CHANGELOG.md

v2.2.1 - 25 December 2025

• Added wild, wild_low, and wild_high options to largest_count() evaluation. These allow specifying outcomes as wild, combining with any outcomes, outcomes that they are lower than, and outcomes that they are higher than respectively.
• standard_pool renamed to d_pool, leaving a deprecation warning.

v2.2.0 - 11 December 2025

Major modifications to map(repeat) Markov process flow.

• Now requires Python >= 3.12 (from 3.10).
• Reroll split into separate Reroll and Restart symbols with different meanings in map(repeat) and Again.
  • In map(repeat), Reroll rerolls only the current stage, while Restart restarts the entire process.
  • With again_count, Reroll rerolls any dice over the limit until they don't roll Again, while Restart restarts the entire process.
  • With again_depth, Reroll rerolls the last depth if going over the limit, while Restart cannot be used with again_depth.
• Break outcome wrapper that can be returned from a function sent to map(repeat) in order to immediately terminate the Markov process.
• map(repeat) now defaults to repeat=None; repeat is now explicitly incompatible with Again.
• Remove separate time_limit parameter to map(); repeat now always returns early if fully absorbed.
• again_end can now be used with again_count.
• .weightless() method for multiset generators that sets the quantity for all possible multisets to 1.
• New product_of_counts() multiset evaluation.
• More specific overloads for map, tupleize, vectorize.
• Rename commonize_denominator() to harmonize_denominator() and add optional weight parameter.
• Fix immediate absorption case for map(repeat='inf').

v2.1.3 - 3 December 2025

• Change max_pair_lowest() and max_pair_highest() to max_pair_keep() and max_pair_drop().
• Add remaining operators to the max_pair methods.
• Rename reroll_to_pool(mode) options from 'highest' to 'high'.
• Allow max_time argument to be omitted from time_to_sum if all outcomes are positive.
• Allow lexicographic comparison between Vectors of different length.
• Prevent unevaluated expressions from being returned from a @multiset_function.

v2.1.2 - 7 August 2025

• Breaking change: difference (but not other operations) now has keep_negative_counts parameter again, which defaults to False.

v2.1.1 - 4 August 2025

• Experimental sort_pair_keep_while() and sort_pair_drop_while() multiset ops.

v2.1.0 - 30 July 2025

• Breaking change: any() multiset evaluation renamed and inverted to empty().
• Breaking change: sort_match() renamed to sort_pair(), maximum_match_highest() renamed to max_pair_highest() and similar for minimum_match_highest().
• Add extra parameter to sort_pair.
• Rename several which and target parameters to outcomes.
• Make max_rerolls and depth parameters of reroll_to_pool() keyword-only, and add default outcomes to reroll.
• Experimental leximin() and leximax() multiset evaluations.
• Experimental versus_any() and versus_all() multiset operations.

v2.0.4 - 9 July 2025

• Fix internal star in map_and_time().
• Experimental mean_time_to_absorb().

v2.0.3 - 3 July 2025

• Add star option for multiset_function.
• Add kwargs option for map and similar functions.
• Skip optimization for multideal.
• Fix Vector comparisons so that they properly serve the dual purpose of element-wise comparison and truth value comparison.
• Fix multideal null pops.
• Experimental Deck.deal_groups().

v2.0.2 - 1 April 2025

• Fix default next_state_key identification.

v2.0.1 - 1 April 2025

• Adjusted behavior of MultisetEvaluator caching between calls to evaluate() depending on the value of next_state_key.
  • By default, next_state_key will only cache if the evaluator is called directly (i.e. not inside a @multiset_function).
  • If a full next_state_key is given, it will cache inside @multiset_function, provided the other contents agree.
  • The special value NoCache will disable caching.
• Support argumentless evaluations.
• Some typing fixes.

v2.0.0 - 31 March 2025

Major rewrite of multiset handling. Things will be more unstable than usual for a while.

