Skip to content

Enhancement/Sum constraints on NumberNode#471

Open
fastbodin wants to merge 34 commits intodwavesystems:mainfrom
fastbodin:enhancement/axis_wise_bounds_numbernode
Open

Enhancement/Sum constraints on NumberNode#471
fastbodin wants to merge 34 commits intodwavesystems:mainfrom
fastbodin:enhancement/axis_wise_bounds_numbernode

Conversation

@fastbodin
Copy link
Contributor

@fastbodin fastbodin commented Jan 30, 2026

Do not merge. Draft PR. Only C++ has been implemented. Simply checking CircleCi.

@fastbodin fastbodin force-pushed the enhancement/axis_wise_bounds_numbernode branch 2 times, most recently from bebd145 to 15f7b67 Compare February 3, 2026 23:11
@fastbodin
Copy link
Contributor Author

Feature is fully implemented.

@fastbodin fastbodin force-pushed the enhancement/axis_wise_bounds_numbernode branch from e3b0255 to 1c177b4 Compare February 4, 2026 00:02
@fastbodin fastbodin marked this pull request as ready for review February 4, 2026 20:24
@fastbodin
Copy link
Contributor Author

fastbodin commented Feb 4, 2026

Closes #216.

@arcondello arcondello requested a review from wbernoudy February 4, 2026 22:49
@arcondello arcondello added the enhancement New feature or request label Feb 4, 2026
@fastbodin
Copy link
Contributor Author

First round of comments addressed.

@fastbodin fastbodin force-pushed the enhancement/axis_wise_bounds_numbernode branch from 4b2205c to c6a93d5 Compare February 5, 2026 06:37
@wbernoudy
Copy link
Member

I do see "hyperslice" used in a lot of comments. Would prefer that is changed to just "slice" as well, or at least a small comment maybe in the header numbers.hpp on what that term means here.

@fastbodin fastbodin force-pushed the enhancement/axis_wise_bounds_numbernode branch from c6a93d5 to 72c6d96 Compare February 6, 2026 02:05
@fastbodin
Copy link
Contributor Author

Second round of comments addressed.

Copy link
Member

@arcondello arcondello left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks great! Just two aesthetic comments

@fastbodin
Copy link
Contributor Author

Third round of comments addressed.

Data is stored at C++ level with the class `AxisBoundInfo` as private
attribute to `NumberNode`. Added relevant C++ tests.
For each bound axis and each hyperslice along said axis, we store the
running sum of the values within the hyperslice. This state dependant
data is stored via `NumberNodeStateData`. If `NumberNode` is initialized
with values, we check that all axis-wise bounds are satisfied.
Added satisfies_axis_wise_bounds(), update_bound_axis_slice_sums(),
axis_wise_bounds(), and bound_axis_sums() to NumberNode. Updated various
NumberNode, IntegerNode, and BinaryNode methods to reference
NumberNodeStateData as opposed to ArrayNodeStateData. Updated all
NumberNode mutate methods to reflect changes to the axis-wise bound
running sums. Added C++ tests to check said mutate methods on
BinaryNode and IntegerNode.
Make use of BufferIterators to compute the sum of the values within each
hyperslice along each bound axis as opposed making a custom method to do
this.
`BoundAxisInfo` -> `AxisBound` and `BoundAxisOperator` -> `Operator`.
`Operator` is now a nested enum classs of `AxisBound`.
Added indicator variable that all bound axis operators are `==`
to reduce redundancy in `NumberNode::exchange()` method.
Made axis, operators, and bounds private members. Added axis(),
num_bounds(), and num_operators() methods. Updated C++ code/tests,
Python, and Cython to reflect this.
Removed `asserts()` that axis-wise bounds were satisfied
to `NumberNode::propagate()`.
@fastbodin fastbodin force-pushed the enhancement/axis_wise_bounds_numbernode branch from a92e2bd to 116d03b Compare February 11, 2026 20:36
@fastbodin fastbodin force-pushed the enhancement/axis_wise_bounds_numbernode branch from f2b1cc4 to 3b10b41 Compare February 12, 2026 23:12
Previously, users could not define a bound for the sum of the values
over an entire `NumberNode` array. This lead to the undesired behaviour
that users could not define a bound on the sum of the values in a
`NumberNode` vector.
@fastbodin
Copy link
Contributor Author

Previously, users could not define a bound for the sum of the values over an entire NumberNode array. This lead to the undesired behaviour that users could not define a bound on the sum of the values in a NumberNode vector.

@fastbodin fastbodin force-pushed the enhancement/axis_wise_bounds_numbernode branch 2 times, most recently from ba928be to 11bcf25 Compare February 27, 2026 22:21
Users may now define a bound on the sum of the values in
the entire `NumberNode` array.
@fastbodin fastbodin force-pushed the enhancement/axis_wise_bounds_numbernode branch from 11bcf25 to 0553d55 Compare February 27, 2026 22:43
In previous commits, we expanded the functionality of `NumberNode` such
that a user may define implicit constraints on the sum of values within
the node. We allow constraints on the entire array and along a fixed
axis. The naming convention axis-wise bounds was given before we allowed
constraints over the entire array. In the latter case, there is no
`axis`. Changing naming convention to `sum constraints`.
@fastbodin fastbodin force-pushed the enhancement/axis_wise_bounds_numbernode branch from 5002a9b to d800bc8 Compare March 6, 2026 23:15
@fastbodin fastbodin force-pushed the enhancement/axis_wise_bounds_numbernode branch from d800bc8 to 8380096 Compare March 6, 2026 23:23
@fastbodin
Copy link
Contributor Author

Changed naming convention to reflect that this features is really a summation constraint that is either applied to the entire array or to each slice along an axis.

@fastbodin fastbodin changed the title Enhancement/axis wise bounds NumberNode Enhancement/Sum constraints on NumberNode Mar 10, 2026
Naming convention change to methods, variables, and comments.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

enhancement New feature or request

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Consider adding the ability to specify one axis bound to BinaryVariable

3 participants