A language dependent cost model for the Plutus evaluator. Note that the EvaluationContext is entirely dependent on the cost model parameters (ie the Map Text Integer) and that this type uses the smart constructor mkCostModel to hide the evaluation context.

Convert cost model parameters to a cost model, making use of the conversion function mkEvaluationContext from the Plutus API.

Note that we always retain the original values that were supplied.

This function attempts to convert a Map with potential cost models into validated CostModels. If it is a valid cost model for a known version of Plutus, it is added to costModelsValid. If it is an invalid cost model for a known version of Plutus, the function will fail with a string version of CostModelApplyError. Lastly, if the Plutus version is unknown, the cost model is also stored in costModelsUnknown.

Encoding for the CostModel. Important to note that it differs from Encoding used by getLanguageView

Initial number of parameters in a CostModel for a specific language when the language was introduced. Starting with Conway we support variable number of parameters, therefore do not expect this number to reflect the reality on the number of supported parameters.

Prior to version 9, each CostModel was expected to be serialized as an array of integers of a specific length (depending on the version of Plutus). Starting in version 9, we allow the decoders to accept lists longer or shorter than what they require, so that new fields can be added in the future. For this reason, we must hard code the length expectation into the deserializers prior to version 9.

Note that the number of elements in the V1 and V2 cost models are allowed to change in the future, they are only fixed prior to version 9.

See https://github.com/intersectmbo/cardano-ledger/issues/2902 and https://github.com/intersectmbo/cardano-ledger/blob/master/docs/adr/2022-12-05_006-cost-model-serialization.md

For a known version of Plutus, attempting to construct a cost model with too few parameters (depending on the version) will result in an error. CostModelApplyError exists to collect these errors in the CostModels type. The CostModels type itself needs to be flexible enough to accept any map of Word8 to '[Int64]', so that cost models can be placed in the protocol parameters ahead of changes to the Plutus evaluation context. In this way, serializing a cost model, updating software, and deserializing can result in errors going away.

Additionally, CostModels needs to be able to store cost models for future version of Plutus, which we cannot yet even validate. These are stored in costModelsUnknown.

Construct an all valid CostModels

Updates the first CostModels with the second one, so that only the cost models that are present in the second one get updated while all the others stay unchanged. Language specific errors and unknown cost models are removed, whenever a valid CostModel for the language is supplied in the update.

Turn a CostModels into a mapping of potential language versions and cost model values, with no distinction between valid and unknown cost models. This is used for serialization, so that judgements about known languages can be made upon deserialization.