• MultisetEvaluator.next_state() now has an explicit parameter with the order in which outcomes are seen.
• Optional MultisetEvaluator.initial_state() method.
• MultisetEvaluator.initial_state() and final_state() now get the following parameters:
  • The order in which outcomes are / were seen.
  • All outcomes that will be / were seen.
  • The sizes of the input multisets, if inferrable with counts being non-negative.
  • Non-multiset keyword arguments that were passed to evaluate().
• ascending and descending variants of next_state no longer exist.
  • Instead, raise UnsupportedOrder() if you don't like the current order. The other order will automatically be tried.
  • This can be done in initial_state() (recommended), next_state(), or final_outcome().
• Multiset operator order is now always attached to the evaluator side rather than the generator side.
  • Unless the operator modifies the generator in-place, but in this case both orders will certainly be supported.
• MultisetEvaluator can optionally provide a key for persistent caching.
• Some existing expressions and evaluators now take advantage of inferred multiset sizes.
  • In particular, keep() and sorted_match().
• MultisetExpression.count() renamed to size().
• @multiset_function now implements late binding like a standard Python method. (Though I still recommend using only pure functions.)
• @multiset_function now accepts variadic arguments.
• @multiset_function now accepts non-multiset keyword arguments.
• @multiset_function now works with joint evaluations where some sub-evaluations don't contain parameters.
• Hopefully better @multiset_function performance.
• Alignment class is retired.
• Deprecated depth=None is removed from Die.reroll().
• Multiset generators now always produce a single count value, with MultiDeal now producing tuple-valued counts rather than taking up multiple argument slots.
• Multiset computations now try to infer multiset sizes if the counts are non-negative. This improves the applicability of keep and sort_match expressions.
• Cartesian products (e.g. tupleize) now return Reroll if any argument is Reroll.

v1.7.2 - 25 February 2025

• Add Population.append() and .remove() methods.
• Improve Vector performance.
• Adjusted typing of mixture expressions.
• Experimental Wallenius noncentral hypergeometric.

v1.7.1 - 10 February 2025

• Fix joint evaluations and @multiset_function interaction with next_state_ascending and next_state_descending.

v1.7.0 - 9 February 2025

• Overhauled multiset expressions. This allows expressions that are given to an evaluator to have the evaluation persistently cached. This makes the caching behavior more consistent: a single expression will be cached in the final evaluator (e.g. (a - b).unique().sum() would be cached in the sum evaluator), and @multiset_function creates an evaluator like any other. Unfortunately, this did come at some performance cost for @multiset_function. I have some ideas on how to claw back some of the performance but I haven't decided whether it's worth the complexity.
• Instead of specifying order() for an evaluator, you can now implement next_state_ascending() and/or next_state_descending().
• Alignment now has a denominator of 1.
• keep(), isdisjoint(), sort_match(), and maximum_match() operations now treat negative incoming counts as zero rather than raising an error.
• Add Population.group_by() method to split a population into a "covering" set of conditional probabilities.
  • Population.group_by[] can also be used to group by index or slice.
• Move split() from Die to the base Population class.
• Straight-related multiset operations can now choose between prioritizing low and high outcomes.
• Store original names of @multiset_function parameters.

v1.6.2 - 23 December 2024

• Deprecate depth=None in favor of depth='inf'.
• Add experimental format_inverse() option that formats probabilities as "1 in N".

v1.6.1 - 17 November 2024

• Add pointwise_max, pointwise_min arguments to take pointwise maximum or minimum of CDFs.
• Add Die.time_to_sum() method.
• Fix identification of absorbing states in the presence of extra_args in map_and_time().
• Add time_limit parameter to map().
• repeat parameter now uses 'inf' to request the absorbing distribution rather than None.

v1.6.0 - 20 September 2024

• Breaking change: outcomes with zero quantities are removed when constructing Die and Deck.
  • Functions and methods relating to zero-quantities are removed: align(), align_range(), Population.has_zero_quantities(), Die.trim(), Die.set_range(), Die.set_outcomes().
  • You can use consecutive() or sorted_union() to get an appropriate superset of sets of outcomes.
• Breaking change: MultisetEvaluator.alignment() is renamed to MultisetEvaluator.extra_outcomes().
  • MultisetEvaluator.range_alignment() is renamed to MultisetEvaluator.consecutive().
  • The Alignment class is no longer public.
• Breaking change: Deck.multiply_counts() and Population.scale_quantities() are replaced/renamed to Population.multiply_quantities() etc.
• Add Deck.sequence() and Die.sequence() method.
• Add Population.pad_to_denominator() method.
• Move zero() and zero_outcome() from Die to Population.
• @ operator now sums left-to-right.
• Remove old compair evaluation.
• min_outcome() and max_outcome() free functions can now be called using a single iterable argument.
• Forward algorithm now has a persistent cache.
• Add skip optimization for single deals with keep tuples.
• Pools now only skip dice, not outcomes. This is a bit slower in some cases but provides more consistent iteration order.
• Add shared evaluator instances for some built-in evaluator for caching.
• Simplify determination of outcome order for multiset evaluations.
• Simplify implementation of generator unbinding.
• Fix extra_args expansion for map_and_time.

