a set of fragments that merge, with its modes laid over them as patches #770, move gap 9 from "what mathspec is missing" to "out by design".