pub trait Score:
Copy
+ Debug
+ Display
+ Default
+ Send
+ Sync
+ PartialEq
+ Eq
+ PartialOrd
+ Ord
+ Add<Output = Self>
+ Sub<Output = Self>
+ Neg<Output = Self>
+ 'static {
Show 24 methods
// Required methods
fn is_feasible(&self) -> bool;
fn zero() -> Self;
fn levels_count() -> usize;
fn level_number(&self, index: usize) -> i64;
fn from_level_numbers(levels: &[i64]) -> Self;
fn multiply(&self, multiplicand: f64) -> Self;
fn divide(&self, divisor: f64) -> Self;
fn scale(&self, factor: i64) -> Self;
fn abs(&self) -> Self;
fn to_scalar(&self) -> f64;
fn level_label(index: usize) -> ScoreLevel;
// Provided methods
fn to_level_numbers(&self) -> Vec<i64> { ... }
fn overflowing_add(self, other: Self) -> (Self, bool) { ... }
fn overflowing_sub(self, other: Self) -> (Self, bool) { ... }
fn overflowing_neg(self) -> (Self, bool) { ... }
fn overflowing_scale(&self, factor: i64) -> (Self, bool) { ... }
fn overflowing_multiply(&self, multiplicand: f64) -> (Self, bool) { ... }
fn compare(&self, other: &Self) -> Ordering { ... }
fn is_better_than(&self, other: &Self) -> bool { ... }
fn is_worse_than(&self, other: &Self) -> bool { ... }
fn is_equal_to(&self, other: &Self) -> bool { ... }
fn one_hard() -> Self { ... }
fn one_soft() -> Self { ... }
fn one_medium() -> Self { ... }
}Expand description
Core trait for all score types in SolverForge.
Scores represent the quality of a planning solution. They are used to:
- Compare solutions (better/worse/equal)
- Guide the optimization process
- Determine feasibility
All score implementations must be:
- Immutable (operations return new instances)
- Thread-safe (Send + Sync)
- Comparable (total ordering)
§Score Levels
Scores can have multiple levels (e.g., hard/soft constraints):
- Hard constraints: Must be satisfied for a solution to be feasible
- Soft constraints: Optimization objectives to maximize/minimize
When comparing scores, higher-priority levels are compared first.
§Overflow
Score arithmetic never wraps. Add, Sub, Neg, Score::abs,
Score::scale, Score::multiply and Score::divide on the built-in
score types panic, in every build profile, when a level does not fit
i64; multiply and divide also panic for a factor that is not finite
(or a zero divisor). A wrapped level would flip sign and turn a penalty
into a reward.
Required Methods§
Sourcefn is_feasible(&self) -> bool
fn is_feasible(&self) -> bool
Returns true if this score represents a feasible solution.
A solution is feasible when all hard constraints are satisfied (i.e., the hard score is >= 0).
Sourcefn levels_count() -> usize
fn levels_count() -> usize
Returns the number of score levels.
For example:
- SoftScore: 1 level
- HardSoftScore: 2 levels
- HardMediumSoftScore: 3 levels
Sourcefn level_number(&self, index: usize) -> i64
fn level_number(&self, index: usize) -> i64
Returns the score value at a priority level.
Level indices follow the same order as to_level_numbers() and
level_label(): highest priority first. This is the primitive
allocation-free score-level accessor used by score-level search behavior.
§Panics
Panics if index >= levels_count().
Sourcefn from_level_numbers(levels: &[i64]) -> Self
fn from_level_numbers(levels: &[i64]) -> Self
Creates a score from level numbers.
§Panics
Panics if the number of levels doesn’t match levels_count().
Sourcefn multiply(&self, multiplicand: f64) -> Self
fn multiply(&self, multiplicand: f64) -> Self
Multiplies every level by a scalar, rounding to the nearest integer.
§Panics
Panics when multiplicand is not finite or a rounded level is outside
i64.