v1.5.0 - 23 August 2024

• Providing only a drop argument to lowest() or highest() will now keep all other elements rather than just the first non-dropped element.
• depth argument to Die.reroll() is now mandatory.
• tuple outcomes are now auto-tupleized again during Die construction.
• Add Die.stochastic_round() method.
• Add Die.reroll_to_pool() method.
• Add Die.keep() method. This works as MultisetGenerator.keep() with an implicit sum.
• Add percent option to Population.probability.
• Add new again_count mode for handling Again, which limits the total number of dice.
• Improved ability to keep from both ends for certain types of multiset expressions.
• Rename func parameters to function.
• MultisetExpression.order() is now public.
• Improved sorting for Symbols; now compares counts in alphabetical order.
• Experimental sort_match, maximum_match_highest, maximum_match_lowest expressions. sort_match replaces the compair evaluations.
• Experimental all_straights_reduce_counts and argsort multiset evaluations.
• Breaking change: nearest, quantity, quantities, probability, probabilities, keep_counts no longer have separate variants for each comparison; instead, they now take a comparison argument. quantities and probabilities now accept a comparison argument but no longer accept a list of outcomes.

v1.4.0 - 1 February 2024

• Rename keep_counts to keep_counts_ge. Add le, lt, gt, eq, and ne variants.
• Add count_subset evaluation that counts how many times the right side is contained in the left.
• Add ImplicitConversionError as subclass of TypeError.
• Add binary multiset operators to Deck.
• Add modulo_counts / % operation on multisets.
• Rebind generators and evaluate when fully bound non-generator expressions are given to an evaluator.
• Fix Symbols intersection.
• Fix argument order in __rfloordiv__.

v1.3.0 - 30 December 2023

• Fix Symbols operator priority with Population, AgainExpression.
• Added experimental map_to_pool and explode_to_pool methods.
• Split compair into compare_lt etc.
• Constructing a mixture of dice now effectively uses the old lcm_joint method, which reduces the denominator more aggressively.

v1.2.0 - 23 December 2023

• Experimental Symbols class representing a multiset of characters.
• marginals now forwards __getattr__ to outcomes, as long as the attribute name doesn't begin with an underscore.
• Operators on expressions now keep negative counts by default. The keep_negative_counts argument is retired.
• Add unary - for MultisetExpression.
• MultisetExpression.isdisjoint() now raises an error for negative counts.
• Small performance optimization for Vector.
• Population.marginals is no longer a Sequence, since the mixins don't make sense.
• Mappings are now properly excluded from Population.common_outcome_length.
• Fixed quoting in repr for populations.

v1.1.2 - 10 December 2023

• Add z(n), which produces a die that runs from 0 to n - 1 inclusive.
• Add Population.to_one_hot(), which converts the die or deck to a one-hot representation.
• Add Die.mean_time_to_sum(), which computes the mean number of rolls until the cumulative sum is greater or equal to the target.

v1.1.1 - 16 November 2023

• Fix non-fully-bound case of MultisetEvaluator.evaluate().
• Add default argument to lowest(), highest(), middle().
• Add Population.entropy().

v1.1.0 - 15 October 2023

• mean(), variance(), etc. now return an exact fractions.Fraction when possible. (Note that Fractions only support float-style formatting from Python 3.12.)
• Rename disjoint_union to additive_union.
• Add keep_negative_counts keyword argument to +, -, &, | binary operators for multiset expressions (default False).
• Symmetric difference (^) for multiset expressions is now a straight absolute difference of counts.
• Add unary + operator for multiset expressions, which is the same as keep_counts(0).

v1.0.0 - 22 July 2023

Improve some error messages.

v0.29.3 - 13 July 2023

Fix a bug in MultisetExpression.keep_outcomes() and drop_outcomes() regarding unbinding variables.

v0.29.2 - 12 July 2023

• MultisetExpression.map_counts() now accepts multiple arguments.
• MultisetExpression.keep_outcomes() and drop_outcomes() now accept an expression as an argument.
• MultisetExpression.highest_outcome_and_count() now returns the min outcome if no outcomes have positive count.

v0.29.1 - 2 July 2023

• highest, lowest, and middle can now take a single iterable argument.
• Add all_straights evaluation.
• all_counts and all_straights now output sizes in descending order.
• Harmonize method and free function versions of map. Both now allow extra arguments and repeat.
• Rename filter_counts to keep_counts.
• expand evaluator now allows order to be set.
• Experimental compair evaluation.

