Knotted Humeri were less useful in the past, so setting their cost to
zero reflected the tendency to have an overabundance of them. They've
become more practical over time, however, so the cost now assumes you go
out of your way for them.
Segmented Ribcages are now obtainable from Port Cecil.
Since the byproduct of the carousel that produces them is
disproportionately large, it has been factored into the cost.
Segmented Ribcages can be used to add more limb slots, but carry a
variety of unusual requirements. In addition to a scaling cost in
Nevercold Brass, they require an unfilled tail slot when applied.
It may be prudent to clean up the partial sum formula using helper
methods in the future, as it currently requires a considerable number of
intermediate terms.
itertools.repeat is now being used to perform exponentiation, rather
than repeating terms directly.
The Segmented Ribcage has no known source outside a specific Whitsun
egg, much like the Obsidian Chitin Tail. As such, it has been given a
placeholder price that prevents it from being used normally.
The cost of adding joints is dependent on the current amount, which may
be affected by both the torso and skulls.
The new implementation still ignores the effects of other Appendage
actions, since joints may be added before doing anything else. This may
warrant revisiting in the future.
OR-Tools 9.3 no longer seems to use more than one worker by default, so
this parameter is once again set by the script.
The parameter has also been changed from num_search_workers to
num_workers internally, though the former is mapped to the latter.
Reusing non-constant integer variables was causing constraints to
propagate in unintended ways, resulting in incorrect restrictions on the
solver's output.
This should not be an issue for constants, as they are already
restricted to one value.
Intermediate variables should not be necessary if the equivalent
expression is a constant or a variable with the same bounds.
This change reduces the total number of variables in the model.
The documentation of BoneMarketModel.NewIntermediateIntVar has been
tweaked to allow for any number of constraints.
The new version of OR-Tools does not necessitate major changes, but has
some performance improvements and support for Python 3.10.
On a related note, Python 3.10 is now required.
Pipfile.lock is ignored because it won't work between different
platforms if the dependency requirements diverge.
Python compilation products are ignored, as they are derived directly
from Python source files.
Buyer requirements now use helper methods and control flow to increase
readability and compactness.
The main benefit of this is no longer having to declare intermediate
variables.
For the sake of maintainability, it is very important the parameters and
results of the helper functions are spelled out explicitly.
Union type hints are being left until Python 3.10 is supported.
This function has been obsoleted by the ability to pass partial methods
as parameters to helper functions.
Use of this function has been replaced with partial methods accordingly.
In addition, the following code style has been applied: partial methods that
contain a nested partial method have the function arguments broken into
distinct lines. This means that the bottom partial method is formatted in the
same manner as a partial method without any nesting.
This function has been obsoleted by the ability to pass partial methods
as parameters to helper functions.
Use of this function has been replaced with partial methods accordingly.
This has allowed some intermediate variables to be removed entirely,
allowing for further simplification.
Intermediate integer variables can now be created for compatible partial
methods, which massively increases the expressiveness of each helper
function.
This change should allow the compound helper functions, such as
AddDivisionMultiplicationEquality, to be removed entirely. It will also
allow even the most complex buyers to avoid intermediate variables.
The Trifling Diplomat fascinations that previously exhibited quasi-cubic
scaling have a new formula using the mean of amalgamy, antiquity, and
menace.
To eliminate unnecessary reptition, all fascinations have been
completely refactored and combined using the new helper methods and design patterns.
A comment has been added to mark The Trifling Diplomat's code, since it
is noticeably less readable than that of other refactored buyers.
This update, despite ostensibly being a minor version, includes breaking
changes that must be accounted for.
Overflow is much more strictly checked, so the magnitude of certain
constants has been decreased.
CP-SAT's default number of workers has been changed to reflect the
default of this script. As such, the script no longer needs to change
that parameter unless a specific number of workers has been specified.
In light of the breaking changes, the OR-Tools version is now pinned at
9.1 instead of being permitted to use future minor versions.
Buyer requirements now use helper methods and control flow to increase
readability and compactness.
The main benefit of this is no longer having to declare intermediate variables directly.
This is a work-in-progress, as some buyers have complex formulae that
necessitate more work or additional helper methods to refactor.
Amalgamy, antiquity, menace, and implausibility are now clamped to be
non-negative. This simplifies a number of calculations, eliminates a few
insiduous calculation errors, and reflects the actual game.
This change impacts the output of the program, which previously showed
negative values even though they weren't actually possible.
Declaration requirements now use helper methods and generator
expressions to increase readability and compactness.
Requirements are ordered and grouped by quantity.
The comments on declaration requirements have been removed, as the new
pattern is far easier to read.
This method allows a series of constraints to be applied all at once
using the same enforcement literal, which can substantially improve
readability and writability.
The replacement method, BoolExpression, accepts a
BoundedLinearExpression directly, and uses memoization to reuse Boolean
variables. This technique is unsuitable for other methods, as
constraints need to be applied separately, but in this case it is fine.
The new method is much easier to understand, and far easier to read at
call sites.
Variable names are generated from the expression, rather than being
explicitly specified.
The new subclass, BoneMarketModel, overrides CpModel's methods with
improvements such as default parameters and support for enforcement
literals.
This should allow substantial improvements to the readability of the
solver in the future.
In addition, various lists have been replaced with tuples where
appropriate.
Jeremy Saklad