Sourcefn divide(&self, divisor: f64) -> Self
fn divide(&self, divisor: f64) -> Self
Divides every level by a scalar, rounding to the nearest integer.
§Panics
Panics when divisor is zero or not finite, or a rounded level is
outside i64.
Sourcefn scale(&self, factor: i64) -> Self
fn scale(&self, factor: i64) -> Self
Multiplies every level by an exact integer factor.
Unlike multiply(f64), this never rounds and stays exact for every i64
level value. Constraint weight multipliers are applied through it.
§Panics
Panics when any level overflows i64, in every build profile.
Sourcefn abs(&self) -> Self
fn abs(&self) -> Self
Returns the absolute value of every level.
§Panics
Panics when a level is i64::MIN, whose absolute value is not an i64.
Sourcefn to_scalar(&self) -> f64
fn to_scalar(&self) -> f64
Converts this score to a single f64 scalar value.
Higher-priority levels are weighted with larger multipliers to preserve their dominance. Used for simulated annealing temperature calculations.
Sourcefn level_label(index: usize) -> ScoreLevel
fn level_label(index: usize) -> ScoreLevel
Returns the semantic label for the score level at the given index.
Level indices follow the same order as to_level_numbers():
highest priority first.
§Panics
Panics if index >= levels_count().
Provided Methods§
Sourcefn to_level_numbers(&self) -> Vec<i64>
fn to_level_numbers(&self) -> Vec<i64>
Returns the score values as a vector of i64.
The order is from highest priority to lowest priority. For HardSoftScore: [hard, soft].
This is an allocating convenience view derived from level_number().
Sourcefn overflowing_add(self, other: Self) -> (Self, bool)
fn overflowing_add(self, other: Self) -> (Self, bool)
Adds other without panicking, returning the sum and whether any level
overflowed i64 (the levels are then wrapped and meaningless).
The scoring engine accumulates a batch of deltas this way and checks the
flag once, instead of branching on every addition. The default delegates
to Add and reports false, so a custom score keeps its own overflow
behavior.
Sourcefn overflowing_sub(self, other: Self) -> (Self, bool)
fn overflowing_sub(self, other: Self) -> (Self, bool)
Subtracts other without panicking; see overflowing_add.
Sourcefn overflowing_neg(self) -> (Self, bool)
fn overflowing_neg(self) -> (Self, bool)
Negates without panicking; see overflowing_add.
Sourcefn overflowing_scale(&self, factor: i64) -> (Self, bool)
fn overflowing_scale(&self, factor: i64) -> (Self, bool)
scale without panicking: the flag is set when a level overflowed;
see overflowing_add.
Sourcefn overflowing_multiply(&self, multiplicand: f64) -> (Self, bool)
fn overflowing_multiply(&self, multiplicand: f64) -> (Self, bool)
multiply without panicking: the flag is set for a non-finite
multiplicand or a rounded level outside i64; see overflowing_add.
Sourcefn compare(&self, other: &Self) -> Ordering
fn compare(&self, other: &Self) -> Ordering
Compares two scores, returning the ordering.
Default implementation uses the Ord trait.
Sourcefn is_better_than(&self, other: &Self) -> bool
fn is_better_than(&self, other: &Self) -> bool
Returns true if this score is better than the other score.
In optimization, “better” typically means higher score.
Sourcefn is_worse_than(&self, other: &Self) -> bool
fn is_worse_than(&self, other: &Self) -> bool
Returns true if this score is worse than the other score.
Sourcefn is_equal_to(&self, other: &Self) -> bool
fn is_equal_to(&self, other: &Self) -> bool
Returns true if this score is equal to the other score.
Sourcefn one_medium() -> Self
fn one_medium() -> Self
Returns a score with 1 at the first Medium-labeled level and 0 elsewhere.
Dyn Compatibility§
This trait is not dyn compatible.
In older versions of Rust, dyn compatibility was called "object safety".