v0.29.0 - 30 April 2023

• Add HTML and BBCode options for population formatting.
• Renamed apply to map and the decorator version to map_function.
• The above now uses guess_star.
• Add default of 1 die for Die.pool().

v0.28.1 - 23 April 2023

Fix mathematical bug in Die.reroll for limited depth.

v0.28.0 - 16 April 2023

Retired implicit elementwise operations on tuples. This is now handled by a new explicit Vector container. cartesian_product() is replaced by two new functions: tupleize() and vectorize(). These compute the Cartesian product and produce a tuple or Vector respectively (or a die/deck thereof).

The Again symbol is now used without parentheses. The name of the underlying type is now AgainExpression.

die.zero() now multiplies all outcomes by 0 to determine the zero-outcome rather than using the default constructor. This allows it to work with Vector's elementwise operations while still producing the expected result for tuples, strings, etc.

Retired the linear algorithm for comparators for Die. While the quadratic algorithm is slower, it allows for non-bool outputs.

die.tuple_len() renamed to common_outcome_length. Now applies to all sized outcome types.

v0.27.1 - 9 April 2023

• Counts type Qs is now invariant with more detailed typing in Deal.

v0.27.0 - 29 March 2023

Incremented two versions because I messed up the last version number.

• commonize_denominator visible at top level.

v0.25.6 - 28 March 2023

• Rename from_cumulative_quantities to from_cumulative and allow die inputs.
• Mark multiset_function experimental again and note more caveats in the docstring.

v0.25.5 - 11 March 2023

• Add missing variants of nearest and quantities methods.
• Add optional outcomes argument to quantities and probabilities methods.

v0.25.4 - 17 February 2023

• map and similar functions will attempt to guess star.
• Changed positive_only parameter to expression.all_counts to filter.
• Changed implementation of bound generators to unbind expressions rather than dynamically splitting.
• Retired apply_sorted and variants.
• Add outcome_function decorator similar to apply.
• Add map_counts expression.
• Reroll in tuple outcomes and joint evaluations causes the whole thing to be rerolled.

v0.25.3 - 28 January 2023

• Tuple outcomes can now be compared with single outcomes.
• Add .keep, .highest, lowest, .middle variants of apply_sorted.
• Recommend multiset_function be used as a decorator, add update_wrapper.
• Add keep_outcomes, drop_outcomes methods to expressions.
• Add any evaluation to expresions.

v0.25.2 - 23 January 2023

Comparisons on dice with tuple outcomes are now performed elementwise.

v0.25.1 - 23 January 2023

Testing GitHub workflows.

0.25.0 - 22 January 2023

Expanded multiset processing with multiset expressions.

• OutcomeCountGenerator, OutcomeCountEvaluator renamed to MultisetGenerator, MultisetEvaluator.
• multiset_function is an easy way to create joint evaluators.
• Pool indexing is now relative rather than absolute.
• Renamed pool.sorted_roll_counts to pool.keep and pool.keep_tuple.
• Die versions of sum_highest etc. renamed to just highest; these always return dice.
• MultisetExpression (including generators like Pool) have highest() returning an expression.
• Die operators now does mixed vector-scalar binary operations by broadcasting the scalar.
• Add middle() methods.
• Remove *generators argument from evaluator.order(), evaluator.final_outcome().
• *generators argument of evaluator.alignment replaced with the union of all generator outcomes.
• Removed suits.
• Stop using __class_getitem__, which is intended for typing only.

0.24.0 - 8 January 2023

Reworked built-in generators and evaluators.

• Generators now support multiset operations. These include:
  • Multiset comparisons (<, <=, >, >=, ==, !=, isdisjoint), which produce Die[bool].
  • Multiset operators (+, -, |, &, ^) which produce wrapped generators.
  • A suite of chainable count adjustments, including multiply_counts, divide_counts, filter_counts, and unique.
• Rename median_left to median_low etc.
• Generators and evaluators are now paramterized by count type as well.
• Move concrete evaluators to a submodule.

0.23.3 - 31 December 2022

• Fixed weighting bug in __matmul__ when the left die has an outcome of 0.
• Retired the names standard and bernoulli. These will be just d and coin respectively.
• apply now accepts Pool arguments, which will have expand() called on them.
• Reinstate automatic Cartesian product in Population construction.
• if_else now runs in two stages.

