Trait Types for SciPy and NumPy

0.2.3 · maintenance · verified Sat Apr 11

traittypes provides trait types extending the `traitlets` library, specifically designed for common data structures in the SciPy ecosystem, such as NumPy arrays and SciPy sparse matrices. It allows developers to define robust type-checked attributes for these data types in `traitlets`-based classes. The current version is 0.2.3. The library appears to be in maintenance mode, with no significant development since early 2021, meaning new features or bug fixes are infrequent.

Warnings

gotcha The `traittypes` project is in maintenance mode and has not seen significant development since early 2021. Users should be aware that new features or bug fixes are unlikely to be implemented.
Fix: Consider the long-term stability and lack of ongoing support when deciding to use `traittypes` for new projects. For existing projects, be prepared to fork or implement custom solutions for any new requirements or bugs.
gotcha When assigning NumPy arrays to `Array` traits, the array object itself is stored; `traittypes` (like `traitlets`) does not create a deep copy by default. This means modifying the assigned array externally will reflect in the trait's value.
Fix: If immutability or defensive copying is required, explicitly create a copy of the array (e.g., `value.copy()`) before assigning it to the trait, or handle mutability within custom validation logic using `@validate`.
gotcha `traittypes` relies on `traitlets`' coercion mechanisms. Values assigned to traits may be implicitly converted to the expected type (e.g., a list might be converted to a NumPy array). This can sometimes hide subtle type mismatches if strict type checking is expected.
Fix: For stricter validation without implicit coercion, use custom validation methods (e.g., `@validate('my_trait')`) to explicitly check the type and content of `proposal['value']` before it's assigned.

Install

pip install traittypes Install stable version

Imports

Array
```
from traittypes import Array
```
ArrayType
```
from traittypes import ArrayType
```
ScipySparse
```
from traittypes import ScipySparse
```
ScipySparseType
```
from traittypes import ScipySparseType
```

Quickstart

This quickstart demonstrates defining a `HasTraits` class with `traittypes.Array` attributes. It shows how to assign NumPy arrays, enforce dimensionality with custom validation, and utilize the `shape` keyword argument for structural constraints.

from traitlets import HasTraits, validate, TraitError
from traittypes import Array
import numpy as np

class DataContainer(HasTraits):
    data = Array(help="A NumPy array container.")
    matrix_shape = Array(shape=(2,), help="Shape of a matrix (rows, cols).")

    @validate('data')
    def _valid_data(self, proposal):
        if proposal['value'].ndim != 1:
            raise TraitError('data must be a 1D array.')
        return proposal['value']

# Instantiate the container
container = DataContainer()

# Assign a valid array
container.data = np.array([1, 2, 3, 4])
print(f"Assigned 1D data: {container.data}")

# Attempt to assign an invalid array (will raise TraitError)
try:
    container.data = np.array([[1, 2], [3, 4]])
except TraitError as e:
    print(f"Error assigning 2D data: {e}")

# Assign a valid shape array
container.matrix_shape = np.array([10, 20])
print(f"Assigned matrix_shape: {container.matrix_shape}")

# Attempt to assign an invalid shape array (will raise TraitError due to 'shape' constraint)
try:
    container.matrix_shape = np.array([5, 6, 7])
except TraitError as e:
    print(f"Error assigning wrong shape: {e}")

view raw JSON →