v0.3.7

[0.3.7] - 2026-01-25

Changed

• Convolve and DyConvolve can take a scalar argument for scales, and will return the transformed signal instead of a list in that case.
• Convolve and DyConvolve can also take a 1D signal and will return a transformed 1D signal in that case
• impulse function now 1D by default
• Misc examples updated to show this feature
• Broke up the benchmark into four separate performance tests

Added

• Improved Chebyshev approximator

v0.3.6

Changed

• Improved library accessors for Laplacians
• Extended documentation
• PDF available for documentation
• More tests for validation

Added

• SGMA modal analysis tool and examples
• Extensive examples of SGMA
• Benchmark performance tests of analytical filters
• Meshes for visualization (bunny, horse, brain)

v0.3.5

Changed

• Expanded test suite to full code and branch coverage, including all edge cases and defensive branches.
• Tests to do NOT cover KLU wrapper, because it is not used in this version.

v0.3.4

Changed

• Migrated the entire test suite from unittest to pytest for a more modern, readable, and feature-rich testing framework. The tests are now included as an installable sub-package sgwt.tests.

Fixed

• Resolved an issue where DyConvolve.addbranch would cause an unhandled error for negative branch weights. It now raises a descriptive ValueError.

v0.3.3

Added

• New documentation section for Chebyshev approximation examples and benchmarks.
• Lazy loading for built-in resources (Laplacians, Signals, Kernels) to improve import time and reduce memory usage.

Fixed

• Offset d_a of Kernel Fitted approximation correctly implemented
• Documentation typos referring to VFKern instead of VFKernel.