0.23.2 - 30 December 2022

• Incremental sorting for all_matching_sets to reduce state space.
• lowest() and highest() now actually visible.
• Improved checking for tuple outcome sortability and types.

0.23.1 - 29 December 2022

Prepend sum_ to OutcomeCountGenerator versions of highest and lowest.

0.23.0 - 29 December 2022

Expanded typing, particularly in terms of parameterizing types.

• Die and Deck constructors now expect exactly one nesting level, e.g. a Sequence of outcomes.
• Removed Die.bool().
• **kwargs replaced with explcit again_depth, again_end. Removed OutcomeCountEvaluator.final_kwargs().
• Split include_steps version of Die.map() into a map_and_time method.
• Split include_outcome version of best_matching_set and best_straight into separate methods.
• Rename keep_highest and highest to sum_highest; add just highest for taking the single highest.
• Same for lowest.
• Added cartesian_product(), took this functionality out of the Die constructor for now.
• Added OutcomeCountGenerator.all_matching_sets().

0.22.0 - 19 December 2022

• Die.sub() renamed to Die.map().
• Die.map() can now include the number of steps taken until absorption.
• Die.reroll_until() renamed to Die.filter().
• wilds arguments marked experimental, pending decision on how ordering should affect them.
• Only tuples get separate columns in tables and not str or bytes.
• Non-recursive algorithm for Again() handling.

0.21.0 - 3 December 2022

• Nested lists are now allowed in the Die() constructor.
• Single outcomes can be sent to the Die() constructor without wrapping them in a list.
• Same for sending a die to the Pool() constructor.
• Add Die.probabilities_lt(), Die.probabilities_gt() methods.
• Add wilds argument to contains_subset() and intersection_size() methods.
• Removed extra_dice argument from Die.sub().
• Added sub() method to Deck, though with less features than the Die version.
• sum() method of generators now has a sub option that maps outcomes before summing.
• expand() method of generators now has a unique option that counts duplicates only once.
• Forwarding of kwargs to Die() replaced with explicit arguments.
• Add comparators to Again.
• Experimental absorbing Markov chain analysis for Die.sub(depth=None).

0.20.1 - 10 September 2022

• Added one_hot function.
• Added experimental suit generator that wraps a generator and produces counts for all suits for each value.

0.20.0 - 6 September 2022

• Retired denominator_method.
• Renamed max_depth parameters to just depth.
• Binary operators delegate to Again's behavior.
• Only single outcomes implicity convert to Die.
• marginals is now a Sequence, and can be iterated over, unpacked, etc.
• Die.sub() now expands extra die arguments into their outcomes.

0.19.1 - 21 August 2022

Fix contains_again checking of sequences.

0.19.0 - 21 August 2022

New feature: Again(), a placeholder that allows to roll again with some modification.

• **kwargs are forwarded to the constructor of the yielded or returned die for sub, if_else, reduce, accumulate, apply, apply_sorted.
• Add optional final_kwargs method to evaluators.
• Rename max_depth parameter of sub() to repeat.

0.18.0 - 19 August 2022

• Rename Die.reduce() to Die.simplify() to avoid confusion with the free function reduce().
• Rename OutcomeCountEvaluator.direction() to order() and add explicitly named Order enums.
• Add is_in, count, and count_in methods to dice.
• Add built-in evaluators as convenience functions of OutcomeCountGenerator.

0.17.4 - 1 August 2022

• Fixes to max_depth=None case of sub().
• This is now marked experimental.

0.17.3 - 1 August 2022

• Pass star parameter to sub() to recursive calls.

0.17.2 - 1 August 2022

• sub() with max_depth=None now handles "monotonic" transitions with finite states. Full absorbing Markov chain calculation still under consideration.
• sub() no longer accept sequence input.
• Some minor formatting fixes.

0.17.1 - 25 June 2022

• Standardize outcome count of bernoulli/coin and comparators.
• standard_pool now accepts a dict argument.
• post_roll_counts renamed again to sorted_roll_counts.
• apply_sorted can be subscripted to set the sorted_roll_counts.
• Pools are no longer resizable after creation.
• Die.pool() now has mandatory argument, now accepts a sequence argument to set sorted_roll_counts.

0.17.0 - 19 June 2022

More renaming, experimental sample() methods.

• Both Die and Deck now have quantities rather than weight, dups, etc.
• Many Die methods moved to base class and are now available to Deck.
• "Eval" is now the full word "Evaluator".
• Parameter and method names are no longer prefixed with "num_".
• reduce_weights renamed back to reduce.
• Decks can be formatted like Die.
• Allow formatting 0-1 probability.
• Experimental sample() methods for OutcomeCountGenerator and OutcomeCountEvaluator.

