Python API Reference

Concise reference for the public Python API exported from stratax.

Imports

from stratax import (
    Shape,
    Vector,
    Matrix,
    Tensor,
    zeros,
    ones,
    full,
    identity,
    to_vector,
    to_matrix,
    to_tensor,
    sum,
    prod,
    max,
    min,
    argmax,
    argmin,
    mean,
    var,
    std,
)

Python containers currently expose double-based storage. Slicing, reshaping, flattening, and conversions return independent containers.

Shape

API	Description
`Shape()`	Create an empty shape.
`Shape(other)`	Copy another `Shape`.
`Shape([d0, d1, ...])`	Create a shape from dimensions.
`shape.rank`	Number of dimensions.
`shape.elements`	Product of all dimensions.
`shape.empty`	Whether the shape has rank 0.
`len(shape)`	Number of dimensions.
`shape[i]`	Dimension at index `i`.
`list(shape)`	Dimensions as a Python list.

Containers

Vector, Matrix, and Tensor share the same basic container operations.

API	Vector	Matrix	Tensor
Empty constructor	`Vector()`	`Matrix()`	`Tensor()`
Copy constructor	`Vector(v)`	`Matrix(m)`	`Tensor(t)`
Shape constructor	`Vector(Shape([n]))`	`Matrix(Shape([r, c]))`	`Tensor(Shape([...]))`
Filled constructor	`Vector(n, value)`	`Matrix(r, c, value)`	`Tensor(shape, value)`
Iterable constructor	`Vector([...])`	`Matrix([[...]])`	`Tensor([...])`
Size	`.size`	`.size`	`.size`
Rank	`.rank`	`.rank`	`.rank`
Empty check	`.empty`	`.empty`	`.empty`
Shape	`.shape`	`.shape`	`.shape`
Row-major strides	`.strides`	`.strides`	`.strides`
Fill	`.fill(value)`	`.fill(value)`	`.fill(value)`
Convert to lists	`.tolist()`	`.tolist()`	`.tolist()`

Matrix also exposes .rows and .cols.

Indexing

from stratax import Matrix, Tensor, Vector
 
v = Vector([1.0, 2.0, 3.0])
v[0] = 10.0
tail = v[1:]
 
m = Matrix([[1.0, 2.0], [3.0, 4.0]])
value = m[1, 0]
column = m[:, 0]
 
t = Tensor([2, 2, 2], 1.0)
t[1, 1, 1] = 5.0
last_axis = t[:, :, -1]

Negative indexes and slice steps are supported. Slice results are independent copies.

Arithmetic

Containers support element-wise arithmetic with matching containers or scalars:

Operation	Methods
Addition	`a + b`, `a += b`
Subtraction	`a - b`, `a -= b`
Multiplication	`a * b`, `a *= b`
Division	`a / b`, `a /= b`
Unary	`+a`, `-a`
Comparison	`a == b`, `a != b`, `a < b`, `a <= b`, `a > b`, `a >= b`

Container-to-container arithmetic uses NumPy-style trailing-dimension broadcasting. Corresponding dimensions must be equal or one of them must be 1; incompatible shapes raise BroadcastError. Comparisons are element-wise, support broadcasting and scalars, and return BoolVector, BoolMatrix, or BoolTensor.

Each comparison is also available as a module-level named function: equal, not_equal, less, less_equal, greater, and greater_equal. These functions accept arrays or scalars in either operand order.

Creation Helpers

Function	Description
`zeros(shape)` / `zeros(size)` / `zeros(rows, cols)`	Zero-filled Tensor, Vector, or Matrix.
`ones(shape)` / `ones(size)` / `ones(rows, cols)`	One-filled Tensor, Vector, or Matrix.
`full(shape, value)` / `full(size, value)` / `full(rows, cols, value)`	Constant-filled Tensor, Vector, or Matrix.
`identity(size)`	Identity Matrix with shape `[size, size]`.

Passing a Shape creates a Tensor, one integer size creates a Vector, and two integer dimensions create a Matrix.

Conversion Helpers

Function	Description
`to_vector(arr)`	Convert a compatible `Vector`, `Matrix`, or `Tensor` to `Vector`.
`to_matrix(arr)`	Convert a compatible `Vector`, `Matrix`, or `Tensor` to `Matrix`.
`to_tensor(arr)`	Convert a `Vector`, `Matrix`, or `Tensor` to `Tensor`.

Conversions return new containers.

Reductions

from stratax import mean, sum
 
total = sum(arr)
last_axis = sum(arr, -1)
kept = mean(arr, 0, keepdims=True)

Function	Result without `axis`	Result with `axis`
`sum(arr)` / `sum(arr, axis, keepdims=False)`	`float`	`Tensor`
`prod(arr)` / `prod(arr, axis, keepdims=False)`	`float`	`Tensor`
`max(arr)` / `max(arr, axis, keepdims=False)`	`float`	`Tensor`
`min(arr)` / `min(arr, axis, keepdims=False)`	`float`	`Tensor`
`argmax(arr)` / `argmax(arr, axis, keepdims=False)`	`int`	`Tensor`
`argmin(arr)` / `argmin(arr, axis, keepdims=False)`	`int`	`Tensor`
`mean(arr)` / `mean(arr, axis, keepdims=False)`	`float`	`Tensor`
`var(arr)` / `var(arr, axis, keepdims=False)`	`float`	`Tensor`
`std(arr)` / `std(arr, axis, keepdims=False)`	`float`	`Tensor`

Axes may be negative. keepdims=True preserves the reduced axis with length 1.

Exceptions

Stratax exports these exception types:

Exception	Typical use
`StrataxError`	Base exception for Stratax errors.
`ShapeError`	Invalid or incompatible shape.
`DimensionError`	Invalid dimension count.
`IndexError`	Out-of-range index.
`TypeError`	Unsupported argument type.
`BroadcastError`	Reserved for broadcasting-related shape failures.
`ZeroDivisionError`	Division by zero.