Version 9.0.0

cooked-validators v9.0.0 comes with 3 major upgrades:

• A better balancing mechanism, with more options, bug fixes, and optimizations.
• A full rewrite using Polysemy. This stable effect system makes cooked-validators more expressive, more consistent and with better separation of concerns.
• An expansion of the LTL DSL to modify traces, with the addition of negation and derived idioms such as implication.

Full CHANGELOG below.

[9.0.0] - 2026-05-28

Added

• New LtlNot primtive, to negate Ltl expressions. This allows to ensure a
  specific tweak will result in mzero and marks a new milestone in the
  expressiveness of the Tweak/Attack DSL.
• New Ltl combinators resulting from the addition of LtlNot, such as
  nowhere, whenAble, ...
• ExtendedStagedMockChain: It is not possible to extend a mockchain run with
  arbitrary effects, while the associated tweaks will also have access to the
  added effects.
• A new capability in a mockchain run, which allows to take note (basically log)
  anything. Functions note(|p|s|w|l) support this functionality.
• A new capability in a mockchain run, which allows to make assertions which
  will be displayed in the final result, and taken into account during
  tests. Functions assert(|') support his functionality.
• A new testBoolMsg function which outputs an error message when given
  False.
• Functions testCookedFromInitDistTemplate and
  testCookedQCFromInitDistTemplate to build tests from the initial
  distribution template (the old default initial distribution).
• The ability to enable/disable everything from the printing of the final result
  of running a mockchain. The pretty printer per se has also been improved.
• New HList type for heterogeneous lists, used to return type-retaining
  results for Utxo searches.
• New TxSkel option txSkelOptDeferFailures to defer validation failures
  occurring during the balancing process. This has the upside of visualizing the
  last modified transaction sent for validation in the log, with the downside of
  a (huge) loss in performance. To be used for debugging purpose only.
• New TxSkel option txSkelOptMaxNbOfBalancingUtxos to limit the total
  maximum number of UTxOs that can be picked as extra inputs during
  balancing. This will only matter if the number of UTxOs owned by the balancing
  wallet is huge, to limit exponential growth, which rarely happens.

Removed

• Tweaks can no longer issue write action such as waiting a certain amount of
  time. As a consequence, waitUntilValidTweak has been removed.

Changed

• The datum hijacking attack has been updated: all the sub-tweaks directly set
  the new outputs, and they all return only the modified outputs.
• The module Ltl.Combinators has been integrated to Ltl. The combinators have
  been enriched and renamed to better match the rest of the Ltl API.
• cooked-validators has been fully migrated from mtl to polysemy. Type class
  MonadBlockChain and its variants have been replaced by DirectMockChain,
  StagedMockChain, ExtendedStagedMockChain and FullMockChain.
• Some datum hijacking params have been changed. ownedByDatumHijackingParams
  now takes a specific user as parameter, while its old behavior is now given by
  typedByDatumHijackingParams. Also, txSkelOutPredDatumHijackingParams has
  been renamed outPredDatumHijackingParams.
• The whole file tree has been updated, in particular under
  Cooked.MockChain. Each effect is given its own file, such as
  Cooked.MockChain.Read, Cooked.MockChain.Error ...
• Generation functions which use to live under
  Cooked.MockChain.GenerateTx.Witness now live in their own
  Cooked.MockChain.GenerateTx.Credential file.
• All the auto adjustment made by cooked are now implemented using tweaks, such
  as autofilling of min ada, auto assignement of reference script...
• Initial distributions have been downgraded from a first class citizen to a
  mere helping structure equivalent to using forceOutputs. By default, the
  mockchain runs now have an empty state, and non-empty initial distribution can
  be fed to runs in one of the following 3 ways: 1. use forceOutputs at the
  beginning of a run. 2. Use runMockChainFromConf or
  runMockChainFromInitiDist while running a trace. 3. use withInitDist when
  running tests. The old default initial distribution still exists as
  initialDistributionTemplate and is simply directly a list of outputs.
• The UTxO searches have been fully reworked. They now happen in 3
  steps: 1. bootstrap the search with a set of UTxOs (beginSearch and
  beginSearchP) 2. filter (ensure, ensurePure, ensureAFoldIs and
  ensureAFoldIsn't) and/or extract elements from the selected outputs
  (extract, extractPure,extractAFold, extractTotal, extractPureTotal
  and extractGetter) in an type-retaining heterogeneous list. 2. retrieve the
  result of the search (getOutputs, getOutputsAndExtracts, getExtracts,
  getTxOutRefs and getTxOutRefandOutputs). Some additional helpers are
  provided for basic searches (utxosAtSearch, allUtxosSearch,
  txSkelOutByRefSearch and txSkelOutByRefSearch') and for basic filters
  (ensureOnlyValueOutputs, ensureVanillaOutputs and
  ensureProperReferenceScript).
• txSkelLabel has been renamed txSkelLabels

Fixed

• The condition for which a balancing solution is considered optimal: The main
  balancing function reachValue which picks the inputs to add and creates an
  additionnal output when needed has been updated. It used to pick the optimal
  result based on how small the additional UTxO was, and it now properly
  minimizes the total overhead in terms of size (and thus cost) that balancing
  adds to the transaction. This process is more involved but also more accurate.
• An optimization bug in the optimal fee computation.