0.16.1 - 18 June 2022

Development of deck API.

• Separate Deal class from Deck, roughly analogous to Pool vs. Die.
• Deal can now output multiple hands. The counts for each hand are provided in order to eval.next_state.
• comb_row now uses iterative rather than recursive memoization. This will prevent some stack overflows.
• Pools are not permitted to be constructed using a raw Die or Deck argument.
• reduce argument of Die.equals() renamed to reduce_weights.

0.16.0 - 17 June 2022

Significant API changes, experimental deck support.

• Experimental Deck class. API still very unstable.
• Pools, Decks, and internal Alignment now have a common base class OutcomeCountGen.
• EvalPool renamed to OutcomeCountEval.
• Die and Deck are now proper Mappings with keys, values, and items.
• The above view types can also be accessed like sequences.
• A Die is now always considered not equal() to non-Die.
• die[] now works like a dict. Use marginals[] to marginalize dimensions.
• Die, Pool, and Deck now take a single sequence or mapping argument rather than a variable number of arguments.
• Die, Pool, and Deck now all have the same name for the second argument: times.
• count_dice renamed to post_roll_counts. No longer accepts None, use [:] instead.
• Linear algorithm for comparators on Die.
• Improvements to internal Count class.
• Add clear_pool_cache function.
• Forward *extra_args for reroll, reroll_until, explode.

0.15.0 - 10 June 2022

Added type hints. Now requires Python 3.10 or later.

Other changes:

• Add apply_sorted() method.
• Add Die.set_range().
• standard() / d() argument is now positional-only.

0.14.1 - 30 May 2022

Reinstate alternate internal EvalPool algorithm, which provides better performance in some cases.

0.14.0 - 30 May 2022

• Added a new EvalPool.alignment() method. This allows to specify an iterable of outcomes that should always be seen by next_state even if they have zero count.
• The free function d() is now simply an alias for standard().
• Removed Die.d().
• The @ operator now casts the right side to a Die like other operators.
• Some internal changes to EvalPool algorithm.

0.13.2 - 23 May 2022

The data of a die resulting from == or != is lazily evaluated.

This saves computation in case the caller is only interested in the truth value.

0.13.1 - 23 May 2022

EvalPool favors the cached direction more.

0.13.0 - 23 May 2022

Major reworking of pool construction.

• Public constructor is now just Pool(*dice).
• In particular, no more truncate_min or truncate_max arguments.
• Pools can be of arbitrary dice, though non-truncative sets of dice will have lower performance. There is some performance penalty overall.
• apply() called with no arguments now calls func once with no arguments.

0.12.1 - 21 May 2022

• Removed Die.keep(). Use Die.pool(...).sum().
• highest/lowest returns empty die if any of the input dice are empty.

0.12.0 - 18 May 2022

• Free-function form of lowest, highest now selects between algorithms for better performance and generality.
• Removed die.lowest, die.highest`.
• die.min_outcome, die.max_outcome no longer take arguments beyond self.
• No more (public) repeat_and_sum, as this is redundant with @ and die.keep().
• No more public die.keep_lowest_single(), die.keep_highest_single().
• count_dice arguments are now keyword-only.
• die.zero() no longer reduces weights to 1.
• Update PyPi classifiers.

0.11.0 - 11 May 2022

• Removed min_outcome parameter from die construction.
• Operations on tuple outcomes are now performed recursively, to match the fact that die expansion on construction is recursive.
• Renamed Die.reduce() to Die.reduce_weights().
• Added reduce() and accumulate() functions analogous to the functions of the same name from functools/itertools.
• Add CSV output.
• Renamed Die.markdown() to Die.format_markdown().
• Added star parameter to sub, explode, reroll, reroll_until methods. If set, this unpacks outcomes before giving them to the supplied function.
• Added experimental JointEval class for performing two evals on the same roll of a pool.

0.10.2 - 8 May 2022

• Operators other than [] are performed element-wise on tuples.
• Rename DicePool to just Pool. Merge the old factory function into the constructor.

0.10.1 - 8 May 2022

• Fix denominator_method='reduce' in die creation.
• Fix outcomes consisting of empty tuple ().
• apply() with no dice produces an empty die.

0.10.0 - 7 May 2022

Retired the EmptyDie / ScalarDie / VectorDie distinction.

