[lisicky]

What

Caches provider answers that cannot change, cuts the calls MeshTxBuilder.complete() makes, and fixes two builds that could not complete at all.

Contents

mesh-common:

• CachedFetcher: an IFetcher wrapper that caches only answers that cannot change.
  • fetchUTxOs(hash): a tx's outputs are fixed by its body, so they are fetched once per hash (whole tx, whatever index is asked for). LRU, 1000 txs by default.
  • fetchProtocolParameters / fetchCostModels: the latest values are kept until the current epoch ends. Values for an explicit past epoch are kept for good. Within 5 minutes after a boundary, values are kept only until those 5 minutes are up, while providers catch up.
  • Every other method (fetchAddressUTxOs, fetchAccountInfo, fetchTxInfo, fetchBlockInfo, asset and governance queries) reflects current chain state. Answers are kept for 10 s by default, half of Cardano's 20 s average block time (1 s slots, active slot coefficient 0.05), so a burst of identical queries costs one request. Configurable through volatileTtlMs (0 disables it). clearVolatile() drops them, and MeshTxBuilder.submitTx calls it after submitting.
  • Concurrent calls share one request. Failures, and empty tx outputs, are not cached. get(url) passes straight through.
  • CachedFetcher.wrap(fetcher) returns one shared cache per fetcher instance.
• resolveEpochEnd(network, ms): the time the current epoch ends, computed from the genesis epoch and slot lengths in SLOT_CONFIG_NETWORK. These change only by hard fork. They were checked against current chain tips: mainnet epoch 661, preprod 318 and preview 1448 all match epoch number and epoch slot.
  • With no network given, CachedFetcher uses the earliest end among mainnet, preprod and preview. "testnet" (custom network) disables epoch caching.

mesh-transaction (MeshTxBuilder):

• Fetches go through CachedFetcher.wrap(fetcher). Builders and evaluators sharing a provider share answers: cost models are fetched once per epoch instead of once per complete(), and reference-script / input txs once per process instead of once per builder.
  • A mainnet change address pins the epoch boundary to mainnet's.
• evaluateRedeemers skips evaluateTx and the serialization before it when nothing is witnessed by a Plutus script. Before, it ran once per coin-selection cost estimate on every transaction, including plain ADA sends.
• Input and reference-script info is looked up first in what the builder already holds: selectUtxosFrom UTxOs, inputForEvaluation UTxOs and chainTx outputs. fetchUTxOs is called only for hashes still unknown.
• A failed or empty response is no longer remembered, so a later complete() retries it instead of failing with "Couldn't find value information".
• fetchCostModels runs alongside the UTxO fetches instead of before them.
• Native (simple) script reference sources are now fetched and completed, including their scriptSize. Before, simpleScriptTxInReference without a hash and size crashed with TypeError: Cannot read properties of undefined (reading 'find'). Native-script inputs given by reference only are now completed too.
• A reference script's hash is derived from scriptRef when the fetcher does not return scriptHash (e.g. ScalusEmulator).
• clone() shares the fetched UTxO cache instead of deep-copying it for every coin-selection callback.
• TxParser fetches missing txs in parallel, through the same cache.

mesh-core-cst (OfflineEvaluatorScalus) and mesh-core-csl (OfflineEvaluator):

• Inputs are resolved through CachedFetcher.wrap(fetcher), in parallel. Each coin-selection round used to refetch every input, collateral and reference input. If the builder already fetched a tx through the same provider, the evaluator reuses it.
• OfflineEvaluator no longer appends resolved UTxOs to the caller's additionalUtxos array.

Provider calls, before → after

Measured with ScalusEmulator behind a call-recording proxy (provider-calls.test.ts):

Scenario	Before	After
ADA send, evaluator set	evaluateTx ×1 (×1+k with extra selection rounds)	0
Two script spends, new builder each, same ref-script UTxO	fetchCostModels ×2, fetchUTxOs ×2 per tx hash	×1, ×1
Spend of a chainTx output given as hash/index	fetchUTxOs for an unsubmitted hash, then "Couldn't find value information"	0 fetches, builds and submits
Native script by reference without hash/size	crash	1 fetch per hash, builds and submits
Two script-spend builds with one OfflineEvaluatorScalus	fetchUTxOs ×4	×2
OfflineEvaluator, same tx evaluated twice	fetchUTxOs ×10	×5

No breaking API changes. The protected getUTxOInfo / queryAllTxInfo keep their signatures.

Validation

npm run build:mesh
cd packages/mesh-common && npx jest
cd packages/mesh-transaction && NODE_OPTIONS=--experimental-vm-modules npx jest
cd packages/mesh-core-cst && npm test
cd packages/mesh-scalus-emulator && npm test
cd packages/mesh-core-csl && npx jest

• mesh-common: 200 tests pass, including cached-fetcher.test.ts and the resolveEpochEnd cases.
• mesh-transaction: 160 tests pass. script-tx-fetcher and script-tx-fail fail to load under --experimental-vm-modules on main too, because they import dist through require. Both pass without the flag.
• mesh-core-cst (179 tests), mesh-scalus-emulator (24) and mesh-wallet (32) pass.
• mesh-core-csl: the new fetch assertions pass. The same 4 tests in offline-providers/evaluator.test.ts fail on main: expected steps differ after the cost-model update (e.g. 164980381 vs 164973765).