• There is now only one Die class; it behaves similarly to how ScalarDie used to.
• There is no more ndim.
• The [] operator now forwards to the outcome, acting similar to what VectorDie.dim[] used to do.
• Removed PoolEval.bind_dice(). It was cute, but I'm not convinced it was worth spending API on.

0.9.1 - 1 May 2022

• This will probably be the last version with a VectorDie distinction.
• Dice cannot have negative weights.
• VectorDie cannot be nested inside tuple outcomes.

0.9.0 - 28 April 2022

• Die and dict arguments to Die() are now expanded, including when nested.
• Add Die.if_else() method, which acts as a ternary conditional operator on outcomes.
• Dice are now hashable. == and != return dice with truth values based on whether the two dice have identical outcomes and weights.
• ndim now uses singletons icepool.Scalar and icepool.Empty.

0.8.0 - 23 April 2022

• EvalPool.eval() can now be provided with single rolls of a pool. This can be a dict-like mapping individual die outcomes to counts or a sequence of individual die outcomes.
• EvalPool.next_state() can not expect that the outcomes it sees are consecutive, though they are guaranteed to be seen in monotonic order.
• FindBestRun() no longer assumes consecutive outcomes.
• Added DicePool.sample().
• Added die.truncate().
• min_outcomes, max_outcomes (for pool definitions) renamed to truncate_min, truncate_max.
• Removed die.getitem().
• DicePool is no longer iterable, since there isn't an intuitive, unambiguous way of doing so.
• align_range() now only operates on scalar outcomes.

0.7.0 - 16 April 2022

• Renamed from hdroller to icepool.
• Primary repository is now https://github.com/HighDiceRoller/icepool.
• Only tuples become VectorDie.
• Weights of 10^30 or above are not shown in tables by default.

0.6.7 - 18 March 2022

• Add max_depth parameter to sub(). If set to None this seeks a fixed point.
• Add 'reduce' option for denominator_method parameters.
• Cache results of repeat_and_sum().

0.6.6 - 24 February 2022

• Add ndim keyword argument to d().
• Removed hitting_time() method; it seems too niche to commit to.
• Several arguments are now keyword-only.
• Fix to reroll_until() for vector dice.

0.6.5 - 20 February 2022

• Add VectorDie.all() and VectorDie.any().
• More fixes to ndim calculation.

0.6.4 - 20 February 2022

• EmptyDie (a die with no outcomes) is now its own class.
• mix() along with its weights argument is folded into the Die() factory.
• Fixes to ndim calculation.
• apply(), EvalPool.final_outcome() can now return dice, hdroller.Reroll, etc.
• Center Ellipsis when indexing a pool can now work on undersized pools, rather than being an error.
• Fix lcm_weighted denominator method.
• len(die) is now the same as die.num_outcomes() again.
• Slicing a die now refers to (sorted) outcome indexes.
• Dimension slicing is now VectorDie.dim[select].

0.6.3 - 18 February 2022

• Die() now takes a variable number of arguments. A sequence provided as a single argument will be treated as a single outcome equal to that sequence. To have one outcome per item in the sequence, the sequence must be unpacked into multiple arguments.
• min_outcome and ndim args to Die() are now keyword-only.
• Non-integers are now allowed for count_dice. Use at your own risk.
• The third argument to a count_dice slice, if provided, changes the size of the pool.
• Up to one Ellipsis (...) may now be used in count_dice. This allows the tuple option to adapt to differently-sized pools.
• VectorDie outcomes are now unpacked when providing them to callable arguments in reroll(), explode(), split().
• if_else() is now a method of ScalarDie rather than a free function.
• Fixed the invert operator.

0.6.2 - 16 February 2022

• Rerolls in pools, mix(), and sub() are now triggered by a sentinel value hdroller.Reroll (rather than None).
• Single-int indexes for count_dice and pool [] operator now produce a die rather than a pool, in analogy to sequence indexing. However, note that this is absolute relative to the whole pool rather than relative to any previous indexing.
• EvalPool can now iterate in both directions even if one of max_outcomes or min_outcomes is given, though at lower efficiency in the disfavored direction.
• Default such direction changed back to ascending.
• Re-implemented clip().
• The minimum outcome of highest() is equal to the highest minimum outcome among all input dice and likewise for lowest(). This will prevent introducing zero-weight outcomes where they did not exist before.

0.6.1 - 12 February 2022

• Pool max_outcomes or min_outcomes outside the range of the fundamental die is now an error.
• Pools no longer automatically shrink the die to fit max_outcomes or min_outcomes. This is to guarantee a predictable iteration order.
• Improved str(die) formatting.
• VectorDie outcomes are automatically converted to tuples.
• ScalarDie now has ndim='scalar' rather than ndim=False.
• Add standard_pool() function for easy creation of standard pools.
• Removed reroll_lt(), etc. as they weren't enough faster than reroll(func) to justify spending extra space.

0.6.0 - 9 February 2022

• Removed initial_state() in favor of using None as the initial state.
• Added simple pool.eval() method.
• None now rerolls in EvalPool.
• Fix trunc, floor, ceil.
• Various improvements to pool algorithm.
• Add some thresholding methods for dice.
• Max/min renamed to highest/lowest.
• Improve empty die handling.
• Improved VectorDie string formatting.
• Disabled reverse().

0.5.2 - 6 February 2022

• Faster algorithm for keeping the single highest/lowest.
• Fix from_sweights().

0.5.1 - 6 February 2022

• Fix max() and min() casting to dice.

0.5.0 - 6 February 2022

• Capitalize Die() and Pool() factory methods.
• PoolEval renamed to EvalPool.
• Implemented min_outcomes for pools. Direction is now determined by EvalPool.
• Allow dice with zero weights or zero outcomes.
• BaseDie subclasses are now ScalarDie and VectorDie.
• == and != compare outcomes again. Dice are no longer hashable by the standard hash().
• Rewored Pool() arguments.
• Added several methods.

0.4.4 - 2 February 2022

• Fix weights_le() etc.
• set_outcomes() is now public.

0.4.3 - 2 February 2022

• min_outcome(), max_outcome() now accept multiple dice.
• Fix slice range going below zero.
• num_dice is now an optional parameter for keep*() and pool(). Exactly one out of num_dice, min_outcomes, and max_outcomes should be provided.
• Removed right-side cast for @ operator, as it leads to confusion with either other operator casting or d() casting.
• Replaced key-value sequence option for die() with a sequence of weight-1 outcomes.
• Add has_zero_weights() method to dice.
• align(), align_range() are now public.

0.4.2 - 31 January 2022

• Remove numpy install dependency.

0.4.1 - 31 January 2022

• Use the @ operator and d() instead of * for "roll the die on the left, then roll that many dice on the right and sum".
• * operator now multiplies.
• Add / operator.
• Add median_left(), median_right(), ppf(), ppf_left(), ppf_right(), sample(), cmp(), sign().

0.4.0 - 30 January 2022

No longer numpy-based. Major changes:

• API completely reworked.
• Weights are Python ints. This decreases performance but produces all exact results.
• New MultiDie class representing multivariate distributions.
• PoolScorer classes merged to a single PoolEval class.

0.3.3 - 27 January 2022

This will most likely be the last numpy-based version.

• Pool scorers skip zero-weight outcomes.
• Removed __bool__ since zero-weight dice are no longer allowed.

0.3.2 - 25 January 2022

• Pool scorers can now return None states, which drops the state from evaluation, effectively performing a full reroll.
• Operations that result in no remaining weight (e.g. rerolling all outcomes) now return None.
• Dice can no longer be constructed with zero total weight.
• The Die() constructor now auto-trims.
• Die._align() is now private and no longer makes all the dice the same weight.
• Fixed Die.explode() weighting.
• Cleaned up unused functions.

0.3.1 - 24 January 2022

• Optimize Pool.pops() by immediately removing all the dice if there are no left in the mask.
• Die.reroll now tracks weights properly.

0.3.0 - 23 January 2022

• Implemented new keep/pool algorithms, which make heavy use of caching. The keep*() function signatures changed.
• Die is now hashable.
• Empty Die is now an error.
• Removed weight normalization.
• Added % and // operators to Die.
• Renamed ccdf to sf following scipy's naming.

0.2.2 - 30 October 2021

• best_set now queries the score_func to determine possible outcomes, which allows more flexibility.

0.2.1 - 30 October 2021

• Die.align() and Die.trim() are now public.
• Optimize keep() algorithm by skipping to results after passing the last kept die.
• Add best_set() method for computing sets of matching dice.

0.2.0 - 22 October 2021

The major change in this version is a new roll-and-keep algorithm. This can keep arbitrary indexes, as well as supporting dissimilar dice as long as they share a common "prefix" of weights.

0.1.1 - 21 September 2021

0.1.0 - 21 September 2021

Initial version.